UNIVERSITY OF HOHENHEIM

DOCTORAL THESIS

United we Stand Divided we Fall -Essays on Knowledge and its Diffusion

in Innovation Networks

A dissertation byKristina Bogner MSc

SupervisorProf Dr Andreas Pyka

A thesis submitted in fulfillment of the degree of doctor oeconomiae (Dr oec)

in the

Faculty of Business Economics and Social SciencesInstitute of Economics

Chair for Economics of Innovation (520I)

2019

First advisor Prof Dr Andreas PykaSecond advisor Prof Dr Bernd EbersbergerChair of the committee Prof Dr Ulrich SchwalbeDean Prof Dr Karsten HadwichDay of defense 30042019

United we Stand Divided we Fall -Essays on Knowledge and its Diffusion in Innovation NetworksKristina BognerDissertationUniversity of Hohenheim2019Copyright ccopy Kristina Bogner

AcknowledgementsFirst and foremost I want to thank my advisor and co-author Prof Dr Andreas

Pyka I want to thank you for your motivation and encouragement and everythingyou enabled Not only the conferences and summer schools my scholarship andthe flexibility in work and research but especially the possibility of doing both atthe same time being a mother and successfully finishing my thesis Studying andworking at the chair had been a motivational and inspiring experience and I amlooking forward to continue working here

I also thank Prof Dr Bernd Ebersberger for accepting to be the second re-viewer of my dissertation the University of Hohenheim (especially for introduc-ing me to neo-Schumpeterian economics) and the Landesgraduiertenfoumlrderungwhich generously funded my dissertation project This scholarship implied notonly trust in my research but gave me the freedom to entirely focus on this doc-toral thesis

This dissertation would not have been possible without the support of andthe collaboration with my colleagues and co-authors constructive criticism fromreviewers and conference participants for which I thank all of them I especiallywant to thank Matthias Muumlller Bianca Janic Sophie Urmetzer Michael SchlaileTobias Buchmann and Muhamed Kudic for being such wonderful colleagues andco-authors Matthias thank you not only for your scientific guidance encourage-ment advice and feedback but also for all the coffee breaks 11 am lunch meetingsand ABM sessions Me learning ABM and writing this thesis would not have beenpossible without you Sophie thank you for your help and support especiallywith my last paper and thank you for always having time to discuss my researchand challenging my way of thinking Micha thank you for always providing mewith the latest research and for showing me how much literature from differentdisciplines a single person can know by heart Tobi and Muhamed thank you forbeing the great colleagues and co-authors you are Last but certainly not leastthank you Bianca for being such a lovely person and the heart and the core of ourchair Working at our chair wouldnrsquot be the same without you

Finally I want to express my sincerest thanks to my whole family and myfriends Omi thank you for being the wonderful brave and funny woman youare Thank you for never giving up and showing me the important things inlife Pood thank you for always being my biggest cheerleader my compass myhiding-place Without you and without you taking care of Finn this dissertationwould not have been possible Nuumlck thank you for always helping me with La-Tex Eva and Hagen thank you for always helping and supporting us and forbeing such wonderful grand-parents to our son

Writing a doctoral thesis needs dedication and devotion And time Thankyou Jan for your love and support for the joy you bring in my life for nevercomplaining about the many many hours and nights I spent working on this the-sis Thank you Finn for being the wonderful little person you are for bringingcolour and joy in our life and for showing me how little sleep a person actuallyneeds for successfully writing a doctoral thesis

Contents

1 Introduction 211 Knowledge and its Diffusion in Innovation Networks 212 Analysing Knowledge Diffusion in Innovation Networks 713 Structure of this Thesis 8

2 The Effect of Structural Disparities on Knowledge Diffusion in Net-works An Agent-Based Simulation Model 1821 Introduction 1922 Knowledge Exchange and Network Formation Mechanisms 20

221 Theoretical Background 20222 Informal Knowledge Exchange within Networks 22223 Algorithms for the Creation of Networks 23

23 Numerical Model Analysis 25231 Path Length Cliquishness and Network Performance 25232 Degree Distribution and Network Performance 28233 Policy Experiment 32

24 Discussion and Conclusions 35

3 Knowledge Diffusion in Formal Networks The Roles of Degree Distri-bution and Cognitive Distance 4131 Introduction 4232 Modelling Knowledge Exchange and Knowledge Diffusion in For-

mal RampD Networks 43321 Network Structure Learning and Knowledge Diffusion Per-

formance in Formal RampD Networks 43322 Modelling Learning and Knowledge Exchange in Formal

RampD Networks 44323 The RampD Network in the Energy Sector An Empirical Ex-

ample 4633 Simulation Analysis 47

331 Four Theoretical Networks 47332 Model Setup and Parametrisation 49333 Knowledge Diffusion Performance in Five Different Networks 49334 Actor Degree and Performance in the RampD Network in the

Energy Sector 5334 Summary and Conclusions 56

4 Exploring the Dedicated Knowledge Base of a Transformation towards aSustainable Bioeconomy 6241 Introduction 6342 Knowledge and Innovation Policy 64

421 Towards a More Comprehensive Conceptualisation of Knowl-edge 65

iii

422 How Knowledge Concepts have Inspired Innovation PolicyMaking 67

423 Knowledge Concepts in Transformative Sustainability Science 6743 Dedicated Knowledge for an SKBBE Transformation 68

431 Systems Knowledge 68432 Normative Knowledge 70433 Transformative Knowledge 71

44 Policy-Relevant Implications 73441 Knowledge-Related Gaps in Current Bioeconomy Policies 73442 Promising (But Fragmented) Building Blocks for Improved

SKBBE Policies 7545 Conclusions 76

5 Knowledge Networks in the German Bioeconomy Network Structure ofPublicly Funded RampD Networks 9351 Introduction 9452 Knowledge and its Diffusion in RampD Networks 95

521 Knowledge Diffusion in Different Network Structures 96522 Knowledge for a Sustainable Knowledge-Based Bioeconomy 100523 On Particularities of Publicly Funded Project Networks 102

53 The RampD Network in the German Bioeconomy 104531 Subsidised RampD Projects in the German Bioeconomy in the

Past 30 Years 105532 Network Structure of the RampD Network 108

54 The RampD Network in the German Bioeconomy and its PotentialPerformance 115

55 Conclusion and Future Research Avenues 119

6 Discussion and Conclusion 12861 Summary and Discussion of Results 12862 Conclusions Limitations and Avenues for Future Research 133

iv

List of Figures

11 Structure of my doctoral thesis 9

21 Visualisation of network topologies created in NetLogo 2522 Mean average knowledge levels of agents 2623 Mean average knowledge levels of agents 2724 Average degree distribution of the agents 2925 Relationship between the variance of the degree distribution and

the mean average knowledge level 2926 Relationship between the time the agents in the network stop trad-

ing and their degree 3027 Relationship between the agentsrsquo mean knowledge levels their de-

gree and the reason why agents eventually stopped trading 3128 Relationship between agentsrsquo mean knowledge levels and the de-

gree difference between them and their partners 3329 Effect of different policy interventions on mean average knowledge

levels 34210 Effect of different policy interventions on the mean average knowl-

edge levels of the 10 smallest agents 35

31 Knowledge gain for different max cognitive distances 4632 Flow-chart of the evolutionary network algorithm 4833 Average degree distribution of the actors 5034 Mean knowledge levels and variance of the average knowledge lev-

els for different max cognitive distances 5135 Mean knowledge levels and variance of the average knowledge lev-

els at different points in time 5236 Histogram of knowledge levels 5437 Relationship between actorsrsquo mean knowledge levels and their de-

gree centralities 55

51 Type and number of actors projects and moneyyear 10852 Visualisation of the knowledge network 10953 Number of nodesedges and network density as well as average

degree 11154 Average path length and average clustering coefficients of the net-

works 11355 Degree distribution of the RampD network in the different observa-

tion periods 11456 Top-15 actors in most joint projects (lhs) and most single projects

(rhs) 126

v

List of Tables

11 Overview of agent-based models on knowledge diffusion 16

21 Average path length and global clustering coefficient (cliquishness)of the four network topologies 27

31 Standard parameter setting 4932 Network characteristics of our five networks 50

41 The elements of dedicated knowledge in the context of SKBBE poli-cies 74

51 Descriptive statistics of both joint and single research projects 10652 Network characteristics of the RampD network in different observa-

tion periods 110

vi

This doctoral thesis is dedicated to three people in particularIt is dedicated to my precious sister for always supporting

and believing in meIt is dedicated to my beloved husband for lighting up my life

and being my best friendIt is dedicated to my dearest son for proving me that I have

forces I never knew I had

vii

Chapter 1

Introduction

Chapter 1

Introduction

Knowledge is the most valuable resource of the futurerdquo (Fraunhofer IMW 2018)Researchers and policy makers alike agree upon the fact that knowledge is a cru-cial economic resource both as an input and output of innovation processes (Lund-vall and Johnson 1994 Foray and Lundvall 1998) However it is not only the inputand output of innovation processes but moreover the solution to problems (Potts2001) This resource is allowing firms to innovate and keep pace with nationaland international competitors It is helping to generate technological progresseconomic growth and prosperity Therefore understanding and managing thecreation and diffusion of knowledge within and outside of firms and innovationnetworks is of utmost importance to make full use of this precious resource

11 Knowledge and its Diffusion in Innovation Networks

How knowledge has been created and managed by entrepreneurs firms and pol-icy makers within the last decades strongly depends on the underlying under-standing and definition of knowledge In mainstream neo-classical economicsknowledge has been understood and treated as an intangible good exhibitingpublic good features as non-excludability and non-rivalry in consumption (Solow1956 1957 Arrow 1972) These (alleged) public good features of knowledge havebeen assumed to cause different problems As other actors cannot be excludedfrom the consumption of a public good new knowledge freely flows from oneactor to another causing spillover effects In this situation some actors can bene-fit from the new knowledge without having invested in its creation ie a typicalfree-riding problem emerges (Pyka et al 2009) Therefore the knowledge creat-ing actor cannot fully appropriate the returns resulting from his research activities(Arrow 1972) leading to market failure ie a situation in which the investment inRampD is below the social optimum In this mainstream neo-classical world knowl-edge falls like manna from heaven (Solow 1956 1957) Knowledge instantly flowsfrom one actor to another (at no costs) and therefore there is no need for learningThis understanding and definition of knowledge has influenced policy makingfor a long period Policies inspired by the concept of knowledge as a public goodhave mainly focused on the mitigation of potential externalities for instance byincentive creation through RampD subsidies and knowledge production by the pub-lic sector (Smith 1994 Chaminade and Esquist 2010) While policies changed toa certain extent when the understanding of knowledge changed nowadays in-novation policies as RampD subsidies (especially for the public sector) in sociallydesirable fields are still common practice (see chapter 5 on this topic)

Challenged by the fact that the mainstream neo-classical understanding lacksin giving an adequate and comprehensive definition of knowledge (evolutionaryor neo-Schumpeterian) innovation economists and management scientists tried

2

Chapter 1 Introduction

providing a much more appropriate analysis of knowledge creation and diffu-sion by considering different features of knowledge affecting its creation anddiffusion Following this understanding knowledge can instead be seen as a la-tent public good (Nelson 1989) exhibiting many non-public good features Thesefeatures have a substantial impact on how to best manage the valuable resourceknowledge Nowadays we know that knowledge is far from being a pure publicgood Knowledge is characterised by cumulativeness (Foray and Mairesse 2002Boschma 2005) relatedness (Morone and Taylor 2010) path dependency (Dosi1982 Rizzello 2004) tacitness (Polanyi 2009) stickiness (von Hippel 1994 Szulan-ski 2002) dispersion (Galunic and Rodan 1998b) context specificity and locality(Galunic and Rodan 1998b) Therefore the neo-Schumpeterian approach espe-cially emphasises the role of knowledge and learning (Hanusch and Pyka 2007Malerba 2007) In line with the neo-Schumpeterian approach anyone that hasever prepared for an exam written a doctoral thesis or even tried to cook a dishfrom a famous cook would agree upon the fact that knowledge does not fall likemanna from heaven and does not freely flow from one actor to another without theother actor actively acquiring the knowledge (learning) Instead it is the case thatthe named non-public good characteristics of knowledge substantially influencelearning and the exchange and diffusion of knowledge between actors and withina network

Inspired by communication theory how the characteristics of knowledgeinfluence its exchange and diffusion can be understood by using the famousShannon-Weaver-Model of communication (Shannon 1948 Shannon and Weaver1964) Following the idea of this model not only information transfer but alsoknowledge transfer can (at least to a certain extent) be seen as a systemic processin which knowledge is transmitted from a sender to a receiver (see eg Schwartz2005) How knowledge can be transferred from one actor to another dependsboth on the characteristics of the sender and the receiver and on the characteris-tics of knowledge itself When talking about knowledge and the impacts of itscharacteristics on knowledge exchange the questions are (i) is it possible to sendthe knowledge (ii) can the receiver re-interpret the knowledge and (iii) can theknowledge be used by the receiver

Characteristics of knowledge that influence knowledge diffusion and (i)whether it is possible to send the knowledge are tacitness stickiness and dis-persion Knowledge is not equal to information Essential parts of knowledgeare tacit ie very difficult to be codified and to be transferred (Galunic and Ro-dan 1998a) Tacit knowledge is rival and excludable and therefore not public(Polanyi 1959) Even if the sender of the knowledge is willing to share the tacit-ness makes it sometimes impossible to send this knowledge Besides knowledgeand its transfer can be sticky (von Hippel 1994 Szulanski 2002) ie the transferof this knowledge requires significantly more effort than the transfer of otherknowledge According to Szulanski (2002) both the knowledge and the processof knowledge exchange might be sticky Reasons can be the kind and amount ofknowledge itself but also attributes of the sender or receiver of knowledge Thedispersion of knowledge also influences the possibility of sending knowledgeGalunic and Rodan (1998a) explain dispersed knowledge by using the exampleof a jigsaw puzzle The authors state that knowledge is distributed if all actorsreceive a photocopy of the picture of the jigsaw puzzle At the same time theknowledge is dispersed if every actor receives one piece of the jigsaw puzzlemeaning that everybody only holds pieces of the knowledge but not the lsquowholersquopicture Dispersed knowledge (or systems-embedded knowledge) is difficult to

3

Chapter 1 Introduction

be sent as detecting dispersed knowledge can be quite problematic (Galunic andRodan 1998a)

Characteristics of knowledge and actors that influence (ii) whether the receivercan re-interpret the knowledge are the cumulative nature of knowledge or theknowledge relatedness agentsrsquo skills or cognitive distances and their absorptivecapacities New knowledge always is a (re-)combination of previous knowledge(Schumpeter 1912) The more complex and industry-specific the knowledge is thehigher the importance of the own knowledge stock and knowledge relatednessTo understand and integrate new knowledge agents need absorptive capacities(Cohen and Levinthal 1989 1990) The higher the cognitive distance between twoactors the more difficult it is for them to exchange and internalise knowledgeSo the lsquooptimalrsquo cognitive distance can be decisive for learning (Nooteboom et al2007)

Characteristics of knowledge that influence (iii) whether the knowledge canactually be used by the receiver are the context-specific or local character of knowl-edge Even if the knowledge is freely available public good features of knowledgemight not be decisive and the knowledge might be of little or no use to the receiverWe have to keep in mind that knowledge itself has no value it only becomes valu-able to someone if the knowledge can be used to eg solve specific problems(Potts 2001) Hence knowledge has different values to different actors and fol-lowing this assumption more knowledge is not always better Actors need theright knowledge in the right context and have to be able to combine this knowl-edge in the right way The valuable resource knowledge might only be relevantand of use in the narrow context for and in which it was developed (Galunic andRodan 1998a)

Making use of this model from communication theory quite impressivelyshows why an adequate comprehensive understanding and definition of knowl-edge and its characteristics is of utmost importance when analysing and manag-ing knowledge diffusion between actors and within networks With an inade-quate understanding and definition of knowledge policy makers will not be ableto manage and foster knowledge creation and diffusion This can for instance beseen by policies inspired by a somewhat linear understanding of knowledge andinnovation processes heavily supporting basic research but failing to transfer theresults into practice and producing successful innovations

It is safe to state that the improved understanding and definition of knowledgepositively influenced innovation policies (see eg study 3 in this thesis) By ap-plying a more realistic understanding of the characteristics of knowledge and howthese influence knowledge creation and diffusion researchers and policy makerswere able to shift their focus on systemic problems instead of the correction ofmarket failures (Chaminade and Esquist 2010) Understanding the importance ofnetworks and their structures for knowledge diffusion performance allows for cre-ating network structures that foster knowledge diffusion However researchersand policy makers alike so far mainly focus on the creation and diffusion of in-formation or mere techno-economic knowledge This intense focus neglects otherimportant characteristics and types of knowledge and therefore is not able to givevalid policy recommendations eg for innovation policies aiming at fosteringtransformation endeavours Especially the politically desired transformation to-wards a sustainable knowledge-based Bioeconomy (SKBBE) has been identified tobe in need of more than the production and diffusion of techno-economic knowl-edge (Urmetzer et al 2018) Inspired by sustainability literature three other typesof knowledge have been identified to be relevant for such transformations (Abson

4

Chapter 1 Introduction

et al 2014) These are systems knowledge normative knowledge and transfor-mative knowledge In chapter 4 the concept of so-called dedicated knowledge in-corporating besides mere techno-economic knowledge also systems normativeand transformative knowledge is presented

Since the effect of network structure on knowledge diffusion performance isstrongly affected by what actually diffuses analysing and incorporating differentkinds of knowledge is important Therefore the first step in this direction is donein study 3 (chapter 4) Study 3 puts special emphasis on different types of knowl-edge necessary to foster the transformation towards a sustainable knowledge-based Bioeconomy (SKBBE) It shows that innovation policies so far put no suffi-cient attention on what actually diffuses throughout innovation networks

As the collection and generation of (new) knowledge give such competi-tive advantages there is a keen interest of firms and policy makers on howto foster the creation and diffusion of (new) knowledge Firms can get accessto new knowledge either by internal knowledge generation processes as egintra-organisational learning or by access to external knowledge sources eginter-organisational cooperation (Malerba 1992) Nowadays strong focus on inter-organisational cooperation can be explained by the fact that invention activityis far from being the outcome of isolated agentsrsquo efforts but that of interactiveand collective processes (Carayol and Roux 2009 p 2) In nearly all industriesand technological areas we observe a pronounced intensity of RampD cooperationand the emergence of innovation networks with dynamically changing compo-sitions over time (Kudic 2015) At the same time we know that the economicactorsrsquo innovativeness is strongly affected by their strategic network positioningand the structural characteristics of the socio-economic environment in which theactors are embedded Networks provide a natural infrastructure for knowledgeexchange and provide the pipes and prisms of markets by enabling informationand knowledge flow (Podolny 2001)

It comes as no surprise that innovation networks and how knowledge diffu-sion within (and outside of) these networks takes place have become the centreof attention within the last decades (Valente 2006 Morone and Taylor 2010 Jack-son and Yariv 2011 Lamberson 2016) Diffusion literature in this context has beenidentified to mainly focus on four different areas or key questions

These area) What diffusesb) How does it diffusec) Where does it diffused) What are the effects (or performance) of the diffusion process and how are

they measured (see Schlaile et al 2018)In line with this the four studies presented in this thesis focus to different

extends on these four questions Study 1 and 2 focus on how different networkstructures influence knowledge diffusion performance (key question c) and d))for different diffusion mechanisms (key question b)) Study 3 focuses on whatactually diffuses within the network (key question a)) and study 4 combines find-ings from study 1 and 2 with those of study 3 namely how network structureinfluences knowledge diffusion in general and how network structure influencesknowledge diffusion of special types of knowledge in particular

These four key areas show that knowledge diffusion research can take variousforms with varying results Especially when analysing the effect of network prop-erties on knowledge diffusion performance as done in this thesis the identifiedeffects are manifold and even ambiguous This is why dependent on a) b) and c)

5

Chapter 1 Introduction

different studies identified different network characteristics and structures as fos-tering or being rsquooptimalrsquo for knowledge diffusion performance (see eg Moroneet al (2007) vs Lin and Li (2010))

While sometimes disagreeing in what actually is the best structure for diffu-sion performance agent-based simulation models within the last 15 years con-firmed that certain network characteristics and structures seem to be relativelyimportant for diffusion These are the networksrsquo average path lengths the net-worksrsquo average or global clustering coefficients the networksrsquo sizes and the net-worksrsquo degree distributions (Morone and Taylor 2004 Cowan and Jonard 20042007 Morone et al 2007 Cassi and Zirulia 2008 Kim and Park 2009 Choi et al2010 Lin and Li 2010 Zhuang et al 2011 Liu et al 2015 Luo et al 2015 Wang etal 2015 Yang et al 2015 Mueller et al 2017 Bogner et al 2018)

Concerning question a) what actually diffuses most studies focus on analysingknowledge (rather oversimplified represented as numbers or vectors) in differ-ent network structures (Morone and Taylor 2004 Cowan and Jonard 2004 2007Morone et al 2007 Cassi and Zirulia 2008 Kim and Park 2009 Lin and Li 2010Zhuang et al 2011 or study 1 and 2 in this thesis) Only a few authors investigatedifferent kinds of knowledge (Morone et al 2007 or study 3 in this thesis)

Regarding question b) how it diffuses researchers either assumed knowledgeexchange as a kind of barter trade in which agents only exchange knowledge ifthis is mutually beneficial (Cowan and Jonard 2004 2007 Morone and Taylor 2004Cassi and Zirulia 2008 Kim and Park 2009) Alternatively they assume knowledgeexchange as a gift transaction ie for some reason agents freely give away theirknowledge (Cowan and Jonard 2007 Morone et al 2007 Lin and Li 2010 Zhuanget al 2011 or study 2 and 4 in this thesis)

Concerning c) where it diffuses most studies focus on knowledge diffusionin different archetypical network structures as eg small-world structures ran-dom structures or regular structures Most of the time the investigated structuresare static (Cowan and Jonard 2004 2007 Morone et al 2007 Cassi and Zirulia2008 Kim and Park 2009 Lin and Li 2010 or study 1 and 2 in this thesis) whilesometimes researchers also investigate dynamic network structures or networkstructures evolving over time (Morone and Taylor 2004 Zhuang et al 2011 orstudy 4 in this thesis)

As the effects of network characteristics and structures on knowledge diffusionperformance depend on many different aspects researchers found ambiguous ef-fects of network characteristics and structures on diffusion Most studies measured) the effect of the diffusion process as efficiency and equity of knowledge diffu-sion ie the mean average knowledge stocks of all agents over time as well as thevariance of knowledge levels (Cowan and Jonard 2004 2007 Morone and Taylor2004 Morone et al 2007 Cassi and Zirulia 2008 Kim and Park 2009 Lin and Li2010 Zhuang et al 2011 and study 1 and 2 of this thesis) (sometimes comple-mented by an analysis of individual knowledge levels (Zhuang et al 2011 andstudy 1 and 2 of this thesis)) However d) the actual performance differs

Many researchers agree that small-world network structures (Watts and Stro-gatz 1998) are favourable (if not best) for knowledge diffusion performance(Cowan and Jonard 2004 Morone and Taylor 2004 Morone et al 2007 Cassiand Zirulia 2008 Kim and Park 2009 and study 1 of this thesis) Nonethelessthis is not always the case and depends on different aspects While Cowan andJonard found that small-world structures lead to the most efficient diffusion ofknowledge these structures at the same time lead to most unequal knowledge

6

Chapter 1 Introduction

distribution (Cowan and Jonard 2004) Morone and Taylor found that small-world networks only provide optimal patterns for diffusion if some lsquobarriers tocommunicationrsquo are removed (Morone and Taylor 2004) Cassi and Zirulia statethe small-world networks only foster diffusion performance for a certain level ofopportunity costs for using the network (Cassi and Zirulia 2008) Moreover otherstudies found that other network structures provide better patterns for knowledgediffusion in general These are for instance scale-free network structures (Lin andLi 2010) or even random network structures (Morone et al 2007) Besides manyresearchers agree that how network structures affect diffusion performance alsodepends on eg agentsrsquo absorptive capacities (Cowan and Jonard 2004) agentsrsquocognitive distances (Morone and Taylor 2004 and study 2 in this thesis) on thediffusion mechanism (Cowan and Jonard 2007 and study 1 and 2 in this thesis)on the special kind of knowledge and its relatedness (Morone et al 2007) on thecosts of using the network (Cassi and Zirulia 2008) and many more ldquoAnalysingthe structure of the network independently of the effective content of the relationcould be therefore misleadingrdquo (Cassi et al 2008 p 285) Summing up it isimpossible to make general statements and give valid policy recommendationsabout network structures lsquooptimalrsquo for knowledge diffusion

As knowledge diffusion performance is affected by so many different inter-dependent aspects Table 11 in the Appendix gives an overview of those studiesconducting experiments on knowledge diffusion used for and in my thesis Amore general comprehensive overview however goes beyond the scope of thisintroduction Such a rather general overview can for instance be found in Va-lente (2006) Morone and Taylor (2010) Jackson and Yariv (2011) or Lamberson(2016)

12 Analysing Knowledge Diffusion in Innovation Net-works

Analysing knowledge diffusion can be quite a challenging task As learningknowledge exchange and knowledge diffusion mainly take place between con-nected agents in different types of networks a method needed and used in studiesanalysing knowledge diffusion within networks is social network analysis (SNA)

Social network analysis aims at describing and exploring ldquopatterns apparentin the social relationships that individuals and groups form with each other () Itseeks to go beyond the visualisation of social relations to an examination of theirstructural properties and their implications for social actionrdquo (Scott 2017 p 2)Since knowledge diffuses throughout networks many studies analyse how theunderlying network structures or properties of agents within the network influ-ence knowledge diffusion (Ahuja 2000 Cowan and Jonard 2004 2007 Mueller etal 2017 Bogner et al 2018 to name just a few) Applying the network perspec-tive ldquoallows new leverage for answering standard social and behavioural scienceresearch questions by giving precise formal definition to aspects of the politicaleconomic or social structural environmentrdquo (Wasserman and Faust 1994 p 3)Hence the network perspective allows not only for analysing eg face-to-faceknowledge exchange between two agents but rather to focus on knowledge sys-tems and therefore to examine knowledge spread and diffusion throughout thewhole network

Due to the impossibility of actually measuring knowledge and its diffusionapproaches in diffusion literature either use empirical methods as measuring

7

Chapter 1 Introduction

patents (Ahuja 2000 Nelson 2009) or questionnaires (Reagans and McEvily 2003)Or they focus on using experiments ranging from laboratory experiments (Mo-rone et al 2007) to quasi-experiments in form of different models such as perco-lation models (Silverberg and Verspagen 2005 Bogner 2015) or other agent-basedmodels (ABMs) (Cowan and Jonard 2004 2007 Mueller et al 2017 Bogner et al2018) The studies presented in this doctoral thesis analyse knowledge diffusionthrough agent-based models (study 1 and 2) and by analysing network structuresand potential diffusion performance by deducing from theoretical considerationsin social network analysis (study 4)

The advantage of using agent-based models for analysing knowledge diffu-sion results from the fact that ABM can help to not only understand agentsrsquo orindividualsrsquo behaviours but also to understand aggregate behaviour and how thebehaviour and interaction of many individual agents lead to large-scale outcomes(Axelrod 1997) This helps understanding fundamental processes that take placeAgent-based modeling is a way of doing thought experiments Although the as-sumptions may be simple the consequences may not be at all obvious (Axelrod1997 p 4) According to Axelrod ABM [] is a third way of doing science(Axelrod 1997 p 3) in addition to deduction and induction

Many scientists argue that traditional modelling tools in economics are nolonger sufficient for analysing innovation processes especially if we want to anal-yse innovations in an increasingly complex and interdependent world (Dawid2006 Pyka and Fagiolo 2007) This is another reason why the method of agent-based modelling has become that famous Dawid (2006) explains this by thefact that ABM is able to incorporate the particularities of innovation processessuch as the unique nature of knowledge and the heterogeneity of actorsrsquo knowl-edge Agent-based simulation allows for modelling and explaining rsquotruersquo non-reversible path dependency feedbacks herding-behaviour and global phenom-ena as for instance the creation of knowledge and its diffusion (Pyka and Fagiolo2007) Therefore the method of agent-based modelling (ABM) is an approach thatbecomes more and more important especially in modelling and studying complexsystems with autonomous decision making agents that interact with each otherand with their environment (North and Macal 2007) as agents that repeatedly ex-change knowledge being arranged according to a particular network algorithm

13 Structure of this Thesis

Given the importance of understanding and even managing knowledge exchangebetween different actors this doctoral thesis aims to use methods as social net-work analysis and agent-based modelling in addition to theoretical considera-tions to explore how different network structures influence knowledge diffusionin different settings Therefore the overall research question of this doctoral the-sis is ldquoWhat are the effects of different network structures on knowledge diffusionperformance in RampD networks This research question can be subdivided in nineresearch questions covered in the four studies presented in this thesis These are

bull Covered in study 1 What are the effects of different network structures onknowledge diffusion performance in informal networks in which knowl-edge is exchanged as a barter trade How does the embeddedness of ac-tors and especially the existence of lsquostarsrsquo affect individual as well as overalldiffusion performance

8

Chapter 1 Introduction

bull Covered in study 2 What are the effects of different network structures onknowledge diffusion performance in formal RampD networks in which knowl-edge diffuses freely between the actors however is limited by the differentcognitive distances of these actors Which cognitive distance between actorsis best in the respective network structure

bull Covered in study 3 Which types and characteristics of knowledge are in-strumental creating policies fostering a transformation towards a sustain-able knowledge-based Bioeconomy (SKBBE) What are the policy-relevantimplications of this extended perspective on the characteristics of knowl-edge

bull Covered in study 4 What is the structure of the publicly funded RampD net-work in the German Bioeconomy From a theoretical point of view howdoes this network structure influence knowledge diffusion of both meretechno-economic knowledge as well as dedicated knowledge Does policymaking allow for or even create network structures that adequately supportthe creation and diffusion of dedicated knowledge in RampD networks

To answer the named research questions the structure of this doctoral thesisis as follows (see also Figure 11)

FIGURE 11 Structure of my doctoral thesis

First chapter 1 gives an introduction to the topics covered the research ques-tions answered the and methods used

In chapter 2 study 1 analyses the effect of different structural disparities onknowledge diffusion by using an agent-based simulation model It focuses onhow different network structures influence knowledge diffusion performanceThis study especially emphasises the effect of an asymmetric degree distributionon knowledge diffusion performance Study 1 complements previous research onknowledge diffusion by showing that (i) besides or even instead of the averagepath length and the average clustering coefficient the (asymmetry of) degreedistribution influences knowledge diffusion In addition (ii) especially smallinadequately embedded agents seem to be a bottleneck for knowledge diffusion

9

Chapter 1 Introduction

in this setting and iii) the identified somewhat negative network structures onthe macro level appear to result from the myopic linking strategies of the actors atthe micro level indicating a trade-off between lsquooptimalrsquo structures at the networkand the actor level

In chapter 3 study 2 uses an agent-based simulation model to analyse theeffect of different network properties on knowledge diffusion performance Asstudy 1 study 2 also focuses on the diffusion of knowledge however in contrastto study 1 this study analyses this relationship in a setting in which knowledge isdiffusing freely throughout an empirical formal RampD network as well as throughthe four benchmark networks already investigated in study 1 Besides the con-cept of cognitive distance and differences in learning between agents in the net-work are taken into account Study 2 complements study 1 and further previousresearch on knowledge diffusion by showing that (i) the (asymmetry of) degreedistribution and the distribution of links between actors in the network indeedinfluence knowledge diffusion performance to a large extent In addition (ii) theextent to which a skewed degree distribution dominates other network character-istics varies depending on the respective cognitive distances between agents

In chapter 4 study 3 analyses how so-called dedicated knowledge can contributeto the transformation towards a sustainable knowledge-based Bioeconomy Inthis study the concept of dedicated knowledge ie besides mere-techno economicknowledge also systems knowledge normative knowledge and transformativeknowledge is first introduced Moreover the characteristics of dedicated knowl-edge which are influencing knowledge diffusion performance are analysed andevaluated according to their importance and potential role for knowledge diffu-sion In addition it is analysed if and how current Bioeconomy innovation policiesaccount for dedicated knowledge This study complements previous research onknowledge and knowledge diffusion by taking a strong focus on different typesof knowledge besides techno-economic knowledge (often overemphasised in pol-icy approaches) It shows that i) different types of knowledge necessarily needto be taken into account when creating policies for knowledge creation and diffu-sion and ii) that especially systems knowledge so far has been only insufficientlyconsidered by current Bioeconomy policy approaches

In chapter 5 study 4 analyses the effect of different structural disparities onknowledge diffusion by deducing from theoretical considerations on networkstructures and diffusion performance This study focuses on how different net-work structures influence knowledge diffusion performance by analysing anempirical RampD network in the German Bioeconomy over the last 30 years bymeans of descriptive statistics and social network analysis The first part of thestudy presents the descriptive statistics of the publicly funded RampD projects andthe research conducting actors within the last 30 years The second part of thestudy shows the network characteristics and structures of the network in six dif-ferent observation periods These network statistics and structures are evaluatedfrom a knowledge diffusion point of view The study tries to answer whether theartificially generated network structures seem favourable for the diffusion of bothmere techno-economic knowledge as well as dedicated knowledge Study 4 es-pecially complements previous research on knowledge diffusion by (i) analysingan empirical network over such a long period of time (i) and (ii) by showing thateven though a network and its structure might be favourable for the diffusionof information or mere techno-economic knowledge this does not imply it alsofosters the creation and diffusion of other types of knowledge (ie dedicated

10

Chapter 1 Introduction

knowledge) necessary for the transformation towards a sustainable knowledge-based Bioeconomy (SKBBE)

In chapter 6 the main results of this doctoral thesis are summarised and dis-cussed and an outlook on possible future research avenues is given

References

Abson David J Baumgaumlrtner Stefan Fischer Joumlrn Hanspach Jan Haumlrdtle WHeinrichs H et al (2014) Ecosystem services as a boundary object for sus-tainability In Ecological Economics 103 pp 29ndash37

Ahuja Gautam (2000) Collaboration Networks Structural Holes and Innova-tion A Longitudinal Study In Administrative science quarterly

Arrow Kenneth Joseph (1972) Economic welfare and the allocation of resourcesfor invention In Readings in Industrial Economics Springer pp 219ndash236

Axelrod Robert (1997) The complexity of cooperation Agent-based models of compe-tition and collaboration Princeton University Press

Barabaacutesi Albert-Laacuteszloacute (2016) Network science Cambridge University Press

Barabaacutesi Albert-Laacuteszloacute Albert Reacuteka (1999) Emergence of Scaling in RandomNetworks In Science 286 (5439) pp 509ndash512

Bjoumlrk Jennie Magnusson Mats (2009) Where Do Good Innovation Ideas ComeFrom Exploring the Influence of Network Connectivity on Innovation IdeaQuality In Journal of Product Innovation Management 26 (6) pp 662ndash670

Bogner Kristina (2015) The Effect of Project Funding on Innovative Performance- An Agent-Based Simulation Model In Hohenheimer Discussion Papers inBusiness Economics and Social Sciences

Bogner Kristina Mueller Matthias Schlaile Michael P (2018) Knowledge dif-fusion in formal networks the roles of degree distribution and cognitivedistance In Int J Computational Economics and Econometrics pp 388ndash407

Boschma Ron A (2005) Proximity and innovation a critical assessment InRegional studies 39 (1) pp 61ndash74

Carayol Nicolas Roux Pascale (2009) Knowledge flows and the geography ofnetworks A strategic model of small world formation In Journal of Eco-nomic Behavior amp Organization 71 (2) pp 414-427

Cassi Lorenzo Corrocher Nicoletta Malerba Franco Vonortas Nicholas (2008)The impact of EU-funded research networks on knowledge diffusion at theregional level In Res Eval 17 (4) pp 283ndash293

Cassi Lorenzo Zirulia Lorenzo (2008) The opportunity cost of social relationsOn the effectiveness of small worlds In J Evol Econ 18 (1) pp 77ndash101

Chaminade Cristina Esquist Charles (2010) Rationales for public policy inter-vention in the innovation process Systems of innovation approach In Thetheory and practice of innovation policy Edward Elgar Publishing

11

Chapter 1 Introduction

Chiu Yen-Ting Helena (2008) How network competence and network locationinfluence innovation performance In Jnl of BusampIndus Marketing 24 (1) pp46ndash55

Choi Hanool Kim Sang-Hoon Lee Jeho (2010) Role of network structure andnetwork effects in diffusion of innovations In Industrial Marketing Manage-ment 39 (1) pp 170ndash177

Cohen Wesley M Levinthal Daniel A (1989) Innovation and learning the twofaces of R amp D In The economic journal 99 (397) pp 569ndash596

Cohen Wesley M Levinthal Daniel A (1990) Absorptive Capacity A NewPerspective on Learning and Innovation In Administrative science quarterly

Cowan Robin Jonard Nicolas (2004) Network structure and the diffusionof knowledge In Journal of Economic Dynamics and Control 28 (8) pp1557ndash1575

Cowan Robin Jonard Nicolas (2007) Structural holes innovation and the dis-tribution of ideas In J Econ Interac Coord 2 (2) pp 93ndash110

Dawid Herbert (2006) Agent-based models of innovation and technologicalchange In Handbook of computational economics 2 pp 1235ndash1272

Dosi Giovanni (1982) Technological paradigms and technological trajectoriesa suggested interpretation of the determinants and directions of technicalchange In Research Policy 11 (3) pp 147ndash162

Erdos Paul Reacutenyi Alfreacuted (1959) On random graphs Publicationes Mathemati-cate pp 290-297

Erdos Paul Reacutenyi Alfreacuted (1960) On the evolution of random graphs In PublMath Inst Hung Acad Sci 5 (1) pp 17ndash60

Foray Dominique Lundvall Bengt-Aringke (1998) The knowledge-based economyfrom the economics of knowledge to the learning economy In The economicimpact of knowledge pp 115ndash121

Foray Dominique Mairesse Jacques (2002) The knowledge dilemma and thegeography of innovation In Institutions and Systems in the Geography of In-novation Springer pp 35ndash54

Fraunhofer IMW (2018) Knowledge - the most valuable resource of the future EdFraunhofer Center for International Management and Knowledge EconomyIMW Available via httpswwwimwfraunhoferdeenpressinterview-annual-reporthtml (accessed on 16 April 2018)

Galunic Charles Rodan Simon (1998) Resource recombinations in the firmKnowledge structures and the potential for Schumpeterian innovation InStrat Mgmt J 19 (12) pp 1193ndash1201

Gilsing Victor Nooteboom Bart Vanhaverbeke Wim Duysters Geert van denOord Ad (2008) Network embeddedness and the exploration of novel tech-nologies Technological distance betweenness centrality and density InResearch Policy 37 (10) pp 1717ndash1731

12

First advisor Prof Dr Andreas PykaSecond advisor Prof Dr Bernd EbersbergerChair of the committee Prof Dr Ulrich SchwalbeDean Prof Dr Karsten HadwichDay of defense 30042019

United we Stand Divided we Fall -Essays on Knowledge and its Diffusion in Innovation NetworksKristina BognerDissertationUniversity of Hohenheim2019Copyright ccopy Kristina Bogner

AcknowledgementsFirst and foremost I want to thank my advisor and co-author Prof Dr Andreas

Pyka I want to thank you for your motivation and encouragement and everythingyou enabled Not only the conferences and summer schools my scholarship andthe flexibility in work and research but especially the possibility of doing both atthe same time being a mother and successfully finishing my thesis Studying andworking at the chair had been a motivational and inspiring experience and I amlooking forward to continue working here

I also thank Prof Dr Bernd Ebersberger for accepting to be the second re-viewer of my dissertation the University of Hohenheim (especially for introduc-ing me to neo-Schumpeterian economics) and the Landesgraduiertenfoumlrderungwhich generously funded my dissertation project This scholarship implied notonly trust in my research but gave me the freedom to entirely focus on this doc-toral thesis

This dissertation would not have been possible without the support of andthe collaboration with my colleagues and co-authors constructive criticism fromreviewers and conference participants for which I thank all of them I especiallywant to thank Matthias Muumlller Bianca Janic Sophie Urmetzer Michael SchlaileTobias Buchmann and Muhamed Kudic for being such wonderful colleagues andco-authors Matthias thank you not only for your scientific guidance encourage-ment advice and feedback but also for all the coffee breaks 11 am lunch meetingsand ABM sessions Me learning ABM and writing this thesis would not have beenpossible without you Sophie thank you for your help and support especiallywith my last paper and thank you for always having time to discuss my researchand challenging my way of thinking Micha thank you for always providing mewith the latest research and for showing me how much literature from differentdisciplines a single person can know by heart Tobi and Muhamed thank you forbeing the great colleagues and co-authors you are Last but certainly not leastthank you Bianca for being such a lovely person and the heart and the core of ourchair Working at our chair wouldnrsquot be the same without you

Finally I want to express my sincerest thanks to my whole family and myfriends Omi thank you for being the wonderful brave and funny woman youare Thank you for never giving up and showing me the important things inlife Pood thank you for always being my biggest cheerleader my compass myhiding-place Without you and without you taking care of Finn this dissertationwould not have been possible Nuumlck thank you for always helping me with La-Tex Eva and Hagen thank you for always helping and supporting us and forbeing such wonderful grand-parents to our son

Writing a doctoral thesis needs dedication and devotion And time Thankyou Jan for your love and support for the joy you bring in my life for nevercomplaining about the many many hours and nights I spent working on this the-sis Thank you Finn for being the wonderful little person you are for bringingcolour and joy in our life and for showing me how little sleep a person actuallyneeds for successfully writing a doctoral thesis

Contents

1 Introduction 211 Knowledge and its Diffusion in Innovation Networks 212 Analysing Knowledge Diffusion in Innovation Networks 713 Structure of this Thesis 8

2 The Effect of Structural Disparities on Knowledge Diffusion in Net-works An Agent-Based Simulation Model 1821 Introduction 1922 Knowledge Exchange and Network Formation Mechanisms 20

221 Theoretical Background 20222 Informal Knowledge Exchange within Networks 22223 Algorithms for the Creation of Networks 23

23 Numerical Model Analysis 25231 Path Length Cliquishness and Network Performance 25232 Degree Distribution and Network Performance 28233 Policy Experiment 32

24 Discussion and Conclusions 35

3 Knowledge Diffusion in Formal Networks The Roles of Degree Distri-bution and Cognitive Distance 4131 Introduction 4232 Modelling Knowledge Exchange and Knowledge Diffusion in For-

mal RampD Networks 43321 Network Structure Learning and Knowledge Diffusion Per-

formance in Formal RampD Networks 43322 Modelling Learning and Knowledge Exchange in Formal

RampD Networks 44323 The RampD Network in the Energy Sector An Empirical Ex-

ample 4633 Simulation Analysis 47

331 Four Theoretical Networks 47332 Model Setup and Parametrisation 49333 Knowledge Diffusion Performance in Five Different Networks 49334 Actor Degree and Performance in the RampD Network in the

Energy Sector 5334 Summary and Conclusions 56

4 Exploring the Dedicated Knowledge Base of a Transformation towards aSustainable Bioeconomy 6241 Introduction 6342 Knowledge and Innovation Policy 64

421 Towards a More Comprehensive Conceptualisation of Knowl-edge 65

iii

422 How Knowledge Concepts have Inspired Innovation PolicyMaking 67

423 Knowledge Concepts in Transformative Sustainability Science 6743 Dedicated Knowledge for an SKBBE Transformation 68

431 Systems Knowledge 68432 Normative Knowledge 70433 Transformative Knowledge 71

44 Policy-Relevant Implications 73441 Knowledge-Related Gaps in Current Bioeconomy Policies 73442 Promising (But Fragmented) Building Blocks for Improved

SKBBE Policies 7545 Conclusions 76

5 Knowledge Networks in the German Bioeconomy Network Structure ofPublicly Funded RampD Networks 9351 Introduction 9452 Knowledge and its Diffusion in RampD Networks 95

521 Knowledge Diffusion in Different Network Structures 96522 Knowledge for a Sustainable Knowledge-Based Bioeconomy 100523 On Particularities of Publicly Funded Project Networks 102

53 The RampD Network in the German Bioeconomy 104531 Subsidised RampD Projects in the German Bioeconomy in the

Past 30 Years 105532 Network Structure of the RampD Network 108

54 The RampD Network in the German Bioeconomy and its PotentialPerformance 115

55 Conclusion and Future Research Avenues 119

6 Discussion and Conclusion 12861 Summary and Discussion of Results 12862 Conclusions Limitations and Avenues for Future Research 133

iv

List of Figures

11 Structure of my doctoral thesis 9

21 Visualisation of network topologies created in NetLogo 2522 Mean average knowledge levels of agents 2623 Mean average knowledge levels of agents 2724 Average degree distribution of the agents 2925 Relationship between the variance of the degree distribution and

the mean average knowledge level 2926 Relationship between the time the agents in the network stop trad-

ing and their degree 3027 Relationship between the agentsrsquo mean knowledge levels their de-

gree and the reason why agents eventually stopped trading 3128 Relationship between agentsrsquo mean knowledge levels and the de-

gree difference between them and their partners 3329 Effect of different policy interventions on mean average knowledge

levels 34210 Effect of different policy interventions on the mean average knowl-

edge levels of the 10 smallest agents 35

31 Knowledge gain for different max cognitive distances 4632 Flow-chart of the evolutionary network algorithm 4833 Average degree distribution of the actors 5034 Mean knowledge levels and variance of the average knowledge lev-

els for different max cognitive distances 5135 Mean knowledge levels and variance of the average knowledge lev-

els at different points in time 5236 Histogram of knowledge levels 5437 Relationship between actorsrsquo mean knowledge levels and their de-

gree centralities 55

51 Type and number of actors projects and moneyyear 10852 Visualisation of the knowledge network 10953 Number of nodesedges and network density as well as average

degree 11154 Average path length and average clustering coefficients of the net-

works 11355 Degree distribution of the RampD network in the different observa-

tion periods 11456 Top-15 actors in most joint projects (lhs) and most single projects

(rhs) 126

v

List of Tables

11 Overview of agent-based models on knowledge diffusion 16

21 Average path length and global clustering coefficient (cliquishness)of the four network topologies 27

31 Standard parameter setting 4932 Network characteristics of our five networks 50

41 The elements of dedicated knowledge in the context of SKBBE poli-cies 74

51 Descriptive statistics of both joint and single research projects 10652 Network characteristics of the RampD network in different observa-

tion periods 110

vi

This doctoral thesis is dedicated to three people in particularIt is dedicated to my precious sister for always supporting

and believing in meIt is dedicated to my beloved husband for lighting up my life

and being my best friendIt is dedicated to my dearest son for proving me that I have

forces I never knew I had

vii

Chapter 1

Introduction

Chapter 1

Introduction

Knowledge is the most valuable resource of the futurerdquo (Fraunhofer IMW 2018)Researchers and policy makers alike agree upon the fact that knowledge is a cru-cial economic resource both as an input and output of innovation processes (Lund-vall and Johnson 1994 Foray and Lundvall 1998) However it is not only the inputand output of innovation processes but moreover the solution to problems (Potts2001) This resource is allowing firms to innovate and keep pace with nationaland international competitors It is helping to generate technological progresseconomic growth and prosperity Therefore understanding and managing thecreation and diffusion of knowledge within and outside of firms and innovationnetworks is of utmost importance to make full use of this precious resource

11 Knowledge and its Diffusion in Innovation Networks

How knowledge has been created and managed by entrepreneurs firms and pol-icy makers within the last decades strongly depends on the underlying under-standing and definition of knowledge In mainstream neo-classical economicsknowledge has been understood and treated as an intangible good exhibitingpublic good features as non-excludability and non-rivalry in consumption (Solow1956 1957 Arrow 1972) These (alleged) public good features of knowledge havebeen assumed to cause different problems As other actors cannot be excludedfrom the consumption of a public good new knowledge freely flows from oneactor to another causing spillover effects In this situation some actors can bene-fit from the new knowledge without having invested in its creation ie a typicalfree-riding problem emerges (Pyka et al 2009) Therefore the knowledge creat-ing actor cannot fully appropriate the returns resulting from his research activities(Arrow 1972) leading to market failure ie a situation in which the investment inRampD is below the social optimum In this mainstream neo-classical world knowl-edge falls like manna from heaven (Solow 1956 1957) Knowledge instantly flowsfrom one actor to another (at no costs) and therefore there is no need for learningThis understanding and definition of knowledge has influenced policy makingfor a long period Policies inspired by the concept of knowledge as a public goodhave mainly focused on the mitigation of potential externalities for instance byincentive creation through RampD subsidies and knowledge production by the pub-lic sector (Smith 1994 Chaminade and Esquist 2010) While policies changed toa certain extent when the understanding of knowledge changed nowadays in-novation policies as RampD subsidies (especially for the public sector) in sociallydesirable fields are still common practice (see chapter 5 on this topic)

Challenged by the fact that the mainstream neo-classical understanding lacksin giving an adequate and comprehensive definition of knowledge (evolutionaryor neo-Schumpeterian) innovation economists and management scientists tried

2

Chapter 1 Introduction

providing a much more appropriate analysis of knowledge creation and diffu-sion by considering different features of knowledge affecting its creation anddiffusion Following this understanding knowledge can instead be seen as a la-tent public good (Nelson 1989) exhibiting many non-public good features Thesefeatures have a substantial impact on how to best manage the valuable resourceknowledge Nowadays we know that knowledge is far from being a pure publicgood Knowledge is characterised by cumulativeness (Foray and Mairesse 2002Boschma 2005) relatedness (Morone and Taylor 2010) path dependency (Dosi1982 Rizzello 2004) tacitness (Polanyi 2009) stickiness (von Hippel 1994 Szulan-ski 2002) dispersion (Galunic and Rodan 1998b) context specificity and locality(Galunic and Rodan 1998b) Therefore the neo-Schumpeterian approach espe-cially emphasises the role of knowledge and learning (Hanusch and Pyka 2007Malerba 2007) In line with the neo-Schumpeterian approach anyone that hasever prepared for an exam written a doctoral thesis or even tried to cook a dishfrom a famous cook would agree upon the fact that knowledge does not fall likemanna from heaven and does not freely flow from one actor to another without theother actor actively acquiring the knowledge (learning) Instead it is the case thatthe named non-public good characteristics of knowledge substantially influencelearning and the exchange and diffusion of knowledge between actors and withina network

Inspired by communication theory how the characteristics of knowledgeinfluence its exchange and diffusion can be understood by using the famousShannon-Weaver-Model of communication (Shannon 1948 Shannon and Weaver1964) Following the idea of this model not only information transfer but alsoknowledge transfer can (at least to a certain extent) be seen as a systemic processin which knowledge is transmitted from a sender to a receiver (see eg Schwartz2005) How knowledge can be transferred from one actor to another dependsboth on the characteristics of the sender and the receiver and on the characteris-tics of knowledge itself When talking about knowledge and the impacts of itscharacteristics on knowledge exchange the questions are (i) is it possible to sendthe knowledge (ii) can the receiver re-interpret the knowledge and (iii) can theknowledge be used by the receiver

Characteristics of knowledge that influence knowledge diffusion and (i)whether it is possible to send the knowledge are tacitness stickiness and dis-persion Knowledge is not equal to information Essential parts of knowledgeare tacit ie very difficult to be codified and to be transferred (Galunic and Ro-dan 1998a) Tacit knowledge is rival and excludable and therefore not public(Polanyi 1959) Even if the sender of the knowledge is willing to share the tacit-ness makes it sometimes impossible to send this knowledge Besides knowledgeand its transfer can be sticky (von Hippel 1994 Szulanski 2002) ie the transferof this knowledge requires significantly more effort than the transfer of otherknowledge According to Szulanski (2002) both the knowledge and the processof knowledge exchange might be sticky Reasons can be the kind and amount ofknowledge itself but also attributes of the sender or receiver of knowledge Thedispersion of knowledge also influences the possibility of sending knowledgeGalunic and Rodan (1998a) explain dispersed knowledge by using the exampleof a jigsaw puzzle The authors state that knowledge is distributed if all actorsreceive a photocopy of the picture of the jigsaw puzzle At the same time theknowledge is dispersed if every actor receives one piece of the jigsaw puzzlemeaning that everybody only holds pieces of the knowledge but not the lsquowholersquopicture Dispersed knowledge (or systems-embedded knowledge) is difficult to

3

Chapter 1 Introduction

be sent as detecting dispersed knowledge can be quite problematic (Galunic andRodan 1998a)

Characteristics of knowledge and actors that influence (ii) whether the receivercan re-interpret the knowledge are the cumulative nature of knowledge or theknowledge relatedness agentsrsquo skills or cognitive distances and their absorptivecapacities New knowledge always is a (re-)combination of previous knowledge(Schumpeter 1912) The more complex and industry-specific the knowledge is thehigher the importance of the own knowledge stock and knowledge relatednessTo understand and integrate new knowledge agents need absorptive capacities(Cohen and Levinthal 1989 1990) The higher the cognitive distance between twoactors the more difficult it is for them to exchange and internalise knowledgeSo the lsquooptimalrsquo cognitive distance can be decisive for learning (Nooteboom et al2007)

Characteristics of knowledge that influence (iii) whether the knowledge canactually be used by the receiver are the context-specific or local character of knowl-edge Even if the knowledge is freely available public good features of knowledgemight not be decisive and the knowledge might be of little or no use to the receiverWe have to keep in mind that knowledge itself has no value it only becomes valu-able to someone if the knowledge can be used to eg solve specific problems(Potts 2001) Hence knowledge has different values to different actors and fol-lowing this assumption more knowledge is not always better Actors need theright knowledge in the right context and have to be able to combine this knowl-edge in the right way The valuable resource knowledge might only be relevantand of use in the narrow context for and in which it was developed (Galunic andRodan 1998a)

Making use of this model from communication theory quite impressivelyshows why an adequate comprehensive understanding and definition of knowl-edge and its characteristics is of utmost importance when analysing and manag-ing knowledge diffusion between actors and within networks With an inade-quate understanding and definition of knowledge policy makers will not be ableto manage and foster knowledge creation and diffusion This can for instance beseen by policies inspired by a somewhat linear understanding of knowledge andinnovation processes heavily supporting basic research but failing to transfer theresults into practice and producing successful innovations

It is safe to state that the improved understanding and definition of knowledgepositively influenced innovation policies (see eg study 3 in this thesis) By ap-plying a more realistic understanding of the characteristics of knowledge and howthese influence knowledge creation and diffusion researchers and policy makerswere able to shift their focus on systemic problems instead of the correction ofmarket failures (Chaminade and Esquist 2010) Understanding the importance ofnetworks and their structures for knowledge diffusion performance allows for cre-ating network structures that foster knowledge diffusion However researchersand policy makers alike so far mainly focus on the creation and diffusion of in-formation or mere techno-economic knowledge This intense focus neglects otherimportant characteristics and types of knowledge and therefore is not able to givevalid policy recommendations eg for innovation policies aiming at fosteringtransformation endeavours Especially the politically desired transformation to-wards a sustainable knowledge-based Bioeconomy (SKBBE) has been identified tobe in need of more than the production and diffusion of techno-economic knowl-edge (Urmetzer et al 2018) Inspired by sustainability literature three other typesof knowledge have been identified to be relevant for such transformations (Abson

4

Chapter 1 Introduction

et al 2014) These are systems knowledge normative knowledge and transfor-mative knowledge In chapter 4 the concept of so-called dedicated knowledge in-corporating besides mere techno-economic knowledge also systems normativeand transformative knowledge is presented

Since the effect of network structure on knowledge diffusion performance isstrongly affected by what actually diffuses analysing and incorporating differentkinds of knowledge is important Therefore the first step in this direction is donein study 3 (chapter 4) Study 3 puts special emphasis on different types of knowl-edge necessary to foster the transformation towards a sustainable knowledge-based Bioeconomy (SKBBE) It shows that innovation policies so far put no suffi-cient attention on what actually diffuses throughout innovation networks

As the collection and generation of (new) knowledge give such competi-tive advantages there is a keen interest of firms and policy makers on howto foster the creation and diffusion of (new) knowledge Firms can get accessto new knowledge either by internal knowledge generation processes as egintra-organisational learning or by access to external knowledge sources eginter-organisational cooperation (Malerba 1992) Nowadays strong focus on inter-organisational cooperation can be explained by the fact that invention activityis far from being the outcome of isolated agentsrsquo efforts but that of interactiveand collective processes (Carayol and Roux 2009 p 2) In nearly all industriesand technological areas we observe a pronounced intensity of RampD cooperationand the emergence of innovation networks with dynamically changing compo-sitions over time (Kudic 2015) At the same time we know that the economicactorsrsquo innovativeness is strongly affected by their strategic network positioningand the structural characteristics of the socio-economic environment in which theactors are embedded Networks provide a natural infrastructure for knowledgeexchange and provide the pipes and prisms of markets by enabling informationand knowledge flow (Podolny 2001)

It comes as no surprise that innovation networks and how knowledge diffu-sion within (and outside of) these networks takes place have become the centreof attention within the last decades (Valente 2006 Morone and Taylor 2010 Jack-son and Yariv 2011 Lamberson 2016) Diffusion literature in this context has beenidentified to mainly focus on four different areas or key questions

These area) What diffusesb) How does it diffusec) Where does it diffused) What are the effects (or performance) of the diffusion process and how are

they measured (see Schlaile et al 2018)In line with this the four studies presented in this thesis focus to different

extends on these four questions Study 1 and 2 focus on how different networkstructures influence knowledge diffusion performance (key question c) and d))for different diffusion mechanisms (key question b)) Study 3 focuses on whatactually diffuses within the network (key question a)) and study 4 combines find-ings from study 1 and 2 with those of study 3 namely how network structureinfluences knowledge diffusion in general and how network structure influencesknowledge diffusion of special types of knowledge in particular

These four key areas show that knowledge diffusion research can take variousforms with varying results Especially when analysing the effect of network prop-erties on knowledge diffusion performance as done in this thesis the identifiedeffects are manifold and even ambiguous This is why dependent on a) b) and c)

5

Chapter 1 Introduction

different studies identified different network characteristics and structures as fos-tering or being rsquooptimalrsquo for knowledge diffusion performance (see eg Moroneet al (2007) vs Lin and Li (2010))

While sometimes disagreeing in what actually is the best structure for diffu-sion performance agent-based simulation models within the last 15 years con-firmed that certain network characteristics and structures seem to be relativelyimportant for diffusion These are the networksrsquo average path lengths the net-worksrsquo average or global clustering coefficients the networksrsquo sizes and the net-worksrsquo degree distributions (Morone and Taylor 2004 Cowan and Jonard 20042007 Morone et al 2007 Cassi and Zirulia 2008 Kim and Park 2009 Choi et al2010 Lin and Li 2010 Zhuang et al 2011 Liu et al 2015 Luo et al 2015 Wang etal 2015 Yang et al 2015 Mueller et al 2017 Bogner et al 2018)

Concerning question a) what actually diffuses most studies focus on analysingknowledge (rather oversimplified represented as numbers or vectors) in differ-ent network structures (Morone and Taylor 2004 Cowan and Jonard 2004 2007Morone et al 2007 Cassi and Zirulia 2008 Kim and Park 2009 Lin and Li 2010Zhuang et al 2011 or study 1 and 2 in this thesis) Only a few authors investigatedifferent kinds of knowledge (Morone et al 2007 or study 3 in this thesis)

Regarding question b) how it diffuses researchers either assumed knowledgeexchange as a kind of barter trade in which agents only exchange knowledge ifthis is mutually beneficial (Cowan and Jonard 2004 2007 Morone and Taylor 2004Cassi and Zirulia 2008 Kim and Park 2009) Alternatively they assume knowledgeexchange as a gift transaction ie for some reason agents freely give away theirknowledge (Cowan and Jonard 2007 Morone et al 2007 Lin and Li 2010 Zhuanget al 2011 or study 2 and 4 in this thesis)

Concerning c) where it diffuses most studies focus on knowledge diffusionin different archetypical network structures as eg small-world structures ran-dom structures or regular structures Most of the time the investigated structuresare static (Cowan and Jonard 2004 2007 Morone et al 2007 Cassi and Zirulia2008 Kim and Park 2009 Lin and Li 2010 or study 1 and 2 in this thesis) whilesometimes researchers also investigate dynamic network structures or networkstructures evolving over time (Morone and Taylor 2004 Zhuang et al 2011 orstudy 4 in this thesis)

As the effects of network characteristics and structures on knowledge diffusionperformance depend on many different aspects researchers found ambiguous ef-fects of network characteristics and structures on diffusion Most studies measured) the effect of the diffusion process as efficiency and equity of knowledge diffu-sion ie the mean average knowledge stocks of all agents over time as well as thevariance of knowledge levels (Cowan and Jonard 2004 2007 Morone and Taylor2004 Morone et al 2007 Cassi and Zirulia 2008 Kim and Park 2009 Lin and Li2010 Zhuang et al 2011 and study 1 and 2 of this thesis) (sometimes comple-mented by an analysis of individual knowledge levels (Zhuang et al 2011 andstudy 1 and 2 of this thesis)) However d) the actual performance differs

Many researchers agree that small-world network structures (Watts and Stro-gatz 1998) are favourable (if not best) for knowledge diffusion performance(Cowan and Jonard 2004 Morone and Taylor 2004 Morone et al 2007 Cassiand Zirulia 2008 Kim and Park 2009 and study 1 of this thesis) Nonethelessthis is not always the case and depends on different aspects While Cowan andJonard found that small-world structures lead to the most efficient diffusion ofknowledge these structures at the same time lead to most unequal knowledge

6

Chapter 1 Introduction

distribution (Cowan and Jonard 2004) Morone and Taylor found that small-world networks only provide optimal patterns for diffusion if some lsquobarriers tocommunicationrsquo are removed (Morone and Taylor 2004) Cassi and Zirulia statethe small-world networks only foster diffusion performance for a certain level ofopportunity costs for using the network (Cassi and Zirulia 2008) Moreover otherstudies found that other network structures provide better patterns for knowledgediffusion in general These are for instance scale-free network structures (Lin andLi 2010) or even random network structures (Morone et al 2007) Besides manyresearchers agree that how network structures affect diffusion performance alsodepends on eg agentsrsquo absorptive capacities (Cowan and Jonard 2004) agentsrsquocognitive distances (Morone and Taylor 2004 and study 2 in this thesis) on thediffusion mechanism (Cowan and Jonard 2007 and study 1 and 2 in this thesis)on the special kind of knowledge and its relatedness (Morone et al 2007) on thecosts of using the network (Cassi and Zirulia 2008) and many more ldquoAnalysingthe structure of the network independently of the effective content of the relationcould be therefore misleadingrdquo (Cassi et al 2008 p 285) Summing up it isimpossible to make general statements and give valid policy recommendationsabout network structures lsquooptimalrsquo for knowledge diffusion

As knowledge diffusion performance is affected by so many different inter-dependent aspects Table 11 in the Appendix gives an overview of those studiesconducting experiments on knowledge diffusion used for and in my thesis Amore general comprehensive overview however goes beyond the scope of thisintroduction Such a rather general overview can for instance be found in Va-lente (2006) Morone and Taylor (2010) Jackson and Yariv (2011) or Lamberson(2016)

12 Analysing Knowledge Diffusion in Innovation Net-works

Analysing knowledge diffusion can be quite a challenging task As learningknowledge exchange and knowledge diffusion mainly take place between con-nected agents in different types of networks a method needed and used in studiesanalysing knowledge diffusion within networks is social network analysis (SNA)

Social network analysis aims at describing and exploring ldquopatterns apparentin the social relationships that individuals and groups form with each other () Itseeks to go beyond the visualisation of social relations to an examination of theirstructural properties and their implications for social actionrdquo (Scott 2017 p 2)Since knowledge diffuses throughout networks many studies analyse how theunderlying network structures or properties of agents within the network influ-ence knowledge diffusion (Ahuja 2000 Cowan and Jonard 2004 2007 Mueller etal 2017 Bogner et al 2018 to name just a few) Applying the network perspec-tive ldquoallows new leverage for answering standard social and behavioural scienceresearch questions by giving precise formal definition to aspects of the politicaleconomic or social structural environmentrdquo (Wasserman and Faust 1994 p 3)Hence the network perspective allows not only for analysing eg face-to-faceknowledge exchange between two agents but rather to focus on knowledge sys-tems and therefore to examine knowledge spread and diffusion throughout thewhole network

Due to the impossibility of actually measuring knowledge and its diffusionapproaches in diffusion literature either use empirical methods as measuring

7

Chapter 1 Introduction

patents (Ahuja 2000 Nelson 2009) or questionnaires (Reagans and McEvily 2003)Or they focus on using experiments ranging from laboratory experiments (Mo-rone et al 2007) to quasi-experiments in form of different models such as perco-lation models (Silverberg and Verspagen 2005 Bogner 2015) or other agent-basedmodels (ABMs) (Cowan and Jonard 2004 2007 Mueller et al 2017 Bogner et al2018) The studies presented in this doctoral thesis analyse knowledge diffusionthrough agent-based models (study 1 and 2) and by analysing network structuresand potential diffusion performance by deducing from theoretical considerationsin social network analysis (study 4)

The advantage of using agent-based models for analysing knowledge diffu-sion results from the fact that ABM can help to not only understand agentsrsquo orindividualsrsquo behaviours but also to understand aggregate behaviour and how thebehaviour and interaction of many individual agents lead to large-scale outcomes(Axelrod 1997) This helps understanding fundamental processes that take placeAgent-based modeling is a way of doing thought experiments Although the as-sumptions may be simple the consequences may not be at all obvious (Axelrod1997 p 4) According to Axelrod ABM [] is a third way of doing science(Axelrod 1997 p 3) in addition to deduction and induction

Many scientists argue that traditional modelling tools in economics are nolonger sufficient for analysing innovation processes especially if we want to anal-yse innovations in an increasingly complex and interdependent world (Dawid2006 Pyka and Fagiolo 2007) This is another reason why the method of agent-based modelling has become that famous Dawid (2006) explains this by thefact that ABM is able to incorporate the particularities of innovation processessuch as the unique nature of knowledge and the heterogeneity of actorsrsquo knowl-edge Agent-based simulation allows for modelling and explaining rsquotruersquo non-reversible path dependency feedbacks herding-behaviour and global phenom-ena as for instance the creation of knowledge and its diffusion (Pyka and Fagiolo2007) Therefore the method of agent-based modelling (ABM) is an approach thatbecomes more and more important especially in modelling and studying complexsystems with autonomous decision making agents that interact with each otherand with their environment (North and Macal 2007) as agents that repeatedly ex-change knowledge being arranged according to a particular network algorithm

13 Structure of this Thesis

Given the importance of understanding and even managing knowledge exchangebetween different actors this doctoral thesis aims to use methods as social net-work analysis and agent-based modelling in addition to theoretical considera-tions to explore how different network structures influence knowledge diffusionin different settings Therefore the overall research question of this doctoral the-sis is ldquoWhat are the effects of different network structures on knowledge diffusionperformance in RampD networks This research question can be subdivided in nineresearch questions covered in the four studies presented in this thesis These are

bull Covered in study 1 What are the effects of different network structures onknowledge diffusion performance in informal networks in which knowl-edge is exchanged as a barter trade How does the embeddedness of ac-tors and especially the existence of lsquostarsrsquo affect individual as well as overalldiffusion performance

8

Chapter 1 Introduction

bull Covered in study 2 What are the effects of different network structures onknowledge diffusion performance in formal RampD networks in which knowl-edge diffuses freely between the actors however is limited by the differentcognitive distances of these actors Which cognitive distance between actorsis best in the respective network structure

bull Covered in study 3 Which types and characteristics of knowledge are in-strumental creating policies fostering a transformation towards a sustain-able knowledge-based Bioeconomy (SKBBE) What are the policy-relevantimplications of this extended perspective on the characteristics of knowl-edge

bull Covered in study 4 What is the structure of the publicly funded RampD net-work in the German Bioeconomy From a theoretical point of view howdoes this network structure influence knowledge diffusion of both meretechno-economic knowledge as well as dedicated knowledge Does policymaking allow for or even create network structures that adequately supportthe creation and diffusion of dedicated knowledge in RampD networks

To answer the named research questions the structure of this doctoral thesisis as follows (see also Figure 11)

FIGURE 11 Structure of my doctoral thesis

First chapter 1 gives an introduction to the topics covered the research ques-tions answered the and methods used

In chapter 2 study 1 analyses the effect of different structural disparities onknowledge diffusion by using an agent-based simulation model It focuses onhow different network structures influence knowledge diffusion performanceThis study especially emphasises the effect of an asymmetric degree distributionon knowledge diffusion performance Study 1 complements previous research onknowledge diffusion by showing that (i) besides or even instead of the averagepath length and the average clustering coefficient the (asymmetry of) degreedistribution influences knowledge diffusion In addition (ii) especially smallinadequately embedded agents seem to be a bottleneck for knowledge diffusion

9

Chapter 1 Introduction

in this setting and iii) the identified somewhat negative network structures onthe macro level appear to result from the myopic linking strategies of the actors atthe micro level indicating a trade-off between lsquooptimalrsquo structures at the networkand the actor level

In chapter 3 study 2 uses an agent-based simulation model to analyse theeffect of different network properties on knowledge diffusion performance Asstudy 1 study 2 also focuses on the diffusion of knowledge however in contrastto study 1 this study analyses this relationship in a setting in which knowledge isdiffusing freely throughout an empirical formal RampD network as well as throughthe four benchmark networks already investigated in study 1 Besides the con-cept of cognitive distance and differences in learning between agents in the net-work are taken into account Study 2 complements study 1 and further previousresearch on knowledge diffusion by showing that (i) the (asymmetry of) degreedistribution and the distribution of links between actors in the network indeedinfluence knowledge diffusion performance to a large extent In addition (ii) theextent to which a skewed degree distribution dominates other network character-istics varies depending on the respective cognitive distances between agents

In chapter 4 study 3 analyses how so-called dedicated knowledge can contributeto the transformation towards a sustainable knowledge-based Bioeconomy Inthis study the concept of dedicated knowledge ie besides mere-techno economicknowledge also systems knowledge normative knowledge and transformativeknowledge is first introduced Moreover the characteristics of dedicated knowl-edge which are influencing knowledge diffusion performance are analysed andevaluated according to their importance and potential role for knowledge diffu-sion In addition it is analysed if and how current Bioeconomy innovation policiesaccount for dedicated knowledge This study complements previous research onknowledge and knowledge diffusion by taking a strong focus on different typesof knowledge besides techno-economic knowledge (often overemphasised in pol-icy approaches) It shows that i) different types of knowledge necessarily needto be taken into account when creating policies for knowledge creation and diffu-sion and ii) that especially systems knowledge so far has been only insufficientlyconsidered by current Bioeconomy policy approaches

In chapter 5 study 4 analyses the effect of different structural disparities onknowledge diffusion by deducing from theoretical considerations on networkstructures and diffusion performance This study focuses on how different net-work structures influence knowledge diffusion performance by analysing anempirical RampD network in the German Bioeconomy over the last 30 years bymeans of descriptive statistics and social network analysis The first part of thestudy presents the descriptive statistics of the publicly funded RampD projects andthe research conducting actors within the last 30 years The second part of thestudy shows the network characteristics and structures of the network in six dif-ferent observation periods These network statistics and structures are evaluatedfrom a knowledge diffusion point of view The study tries to answer whether theartificially generated network structures seem favourable for the diffusion of bothmere techno-economic knowledge as well as dedicated knowledge Study 4 es-pecially complements previous research on knowledge diffusion by (i) analysingan empirical network over such a long period of time (i) and (ii) by showing thateven though a network and its structure might be favourable for the diffusionof information or mere techno-economic knowledge this does not imply it alsofosters the creation and diffusion of other types of knowledge (ie dedicated

10

Chapter 1 Introduction

knowledge) necessary for the transformation towards a sustainable knowledge-based Bioeconomy (SKBBE)

In chapter 6 the main results of this doctoral thesis are summarised and dis-cussed and an outlook on possible future research avenues is given

References

Abson David J Baumgaumlrtner Stefan Fischer Joumlrn Hanspach Jan Haumlrdtle WHeinrichs H et al (2014) Ecosystem services as a boundary object for sus-tainability In Ecological Economics 103 pp 29ndash37

Ahuja Gautam (2000) Collaboration Networks Structural Holes and Innova-tion A Longitudinal Study In Administrative science quarterly

Arrow Kenneth Joseph (1972) Economic welfare and the allocation of resourcesfor invention In Readings in Industrial Economics Springer pp 219ndash236

Axelrod Robert (1997) The complexity of cooperation Agent-based models of compe-tition and collaboration Princeton University Press

Barabaacutesi Albert-Laacuteszloacute (2016) Network science Cambridge University Press

Barabaacutesi Albert-Laacuteszloacute Albert Reacuteka (1999) Emergence of Scaling in RandomNetworks In Science 286 (5439) pp 509ndash512

Bjoumlrk Jennie Magnusson Mats (2009) Where Do Good Innovation Ideas ComeFrom Exploring the Influence of Network Connectivity on Innovation IdeaQuality In Journal of Product Innovation Management 26 (6) pp 662ndash670

Bogner Kristina (2015) The Effect of Project Funding on Innovative Performance- An Agent-Based Simulation Model In Hohenheimer Discussion Papers inBusiness Economics and Social Sciences

Bogner Kristina Mueller Matthias Schlaile Michael P (2018) Knowledge dif-fusion in formal networks the roles of degree distribution and cognitivedistance In Int J Computational Economics and Econometrics pp 388ndash407

Boschma Ron A (2005) Proximity and innovation a critical assessment InRegional studies 39 (1) pp 61ndash74

Carayol Nicolas Roux Pascale (2009) Knowledge flows and the geography ofnetworks A strategic model of small world formation In Journal of Eco-nomic Behavior amp Organization 71 (2) pp 414-427

Cassi Lorenzo Corrocher Nicoletta Malerba Franco Vonortas Nicholas (2008)The impact of EU-funded research networks on knowledge diffusion at theregional level In Res Eval 17 (4) pp 283ndash293

Cassi Lorenzo Zirulia Lorenzo (2008) The opportunity cost of social relationsOn the effectiveness of small worlds In J Evol Econ 18 (1) pp 77ndash101

Chaminade Cristina Esquist Charles (2010) Rationales for public policy inter-vention in the innovation process Systems of innovation approach In Thetheory and practice of innovation policy Edward Elgar Publishing

11

Chapter 1 Introduction

Chiu Yen-Ting Helena (2008) How network competence and network locationinfluence innovation performance In Jnl of BusampIndus Marketing 24 (1) pp46ndash55

Choi Hanool Kim Sang-Hoon Lee Jeho (2010) Role of network structure andnetwork effects in diffusion of innovations In Industrial Marketing Manage-ment 39 (1) pp 170ndash177

Cohen Wesley M Levinthal Daniel A (1989) Innovation and learning the twofaces of R amp D In The economic journal 99 (397) pp 569ndash596

Cohen Wesley M Levinthal Daniel A (1990) Absorptive Capacity A NewPerspective on Learning and Innovation In Administrative science quarterly

Cowan Robin Jonard Nicolas (2004) Network structure and the diffusionof knowledge In Journal of Economic Dynamics and Control 28 (8) pp1557ndash1575

Cowan Robin Jonard Nicolas (2007) Structural holes innovation and the dis-tribution of ideas In J Econ Interac Coord 2 (2) pp 93ndash110

Dawid Herbert (2006) Agent-based models of innovation and technologicalchange In Handbook of computational economics 2 pp 1235ndash1272

Dosi Giovanni (1982) Technological paradigms and technological trajectoriesa suggested interpretation of the determinants and directions of technicalchange In Research Policy 11 (3) pp 147ndash162

Erdos Paul Reacutenyi Alfreacuted (1959) On random graphs Publicationes Mathemati-cate pp 290-297

Erdos Paul Reacutenyi Alfreacuted (1960) On the evolution of random graphs In PublMath Inst Hung Acad Sci 5 (1) pp 17ndash60

Foray Dominique Lundvall Bengt-Aringke (1998) The knowledge-based economyfrom the economics of knowledge to the learning economy In The economicimpact of knowledge pp 115ndash121

Foray Dominique Mairesse Jacques (2002) The knowledge dilemma and thegeography of innovation In Institutions and Systems in the Geography of In-novation Springer pp 35ndash54

Fraunhofer IMW (2018) Knowledge - the most valuable resource of the future EdFraunhofer Center for International Management and Knowledge EconomyIMW Available via httpswwwimwfraunhoferdeenpressinterview-annual-reporthtml (accessed on 16 April 2018)

Galunic Charles Rodan Simon (1998) Resource recombinations in the firmKnowledge structures and the potential for Schumpeterian innovation InStrat Mgmt J 19 (12) pp 1193ndash1201

Gilsing Victor Nooteboom Bart Vanhaverbeke Wim Duysters Geert van denOord Ad (2008) Network embeddedness and the exploration of novel tech-nologies Technological distance betweenness centrality and density InResearch Policy 37 (10) pp 1717ndash1731

12

AcknowledgementsFirst and foremost I want to thank my advisor and co-author Prof Dr Andreas

Pyka I want to thank you for your motivation and encouragement and everythingyou enabled Not only the conferences and summer schools my scholarship andthe flexibility in work and research but especially the possibility of doing both atthe same time being a mother and successfully finishing my thesis Studying andworking at the chair had been a motivational and inspiring experience and I amlooking forward to continue working here

I also thank Prof Dr Bernd Ebersberger for accepting to be the second re-viewer of my dissertation the University of Hohenheim (especially for introduc-ing me to neo-Schumpeterian economics) and the Landesgraduiertenfoumlrderungwhich generously funded my dissertation project This scholarship implied notonly trust in my research but gave me the freedom to entirely focus on this doc-toral thesis

This dissertation would not have been possible without the support of andthe collaboration with my colleagues and co-authors constructive criticism fromreviewers and conference participants for which I thank all of them I especiallywant to thank Matthias Muumlller Bianca Janic Sophie Urmetzer Michael SchlaileTobias Buchmann and Muhamed Kudic for being such wonderful colleagues andco-authors Matthias thank you not only for your scientific guidance encourage-ment advice and feedback but also for all the coffee breaks 11 am lunch meetingsand ABM sessions Me learning ABM and writing this thesis would not have beenpossible without you Sophie thank you for your help and support especiallywith my last paper and thank you for always having time to discuss my researchand challenging my way of thinking Micha thank you for always providing mewith the latest research and for showing me how much literature from differentdisciplines a single person can know by heart Tobi and Muhamed thank you forbeing the great colleagues and co-authors you are Last but certainly not leastthank you Bianca for being such a lovely person and the heart and the core of ourchair Working at our chair wouldnrsquot be the same without you

Finally I want to express my sincerest thanks to my whole family and myfriends Omi thank you for being the wonderful brave and funny woman youare Thank you for never giving up and showing me the important things inlife Pood thank you for always being my biggest cheerleader my compass myhiding-place Without you and without you taking care of Finn this dissertationwould not have been possible Nuumlck thank you for always helping me with La-Tex Eva and Hagen thank you for always helping and supporting us and forbeing such wonderful grand-parents to our son

Writing a doctoral thesis needs dedication and devotion And time Thankyou Jan for your love and support for the joy you bring in my life for nevercomplaining about the many many hours and nights I spent working on this the-sis Thank you Finn for being the wonderful little person you are for bringingcolour and joy in our life and for showing me how little sleep a person actuallyneeds for successfully writing a doctoral thesis

Contents

1 Introduction 211 Knowledge and its Diffusion in Innovation Networks 212 Analysing Knowledge Diffusion in Innovation Networks 713 Structure of this Thesis 8

2 The Effect of Structural Disparities on Knowledge Diffusion in Net-works An Agent-Based Simulation Model 1821 Introduction 1922 Knowledge Exchange and Network Formation Mechanisms 20

221 Theoretical Background 20222 Informal Knowledge Exchange within Networks 22223 Algorithms for the Creation of Networks 23

23 Numerical Model Analysis 25231 Path Length Cliquishness and Network Performance 25232 Degree Distribution and Network Performance 28233 Policy Experiment 32

24 Discussion and Conclusions 35

3 Knowledge Diffusion in Formal Networks The Roles of Degree Distri-bution and Cognitive Distance 4131 Introduction 4232 Modelling Knowledge Exchange and Knowledge Diffusion in For-

mal RampD Networks 43321 Network Structure Learning and Knowledge Diffusion Per-

formance in Formal RampD Networks 43322 Modelling Learning and Knowledge Exchange in Formal

RampD Networks 44323 The RampD Network in the Energy Sector An Empirical Ex-

ample 4633 Simulation Analysis 47

331 Four Theoretical Networks 47332 Model Setup and Parametrisation 49333 Knowledge Diffusion Performance in Five Different Networks 49334 Actor Degree and Performance in the RampD Network in the

Energy Sector 5334 Summary and Conclusions 56

4 Exploring the Dedicated Knowledge Base of a Transformation towards aSustainable Bioeconomy 6241 Introduction 6342 Knowledge and Innovation Policy 64

421 Towards a More Comprehensive Conceptualisation of Knowl-edge 65

iii

422 How Knowledge Concepts have Inspired Innovation PolicyMaking 67

423 Knowledge Concepts in Transformative Sustainability Science 6743 Dedicated Knowledge for an SKBBE Transformation 68

431 Systems Knowledge 68432 Normative Knowledge 70433 Transformative Knowledge 71

44 Policy-Relevant Implications 73441 Knowledge-Related Gaps in Current Bioeconomy Policies 73442 Promising (But Fragmented) Building Blocks for Improved

SKBBE Policies 7545 Conclusions 76

5 Knowledge Networks in the German Bioeconomy Network Structure ofPublicly Funded RampD Networks 9351 Introduction 9452 Knowledge and its Diffusion in RampD Networks 95

521 Knowledge Diffusion in Different Network Structures 96522 Knowledge for a Sustainable Knowledge-Based Bioeconomy 100523 On Particularities of Publicly Funded Project Networks 102

53 The RampD Network in the German Bioeconomy 104531 Subsidised RampD Projects in the German Bioeconomy in the

Past 30 Years 105532 Network Structure of the RampD Network 108

54 The RampD Network in the German Bioeconomy and its PotentialPerformance 115

55 Conclusion and Future Research Avenues 119

6 Discussion and Conclusion 12861 Summary and Discussion of Results 12862 Conclusions Limitations and Avenues for Future Research 133

iv

List of Figures

11 Structure of my doctoral thesis 9

21 Visualisation of network topologies created in NetLogo 2522 Mean average knowledge levels of agents 2623 Mean average knowledge levels of agents 2724 Average degree distribution of the agents 2925 Relationship between the variance of the degree distribution and

the mean average knowledge level 2926 Relationship between the time the agents in the network stop trad-

ing and their degree 3027 Relationship between the agentsrsquo mean knowledge levels their de-

gree and the reason why agents eventually stopped trading 3128 Relationship between agentsrsquo mean knowledge levels and the de-

gree difference between them and their partners 3329 Effect of different policy interventions on mean average knowledge

levels 34210 Effect of different policy interventions on the mean average knowl-

edge levels of the 10 smallest agents 35

31 Knowledge gain for different max cognitive distances 4632 Flow-chart of the evolutionary network algorithm 4833 Average degree distribution of the actors 5034 Mean knowledge levels and variance of the average knowledge lev-

els for different max cognitive distances 5135 Mean knowledge levels and variance of the average knowledge lev-

els at different points in time 5236 Histogram of knowledge levels 5437 Relationship between actorsrsquo mean knowledge levels and their de-

gree centralities 55

51 Type and number of actors projects and moneyyear 10852 Visualisation of the knowledge network 10953 Number of nodesedges and network density as well as average

degree 11154 Average path length and average clustering coefficients of the net-

works 11355 Degree distribution of the RampD network in the different observa-

tion periods 11456 Top-15 actors in most joint projects (lhs) and most single projects

(rhs) 126

v

List of Tables

11 Overview of agent-based models on knowledge diffusion 16

21 Average path length and global clustering coefficient (cliquishness)of the four network topologies 27

31 Standard parameter setting 4932 Network characteristics of our five networks 50

41 The elements of dedicated knowledge in the context of SKBBE poli-cies 74

51 Descriptive statistics of both joint and single research projects 10652 Network characteristics of the RampD network in different observa-

tion periods 110

vi

This doctoral thesis is dedicated to three people in particularIt is dedicated to my precious sister for always supporting

and believing in meIt is dedicated to my beloved husband for lighting up my life

and being my best friendIt is dedicated to my dearest son for proving me that I have

forces I never knew I had

vii

Chapter 1

Introduction

Chapter 1

Introduction

Knowledge is the most valuable resource of the futurerdquo (Fraunhofer IMW 2018)Researchers and policy makers alike agree upon the fact that knowledge is a cru-cial economic resource both as an input and output of innovation processes (Lund-vall and Johnson 1994 Foray and Lundvall 1998) However it is not only the inputand output of innovation processes but moreover the solution to problems (Potts2001) This resource is allowing firms to innovate and keep pace with nationaland international competitors It is helping to generate technological progresseconomic growth and prosperity Therefore understanding and managing thecreation and diffusion of knowledge within and outside of firms and innovationnetworks is of utmost importance to make full use of this precious resource

11 Knowledge and its Diffusion in Innovation Networks

How knowledge has been created and managed by entrepreneurs firms and pol-icy makers within the last decades strongly depends on the underlying under-standing and definition of knowledge In mainstream neo-classical economicsknowledge has been understood and treated as an intangible good exhibitingpublic good features as non-excludability and non-rivalry in consumption (Solow1956 1957 Arrow 1972) These (alleged) public good features of knowledge havebeen assumed to cause different problems As other actors cannot be excludedfrom the consumption of a public good new knowledge freely flows from oneactor to another causing spillover effects In this situation some actors can bene-fit from the new knowledge without having invested in its creation ie a typicalfree-riding problem emerges (Pyka et al 2009) Therefore the knowledge creat-ing actor cannot fully appropriate the returns resulting from his research activities(Arrow 1972) leading to market failure ie a situation in which the investment inRampD is below the social optimum In this mainstream neo-classical world knowl-edge falls like manna from heaven (Solow 1956 1957) Knowledge instantly flowsfrom one actor to another (at no costs) and therefore there is no need for learningThis understanding and definition of knowledge has influenced policy makingfor a long period Policies inspired by the concept of knowledge as a public goodhave mainly focused on the mitigation of potential externalities for instance byincentive creation through RampD subsidies and knowledge production by the pub-lic sector (Smith 1994 Chaminade and Esquist 2010) While policies changed toa certain extent when the understanding of knowledge changed nowadays in-novation policies as RampD subsidies (especially for the public sector) in sociallydesirable fields are still common practice (see chapter 5 on this topic)

Challenged by the fact that the mainstream neo-classical understanding lacksin giving an adequate and comprehensive definition of knowledge (evolutionaryor neo-Schumpeterian) innovation economists and management scientists tried

2

Chapter 1 Introduction

providing a much more appropriate analysis of knowledge creation and diffu-sion by considering different features of knowledge affecting its creation anddiffusion Following this understanding knowledge can instead be seen as a la-tent public good (Nelson 1989) exhibiting many non-public good features Thesefeatures have a substantial impact on how to best manage the valuable resourceknowledge Nowadays we know that knowledge is far from being a pure publicgood Knowledge is characterised by cumulativeness (Foray and Mairesse 2002Boschma 2005) relatedness (Morone and Taylor 2010) path dependency (Dosi1982 Rizzello 2004) tacitness (Polanyi 2009) stickiness (von Hippel 1994 Szulan-ski 2002) dispersion (Galunic and Rodan 1998b) context specificity and locality(Galunic and Rodan 1998b) Therefore the neo-Schumpeterian approach espe-cially emphasises the role of knowledge and learning (Hanusch and Pyka 2007Malerba 2007) In line with the neo-Schumpeterian approach anyone that hasever prepared for an exam written a doctoral thesis or even tried to cook a dishfrom a famous cook would agree upon the fact that knowledge does not fall likemanna from heaven and does not freely flow from one actor to another without theother actor actively acquiring the knowledge (learning) Instead it is the case thatthe named non-public good characteristics of knowledge substantially influencelearning and the exchange and diffusion of knowledge between actors and withina network

Inspired by communication theory how the characteristics of knowledgeinfluence its exchange and diffusion can be understood by using the famousShannon-Weaver-Model of communication (Shannon 1948 Shannon and Weaver1964) Following the idea of this model not only information transfer but alsoknowledge transfer can (at least to a certain extent) be seen as a systemic processin which knowledge is transmitted from a sender to a receiver (see eg Schwartz2005) How knowledge can be transferred from one actor to another dependsboth on the characteristics of the sender and the receiver and on the characteris-tics of knowledge itself When talking about knowledge and the impacts of itscharacteristics on knowledge exchange the questions are (i) is it possible to sendthe knowledge (ii) can the receiver re-interpret the knowledge and (iii) can theknowledge be used by the receiver

Characteristics of knowledge that influence knowledge diffusion and (i)whether it is possible to send the knowledge are tacitness stickiness and dis-persion Knowledge is not equal to information Essential parts of knowledgeare tacit ie very difficult to be codified and to be transferred (Galunic and Ro-dan 1998a) Tacit knowledge is rival and excludable and therefore not public(Polanyi 1959) Even if the sender of the knowledge is willing to share the tacit-ness makes it sometimes impossible to send this knowledge Besides knowledgeand its transfer can be sticky (von Hippel 1994 Szulanski 2002) ie the transferof this knowledge requires significantly more effort than the transfer of otherknowledge According to Szulanski (2002) both the knowledge and the processof knowledge exchange might be sticky Reasons can be the kind and amount ofknowledge itself but also attributes of the sender or receiver of knowledge Thedispersion of knowledge also influences the possibility of sending knowledgeGalunic and Rodan (1998a) explain dispersed knowledge by using the exampleof a jigsaw puzzle The authors state that knowledge is distributed if all actorsreceive a photocopy of the picture of the jigsaw puzzle At the same time theknowledge is dispersed if every actor receives one piece of the jigsaw puzzlemeaning that everybody only holds pieces of the knowledge but not the lsquowholersquopicture Dispersed knowledge (or systems-embedded knowledge) is difficult to

3

Chapter 1 Introduction

be sent as detecting dispersed knowledge can be quite problematic (Galunic andRodan 1998a)

Characteristics of knowledge and actors that influence (ii) whether the receivercan re-interpret the knowledge are the cumulative nature of knowledge or theknowledge relatedness agentsrsquo skills or cognitive distances and their absorptivecapacities New knowledge always is a (re-)combination of previous knowledge(Schumpeter 1912) The more complex and industry-specific the knowledge is thehigher the importance of the own knowledge stock and knowledge relatednessTo understand and integrate new knowledge agents need absorptive capacities(Cohen and Levinthal 1989 1990) The higher the cognitive distance between twoactors the more difficult it is for them to exchange and internalise knowledgeSo the lsquooptimalrsquo cognitive distance can be decisive for learning (Nooteboom et al2007)

Characteristics of knowledge that influence (iii) whether the knowledge canactually be used by the receiver are the context-specific or local character of knowl-edge Even if the knowledge is freely available public good features of knowledgemight not be decisive and the knowledge might be of little or no use to the receiverWe have to keep in mind that knowledge itself has no value it only becomes valu-able to someone if the knowledge can be used to eg solve specific problems(Potts 2001) Hence knowledge has different values to different actors and fol-lowing this assumption more knowledge is not always better Actors need theright knowledge in the right context and have to be able to combine this knowl-edge in the right way The valuable resource knowledge might only be relevantand of use in the narrow context for and in which it was developed (Galunic andRodan 1998a)

Making use of this model from communication theory quite impressivelyshows why an adequate comprehensive understanding and definition of knowl-edge and its characteristics is of utmost importance when analysing and manag-ing knowledge diffusion between actors and within networks With an inade-quate understanding and definition of knowledge policy makers will not be ableto manage and foster knowledge creation and diffusion This can for instance beseen by policies inspired by a somewhat linear understanding of knowledge andinnovation processes heavily supporting basic research but failing to transfer theresults into practice and producing successful innovations

It is safe to state that the improved understanding and definition of knowledgepositively influenced innovation policies (see eg study 3 in this thesis) By ap-plying a more realistic understanding of the characteristics of knowledge and howthese influence knowledge creation and diffusion researchers and policy makerswere able to shift their focus on systemic problems instead of the correction ofmarket failures (Chaminade and Esquist 2010) Understanding the importance ofnetworks and their structures for knowledge diffusion performance allows for cre-ating network structures that foster knowledge diffusion However researchersand policy makers alike so far mainly focus on the creation and diffusion of in-formation or mere techno-economic knowledge This intense focus neglects otherimportant characteristics and types of knowledge and therefore is not able to givevalid policy recommendations eg for innovation policies aiming at fosteringtransformation endeavours Especially the politically desired transformation to-wards a sustainable knowledge-based Bioeconomy (SKBBE) has been identified tobe in need of more than the production and diffusion of techno-economic knowl-edge (Urmetzer et al 2018) Inspired by sustainability literature three other typesof knowledge have been identified to be relevant for such transformations (Abson

4

Chapter 1 Introduction

et al 2014) These are systems knowledge normative knowledge and transfor-mative knowledge In chapter 4 the concept of so-called dedicated knowledge in-corporating besides mere techno-economic knowledge also systems normativeand transformative knowledge is presented

Since the effect of network structure on knowledge diffusion performance isstrongly affected by what actually diffuses analysing and incorporating differentkinds of knowledge is important Therefore the first step in this direction is donein study 3 (chapter 4) Study 3 puts special emphasis on different types of knowl-edge necessary to foster the transformation towards a sustainable knowledge-based Bioeconomy (SKBBE) It shows that innovation policies so far put no suffi-cient attention on what actually diffuses throughout innovation networks

As the collection and generation of (new) knowledge give such competi-tive advantages there is a keen interest of firms and policy makers on howto foster the creation and diffusion of (new) knowledge Firms can get accessto new knowledge either by internal knowledge generation processes as egintra-organisational learning or by access to external knowledge sources eginter-organisational cooperation (Malerba 1992) Nowadays strong focus on inter-organisational cooperation can be explained by the fact that invention activityis far from being the outcome of isolated agentsrsquo efforts but that of interactiveand collective processes (Carayol and Roux 2009 p 2) In nearly all industriesand technological areas we observe a pronounced intensity of RampD cooperationand the emergence of innovation networks with dynamically changing compo-sitions over time (Kudic 2015) At the same time we know that the economicactorsrsquo innovativeness is strongly affected by their strategic network positioningand the structural characteristics of the socio-economic environment in which theactors are embedded Networks provide a natural infrastructure for knowledgeexchange and provide the pipes and prisms of markets by enabling informationand knowledge flow (Podolny 2001)

It comes as no surprise that innovation networks and how knowledge diffu-sion within (and outside of) these networks takes place have become the centreof attention within the last decades (Valente 2006 Morone and Taylor 2010 Jack-son and Yariv 2011 Lamberson 2016) Diffusion literature in this context has beenidentified to mainly focus on four different areas or key questions

These area) What diffusesb) How does it diffusec) Where does it diffused) What are the effects (or performance) of the diffusion process and how are

they measured (see Schlaile et al 2018)In line with this the four studies presented in this thesis focus to different

extends on these four questions Study 1 and 2 focus on how different networkstructures influence knowledge diffusion performance (key question c) and d))for different diffusion mechanisms (key question b)) Study 3 focuses on whatactually diffuses within the network (key question a)) and study 4 combines find-ings from study 1 and 2 with those of study 3 namely how network structureinfluences knowledge diffusion in general and how network structure influencesknowledge diffusion of special types of knowledge in particular

These four key areas show that knowledge diffusion research can take variousforms with varying results Especially when analysing the effect of network prop-erties on knowledge diffusion performance as done in this thesis the identifiedeffects are manifold and even ambiguous This is why dependent on a) b) and c)

5

Chapter 1 Introduction

different studies identified different network characteristics and structures as fos-tering or being rsquooptimalrsquo for knowledge diffusion performance (see eg Moroneet al (2007) vs Lin and Li (2010))

While sometimes disagreeing in what actually is the best structure for diffu-sion performance agent-based simulation models within the last 15 years con-firmed that certain network characteristics and structures seem to be relativelyimportant for diffusion These are the networksrsquo average path lengths the net-worksrsquo average or global clustering coefficients the networksrsquo sizes and the net-worksrsquo degree distributions (Morone and Taylor 2004 Cowan and Jonard 20042007 Morone et al 2007 Cassi and Zirulia 2008 Kim and Park 2009 Choi et al2010 Lin and Li 2010 Zhuang et al 2011 Liu et al 2015 Luo et al 2015 Wang etal 2015 Yang et al 2015 Mueller et al 2017 Bogner et al 2018)

Concerning question a) what actually diffuses most studies focus on analysingknowledge (rather oversimplified represented as numbers or vectors) in differ-ent network structures (Morone and Taylor 2004 Cowan and Jonard 2004 2007Morone et al 2007 Cassi and Zirulia 2008 Kim and Park 2009 Lin and Li 2010Zhuang et al 2011 or study 1 and 2 in this thesis) Only a few authors investigatedifferent kinds of knowledge (Morone et al 2007 or study 3 in this thesis)

Regarding question b) how it diffuses researchers either assumed knowledgeexchange as a kind of barter trade in which agents only exchange knowledge ifthis is mutually beneficial (Cowan and Jonard 2004 2007 Morone and Taylor 2004Cassi and Zirulia 2008 Kim and Park 2009) Alternatively they assume knowledgeexchange as a gift transaction ie for some reason agents freely give away theirknowledge (Cowan and Jonard 2007 Morone et al 2007 Lin and Li 2010 Zhuanget al 2011 or study 2 and 4 in this thesis)

Concerning c) where it diffuses most studies focus on knowledge diffusionin different archetypical network structures as eg small-world structures ran-dom structures or regular structures Most of the time the investigated structuresare static (Cowan and Jonard 2004 2007 Morone et al 2007 Cassi and Zirulia2008 Kim and Park 2009 Lin and Li 2010 or study 1 and 2 in this thesis) whilesometimes researchers also investigate dynamic network structures or networkstructures evolving over time (Morone and Taylor 2004 Zhuang et al 2011 orstudy 4 in this thesis)

As the effects of network characteristics and structures on knowledge diffusionperformance depend on many different aspects researchers found ambiguous ef-fects of network characteristics and structures on diffusion Most studies measured) the effect of the diffusion process as efficiency and equity of knowledge diffu-sion ie the mean average knowledge stocks of all agents over time as well as thevariance of knowledge levels (Cowan and Jonard 2004 2007 Morone and Taylor2004 Morone et al 2007 Cassi and Zirulia 2008 Kim and Park 2009 Lin and Li2010 Zhuang et al 2011 and study 1 and 2 of this thesis) (sometimes comple-mented by an analysis of individual knowledge levels (Zhuang et al 2011 andstudy 1 and 2 of this thesis)) However d) the actual performance differs

Many researchers agree that small-world network structures (Watts and Stro-gatz 1998) are favourable (if not best) for knowledge diffusion performance(Cowan and Jonard 2004 Morone and Taylor 2004 Morone et al 2007 Cassiand Zirulia 2008 Kim and Park 2009 and study 1 of this thesis) Nonethelessthis is not always the case and depends on different aspects While Cowan andJonard found that small-world structures lead to the most efficient diffusion ofknowledge these structures at the same time lead to most unequal knowledge

6

Chapter 1 Introduction

distribution (Cowan and Jonard 2004) Morone and Taylor found that small-world networks only provide optimal patterns for diffusion if some lsquobarriers tocommunicationrsquo are removed (Morone and Taylor 2004) Cassi and Zirulia statethe small-world networks only foster diffusion performance for a certain level ofopportunity costs for using the network (Cassi and Zirulia 2008) Moreover otherstudies found that other network structures provide better patterns for knowledgediffusion in general These are for instance scale-free network structures (Lin andLi 2010) or even random network structures (Morone et al 2007) Besides manyresearchers agree that how network structures affect diffusion performance alsodepends on eg agentsrsquo absorptive capacities (Cowan and Jonard 2004) agentsrsquocognitive distances (Morone and Taylor 2004 and study 2 in this thesis) on thediffusion mechanism (Cowan and Jonard 2007 and study 1 and 2 in this thesis)on the special kind of knowledge and its relatedness (Morone et al 2007) on thecosts of using the network (Cassi and Zirulia 2008) and many more ldquoAnalysingthe structure of the network independently of the effective content of the relationcould be therefore misleadingrdquo (Cassi et al 2008 p 285) Summing up it isimpossible to make general statements and give valid policy recommendationsabout network structures lsquooptimalrsquo for knowledge diffusion

As knowledge diffusion performance is affected by so many different inter-dependent aspects Table 11 in the Appendix gives an overview of those studiesconducting experiments on knowledge diffusion used for and in my thesis Amore general comprehensive overview however goes beyond the scope of thisintroduction Such a rather general overview can for instance be found in Va-lente (2006) Morone and Taylor (2010) Jackson and Yariv (2011) or Lamberson(2016)

12 Analysing Knowledge Diffusion in Innovation Net-works

Analysing knowledge diffusion can be quite a challenging task As learningknowledge exchange and knowledge diffusion mainly take place between con-nected agents in different types of networks a method needed and used in studiesanalysing knowledge diffusion within networks is social network analysis (SNA)

Social network analysis aims at describing and exploring ldquopatterns apparentin the social relationships that individuals and groups form with each other () Itseeks to go beyond the visualisation of social relations to an examination of theirstructural properties and their implications for social actionrdquo (Scott 2017 p 2)Since knowledge diffuses throughout networks many studies analyse how theunderlying network structures or properties of agents within the network influ-ence knowledge diffusion (Ahuja 2000 Cowan and Jonard 2004 2007 Mueller etal 2017 Bogner et al 2018 to name just a few) Applying the network perspec-tive ldquoallows new leverage for answering standard social and behavioural scienceresearch questions by giving precise formal definition to aspects of the politicaleconomic or social structural environmentrdquo (Wasserman and Faust 1994 p 3)Hence the network perspective allows not only for analysing eg face-to-faceknowledge exchange between two agents but rather to focus on knowledge sys-tems and therefore to examine knowledge spread and diffusion throughout thewhole network

Due to the impossibility of actually measuring knowledge and its diffusionapproaches in diffusion literature either use empirical methods as measuring

7

Chapter 1 Introduction

patents (Ahuja 2000 Nelson 2009) or questionnaires (Reagans and McEvily 2003)Or they focus on using experiments ranging from laboratory experiments (Mo-rone et al 2007) to quasi-experiments in form of different models such as perco-lation models (Silverberg and Verspagen 2005 Bogner 2015) or other agent-basedmodels (ABMs) (Cowan and Jonard 2004 2007 Mueller et al 2017 Bogner et al2018) The studies presented in this doctoral thesis analyse knowledge diffusionthrough agent-based models (study 1 and 2) and by analysing network structuresand potential diffusion performance by deducing from theoretical considerationsin social network analysis (study 4)

The advantage of using agent-based models for analysing knowledge diffu-sion results from the fact that ABM can help to not only understand agentsrsquo orindividualsrsquo behaviours but also to understand aggregate behaviour and how thebehaviour and interaction of many individual agents lead to large-scale outcomes(Axelrod 1997) This helps understanding fundamental processes that take placeAgent-based modeling is a way of doing thought experiments Although the as-sumptions may be simple the consequences may not be at all obvious (Axelrod1997 p 4) According to Axelrod ABM [] is a third way of doing science(Axelrod 1997 p 3) in addition to deduction and induction

Many scientists argue that traditional modelling tools in economics are nolonger sufficient for analysing innovation processes especially if we want to anal-yse innovations in an increasingly complex and interdependent world (Dawid2006 Pyka and Fagiolo 2007) This is another reason why the method of agent-based modelling has become that famous Dawid (2006) explains this by thefact that ABM is able to incorporate the particularities of innovation processessuch as the unique nature of knowledge and the heterogeneity of actorsrsquo knowl-edge Agent-based simulation allows for modelling and explaining rsquotruersquo non-reversible path dependency feedbacks herding-behaviour and global phenom-ena as for instance the creation of knowledge and its diffusion (Pyka and Fagiolo2007) Therefore the method of agent-based modelling (ABM) is an approach thatbecomes more and more important especially in modelling and studying complexsystems with autonomous decision making agents that interact with each otherand with their environment (North and Macal 2007) as agents that repeatedly ex-change knowledge being arranged according to a particular network algorithm

13 Structure of this Thesis

Given the importance of understanding and even managing knowledge exchangebetween different actors this doctoral thesis aims to use methods as social net-work analysis and agent-based modelling in addition to theoretical considera-tions to explore how different network structures influence knowledge diffusionin different settings Therefore the overall research question of this doctoral the-sis is ldquoWhat are the effects of different network structures on knowledge diffusionperformance in RampD networks This research question can be subdivided in nineresearch questions covered in the four studies presented in this thesis These are

bull Covered in study 1 What are the effects of different network structures onknowledge diffusion performance in informal networks in which knowl-edge is exchanged as a barter trade How does the embeddedness of ac-tors and especially the existence of lsquostarsrsquo affect individual as well as overalldiffusion performance

8

Chapter 1 Introduction

bull Covered in study 2 What are the effects of different network structures onknowledge diffusion performance in formal RampD networks in which knowl-edge diffuses freely between the actors however is limited by the differentcognitive distances of these actors Which cognitive distance between actorsis best in the respective network structure

bull Covered in study 3 Which types and characteristics of knowledge are in-strumental creating policies fostering a transformation towards a sustain-able knowledge-based Bioeconomy (SKBBE) What are the policy-relevantimplications of this extended perspective on the characteristics of knowl-edge

bull Covered in study 4 What is the structure of the publicly funded RampD net-work in the German Bioeconomy From a theoretical point of view howdoes this network structure influence knowledge diffusion of both meretechno-economic knowledge as well as dedicated knowledge Does policymaking allow for or even create network structures that adequately supportthe creation and diffusion of dedicated knowledge in RampD networks

To answer the named research questions the structure of this doctoral thesisis as follows (see also Figure 11)

FIGURE 11 Structure of my doctoral thesis

First chapter 1 gives an introduction to the topics covered the research ques-tions answered the and methods used

In chapter 2 study 1 analyses the effect of different structural disparities onknowledge diffusion by using an agent-based simulation model It focuses onhow different network structures influence knowledge diffusion performanceThis study especially emphasises the effect of an asymmetric degree distributionon knowledge diffusion performance Study 1 complements previous research onknowledge diffusion by showing that (i) besides or even instead of the averagepath length and the average clustering coefficient the (asymmetry of) degreedistribution influences knowledge diffusion In addition (ii) especially smallinadequately embedded agents seem to be a bottleneck for knowledge diffusion

9

Chapter 1 Introduction

in this setting and iii) the identified somewhat negative network structures onthe macro level appear to result from the myopic linking strategies of the actors atthe micro level indicating a trade-off between lsquooptimalrsquo structures at the networkand the actor level

In chapter 3 study 2 uses an agent-based simulation model to analyse theeffect of different network properties on knowledge diffusion performance Asstudy 1 study 2 also focuses on the diffusion of knowledge however in contrastto study 1 this study analyses this relationship in a setting in which knowledge isdiffusing freely throughout an empirical formal RampD network as well as throughthe four benchmark networks already investigated in study 1 Besides the con-cept of cognitive distance and differences in learning between agents in the net-work are taken into account Study 2 complements study 1 and further previousresearch on knowledge diffusion by showing that (i) the (asymmetry of) degreedistribution and the distribution of links between actors in the network indeedinfluence knowledge diffusion performance to a large extent In addition (ii) theextent to which a skewed degree distribution dominates other network character-istics varies depending on the respective cognitive distances between agents

In chapter 4 study 3 analyses how so-called dedicated knowledge can contributeto the transformation towards a sustainable knowledge-based Bioeconomy Inthis study the concept of dedicated knowledge ie besides mere-techno economicknowledge also systems knowledge normative knowledge and transformativeknowledge is first introduced Moreover the characteristics of dedicated knowl-edge which are influencing knowledge diffusion performance are analysed andevaluated according to their importance and potential role for knowledge diffu-sion In addition it is analysed if and how current Bioeconomy innovation policiesaccount for dedicated knowledge This study complements previous research onknowledge and knowledge diffusion by taking a strong focus on different typesof knowledge besides techno-economic knowledge (often overemphasised in pol-icy approaches) It shows that i) different types of knowledge necessarily needto be taken into account when creating policies for knowledge creation and diffu-sion and ii) that especially systems knowledge so far has been only insufficientlyconsidered by current Bioeconomy policy approaches

In chapter 5 study 4 analyses the effect of different structural disparities onknowledge diffusion by deducing from theoretical considerations on networkstructures and diffusion performance This study focuses on how different net-work structures influence knowledge diffusion performance by analysing anempirical RampD network in the German Bioeconomy over the last 30 years bymeans of descriptive statistics and social network analysis The first part of thestudy presents the descriptive statistics of the publicly funded RampD projects andthe research conducting actors within the last 30 years The second part of thestudy shows the network characteristics and structures of the network in six dif-ferent observation periods These network statistics and structures are evaluatedfrom a knowledge diffusion point of view The study tries to answer whether theartificially generated network structures seem favourable for the diffusion of bothmere techno-economic knowledge as well as dedicated knowledge Study 4 es-pecially complements previous research on knowledge diffusion by (i) analysingan empirical network over such a long period of time (i) and (ii) by showing thateven though a network and its structure might be favourable for the diffusionof information or mere techno-economic knowledge this does not imply it alsofosters the creation and diffusion of other types of knowledge (ie dedicated

10

Chapter 1 Introduction

knowledge) necessary for the transformation towards a sustainable knowledge-based Bioeconomy (SKBBE)

In chapter 6 the main results of this doctoral thesis are summarised and dis-cussed and an outlook on possible future research avenues is given

References

Abson David J Baumgaumlrtner Stefan Fischer Joumlrn Hanspach Jan Haumlrdtle WHeinrichs H et al (2014) Ecosystem services as a boundary object for sus-tainability In Ecological Economics 103 pp 29ndash37

Ahuja Gautam (2000) Collaboration Networks Structural Holes and Innova-tion A Longitudinal Study In Administrative science quarterly

Arrow Kenneth Joseph (1972) Economic welfare and the allocation of resourcesfor invention In Readings in Industrial Economics Springer pp 219ndash236

Axelrod Robert (1997) The complexity of cooperation Agent-based models of compe-tition and collaboration Princeton University Press

Barabaacutesi Albert-Laacuteszloacute (2016) Network science Cambridge University Press

Barabaacutesi Albert-Laacuteszloacute Albert Reacuteka (1999) Emergence of Scaling in RandomNetworks In Science 286 (5439) pp 509ndash512

Bjoumlrk Jennie Magnusson Mats (2009) Where Do Good Innovation Ideas ComeFrom Exploring the Influence of Network Connectivity on Innovation IdeaQuality In Journal of Product Innovation Management 26 (6) pp 662ndash670

Bogner Kristina (2015) The Effect of Project Funding on Innovative Performance- An Agent-Based Simulation Model In Hohenheimer Discussion Papers inBusiness Economics and Social Sciences

Bogner Kristina Mueller Matthias Schlaile Michael P (2018) Knowledge dif-fusion in formal networks the roles of degree distribution and cognitivedistance In Int J Computational Economics and Econometrics pp 388ndash407

Boschma Ron A (2005) Proximity and innovation a critical assessment InRegional studies 39 (1) pp 61ndash74

Carayol Nicolas Roux Pascale (2009) Knowledge flows and the geography ofnetworks A strategic model of small world formation In Journal of Eco-nomic Behavior amp Organization 71 (2) pp 414-427

Cassi Lorenzo Corrocher Nicoletta Malerba Franco Vonortas Nicholas (2008)The impact of EU-funded research networks on knowledge diffusion at theregional level In Res Eval 17 (4) pp 283ndash293

Cassi Lorenzo Zirulia Lorenzo (2008) The opportunity cost of social relationsOn the effectiveness of small worlds In J Evol Econ 18 (1) pp 77ndash101

Chaminade Cristina Esquist Charles (2010) Rationales for public policy inter-vention in the innovation process Systems of innovation approach In Thetheory and practice of innovation policy Edward Elgar Publishing

11

Chapter 1 Introduction

Chiu Yen-Ting Helena (2008) How network competence and network locationinfluence innovation performance In Jnl of BusampIndus Marketing 24 (1) pp46ndash55

Choi Hanool Kim Sang-Hoon Lee Jeho (2010) Role of network structure andnetwork effects in diffusion of innovations In Industrial Marketing Manage-ment 39 (1) pp 170ndash177

Cohen Wesley M Levinthal Daniel A (1989) Innovation and learning the twofaces of R amp D In The economic journal 99 (397) pp 569ndash596

Cohen Wesley M Levinthal Daniel A (1990) Absorptive Capacity A NewPerspective on Learning and Innovation In Administrative science quarterly

Cowan Robin Jonard Nicolas (2004) Network structure and the diffusionof knowledge In Journal of Economic Dynamics and Control 28 (8) pp1557ndash1575

Cowan Robin Jonard Nicolas (2007) Structural holes innovation and the dis-tribution of ideas In J Econ Interac Coord 2 (2) pp 93ndash110

Dawid Herbert (2006) Agent-based models of innovation and technologicalchange In Handbook of computational economics 2 pp 1235ndash1272

Dosi Giovanni (1982) Technological paradigms and technological trajectoriesa suggested interpretation of the determinants and directions of technicalchange In Research Policy 11 (3) pp 147ndash162

Erdos Paul Reacutenyi Alfreacuted (1959) On random graphs Publicationes Mathemati-cate pp 290-297

Erdos Paul Reacutenyi Alfreacuted (1960) On the evolution of random graphs In PublMath Inst Hung Acad Sci 5 (1) pp 17ndash60

Foray Dominique Lundvall Bengt-Aringke (1998) The knowledge-based economyfrom the economics of knowledge to the learning economy In The economicimpact of knowledge pp 115ndash121

Foray Dominique Mairesse Jacques (2002) The knowledge dilemma and thegeography of innovation In Institutions and Systems in the Geography of In-novation Springer pp 35ndash54

Fraunhofer IMW (2018) Knowledge - the most valuable resource of the future EdFraunhofer Center for International Management and Knowledge EconomyIMW Available via httpswwwimwfraunhoferdeenpressinterview-annual-reporthtml (accessed on 16 April 2018)

Galunic Charles Rodan Simon (1998) Resource recombinations in the firmKnowledge structures and the potential for Schumpeterian innovation InStrat Mgmt J 19 (12) pp 1193ndash1201

Gilsing Victor Nooteboom Bart Vanhaverbeke Wim Duysters Geert van denOord Ad (2008) Network embeddedness and the exploration of novel tech-nologies Technological distance betweenness centrality and density InResearch Policy 37 (10) pp 1717ndash1731

12

Contents

1 Introduction 211 Knowledge and its Diffusion in Innovation Networks 212 Analysing Knowledge Diffusion in Innovation Networks 713 Structure of this Thesis 8

2 The Effect of Structural Disparities on Knowledge Diffusion in Net-works An Agent-Based Simulation Model 1821 Introduction 1922 Knowledge Exchange and Network Formation Mechanisms 20

221 Theoretical Background 20222 Informal Knowledge Exchange within Networks 22223 Algorithms for the Creation of Networks 23

23 Numerical Model Analysis 25231 Path Length Cliquishness and Network Performance 25232 Degree Distribution and Network Performance 28233 Policy Experiment 32

24 Discussion and Conclusions 35

3 Knowledge Diffusion in Formal Networks The Roles of Degree Distri-bution and Cognitive Distance 4131 Introduction 4232 Modelling Knowledge Exchange and Knowledge Diffusion in For-

mal RampD Networks 43321 Network Structure Learning and Knowledge Diffusion Per-

formance in Formal RampD Networks 43322 Modelling Learning and Knowledge Exchange in Formal

RampD Networks 44323 The RampD Network in the Energy Sector An Empirical Ex-

ample 4633 Simulation Analysis 47

331 Four Theoretical Networks 47332 Model Setup and Parametrisation 49333 Knowledge Diffusion Performance in Five Different Networks 49334 Actor Degree and Performance in the RampD Network in the

Energy Sector 5334 Summary and Conclusions 56

4 Exploring the Dedicated Knowledge Base of a Transformation towards aSustainable Bioeconomy 6241 Introduction 6342 Knowledge and Innovation Policy 64

421 Towards a More Comprehensive Conceptualisation of Knowl-edge 65

iii

422 How Knowledge Concepts have Inspired Innovation PolicyMaking 67

423 Knowledge Concepts in Transformative Sustainability Science 6743 Dedicated Knowledge for an SKBBE Transformation 68

431 Systems Knowledge 68432 Normative Knowledge 70433 Transformative Knowledge 71

44 Policy-Relevant Implications 73441 Knowledge-Related Gaps in Current Bioeconomy Policies 73442 Promising (But Fragmented) Building Blocks for Improved

SKBBE Policies 7545 Conclusions 76

5 Knowledge Networks in the German Bioeconomy Network Structure ofPublicly Funded RampD Networks 9351 Introduction 9452 Knowledge and its Diffusion in RampD Networks 95

521 Knowledge Diffusion in Different Network Structures 96522 Knowledge for a Sustainable Knowledge-Based Bioeconomy 100523 On Particularities of Publicly Funded Project Networks 102

53 The RampD Network in the German Bioeconomy 104531 Subsidised RampD Projects in the German Bioeconomy in the

Past 30 Years 105532 Network Structure of the RampD Network 108

54 The RampD Network in the German Bioeconomy and its PotentialPerformance 115

55 Conclusion and Future Research Avenues 119

6 Discussion and Conclusion 12861 Summary and Discussion of Results 12862 Conclusions Limitations and Avenues for Future Research 133

iv

List of Figures

11 Structure of my doctoral thesis 9

21 Visualisation of network topologies created in NetLogo 2522 Mean average knowledge levels of agents 2623 Mean average knowledge levels of agents 2724 Average degree distribution of the agents 2925 Relationship between the variance of the degree distribution and

the mean average knowledge level 2926 Relationship between the time the agents in the network stop trad-

ing and their degree 3027 Relationship between the agentsrsquo mean knowledge levels their de-

gree and the reason why agents eventually stopped trading 3128 Relationship between agentsrsquo mean knowledge levels and the de-

gree difference between them and their partners 3329 Effect of different policy interventions on mean average knowledge

levels 34210 Effect of different policy interventions on the mean average knowl-

edge levels of the 10 smallest agents 35

31 Knowledge gain for different max cognitive distances 4632 Flow-chart of the evolutionary network algorithm 4833 Average degree distribution of the actors 5034 Mean knowledge levels and variance of the average knowledge lev-

els for different max cognitive distances 5135 Mean knowledge levels and variance of the average knowledge lev-

els at different points in time 5236 Histogram of knowledge levels 5437 Relationship between actorsrsquo mean knowledge levels and their de-

gree centralities 55

51 Type and number of actors projects and moneyyear 10852 Visualisation of the knowledge network 10953 Number of nodesedges and network density as well as average

degree 11154 Average path length and average clustering coefficients of the net-

works 11355 Degree distribution of the RampD network in the different observa-

tion periods 11456 Top-15 actors in most joint projects (lhs) and most single projects

(rhs) 126

v

List of Tables

11 Overview of agent-based models on knowledge diffusion 16

21 Average path length and global clustering coefficient (cliquishness)of the four network topologies 27

31 Standard parameter setting 4932 Network characteristics of our five networks 50

41 The elements of dedicated knowledge in the context of SKBBE poli-cies 74

51 Descriptive statistics of both joint and single research projects 10652 Network characteristics of the RampD network in different observa-

tion periods 110

vi

This doctoral thesis is dedicated to three people in particularIt is dedicated to my precious sister for always supporting

and believing in meIt is dedicated to my beloved husband for lighting up my life

and being my best friendIt is dedicated to my dearest son for proving me that I have

forces I never knew I had

vii

Chapter 1

Introduction

Chapter 1

Introduction

Knowledge is the most valuable resource of the futurerdquo (Fraunhofer IMW 2018)Researchers and policy makers alike agree upon the fact that knowledge is a cru-cial economic resource both as an input and output of innovation processes (Lund-vall and Johnson 1994 Foray and Lundvall 1998) However it is not only the inputand output of innovation processes but moreover the solution to problems (Potts2001) This resource is allowing firms to innovate and keep pace with nationaland international competitors It is helping to generate technological progresseconomic growth and prosperity Therefore understanding and managing thecreation and diffusion of knowledge within and outside of firms and innovationnetworks is of utmost importance to make full use of this precious resource

11 Knowledge and its Diffusion in Innovation Networks

How knowledge has been created and managed by entrepreneurs firms and pol-icy makers within the last decades strongly depends on the underlying under-standing and definition of knowledge In mainstream neo-classical economicsknowledge has been understood and treated as an intangible good exhibitingpublic good features as non-excludability and non-rivalry in consumption (Solow1956 1957 Arrow 1972) These (alleged) public good features of knowledge havebeen assumed to cause different problems As other actors cannot be excludedfrom the consumption of a public good new knowledge freely flows from oneactor to another causing spillover effects In this situation some actors can bene-fit from the new knowledge without having invested in its creation ie a typicalfree-riding problem emerges (Pyka et al 2009) Therefore the knowledge creat-ing actor cannot fully appropriate the returns resulting from his research activities(Arrow 1972) leading to market failure ie a situation in which the investment inRampD is below the social optimum In this mainstream neo-classical world knowl-edge falls like manna from heaven (Solow 1956 1957) Knowledge instantly flowsfrom one actor to another (at no costs) and therefore there is no need for learningThis understanding and definition of knowledge has influenced policy makingfor a long period Policies inspired by the concept of knowledge as a public goodhave mainly focused on the mitigation of potential externalities for instance byincentive creation through RampD subsidies and knowledge production by the pub-lic sector (Smith 1994 Chaminade and Esquist 2010) While policies changed toa certain extent when the understanding of knowledge changed nowadays in-novation policies as RampD subsidies (especially for the public sector) in sociallydesirable fields are still common practice (see chapter 5 on this topic)

Challenged by the fact that the mainstream neo-classical understanding lacksin giving an adequate and comprehensive definition of knowledge (evolutionaryor neo-Schumpeterian) innovation economists and management scientists tried

2

Chapter 1 Introduction

providing a much more appropriate analysis of knowledge creation and diffu-sion by considering different features of knowledge affecting its creation anddiffusion Following this understanding knowledge can instead be seen as a la-tent public good (Nelson 1989) exhibiting many non-public good features Thesefeatures have a substantial impact on how to best manage the valuable resourceknowledge Nowadays we know that knowledge is far from being a pure publicgood Knowledge is characterised by cumulativeness (Foray and Mairesse 2002Boschma 2005) relatedness (Morone and Taylor 2010) path dependency (Dosi1982 Rizzello 2004) tacitness (Polanyi 2009) stickiness (von Hippel 1994 Szulan-ski 2002) dispersion (Galunic and Rodan 1998b) context specificity and locality(Galunic and Rodan 1998b) Therefore the neo-Schumpeterian approach espe-cially emphasises the role of knowledge and learning (Hanusch and Pyka 2007Malerba 2007) In line with the neo-Schumpeterian approach anyone that hasever prepared for an exam written a doctoral thesis or even tried to cook a dishfrom a famous cook would agree upon the fact that knowledge does not fall likemanna from heaven and does not freely flow from one actor to another without theother actor actively acquiring the knowledge (learning) Instead it is the case thatthe named non-public good characteristics of knowledge substantially influencelearning and the exchange and diffusion of knowledge between actors and withina network

Inspired by communication theory how the characteristics of knowledgeinfluence its exchange and diffusion can be understood by using the famousShannon-Weaver-Model of communication (Shannon 1948 Shannon and Weaver1964) Following the idea of this model not only information transfer but alsoknowledge transfer can (at least to a certain extent) be seen as a systemic processin which knowledge is transmitted from a sender to a receiver (see eg Schwartz2005) How knowledge can be transferred from one actor to another dependsboth on the characteristics of the sender and the receiver and on the characteris-tics of knowledge itself When talking about knowledge and the impacts of itscharacteristics on knowledge exchange the questions are (i) is it possible to sendthe knowledge (ii) can the receiver re-interpret the knowledge and (iii) can theknowledge be used by the receiver

Characteristics of knowledge that influence knowledge diffusion and (i)whether it is possible to send the knowledge are tacitness stickiness and dis-persion Knowledge is not equal to information Essential parts of knowledgeare tacit ie very difficult to be codified and to be transferred (Galunic and Ro-dan 1998a) Tacit knowledge is rival and excludable and therefore not public(Polanyi 1959) Even if the sender of the knowledge is willing to share the tacit-ness makes it sometimes impossible to send this knowledge Besides knowledgeand its transfer can be sticky (von Hippel 1994 Szulanski 2002) ie the transferof this knowledge requires significantly more effort than the transfer of otherknowledge According to Szulanski (2002) both the knowledge and the processof knowledge exchange might be sticky Reasons can be the kind and amount ofknowledge itself but also attributes of the sender or receiver of knowledge Thedispersion of knowledge also influences the possibility of sending knowledgeGalunic and Rodan (1998a) explain dispersed knowledge by using the exampleof a jigsaw puzzle The authors state that knowledge is distributed if all actorsreceive a photocopy of the picture of the jigsaw puzzle At the same time theknowledge is dispersed if every actor receives one piece of the jigsaw puzzlemeaning that everybody only holds pieces of the knowledge but not the lsquowholersquopicture Dispersed knowledge (or systems-embedded knowledge) is difficult to

3

Chapter 1 Introduction

be sent as detecting dispersed knowledge can be quite problematic (Galunic andRodan 1998a)

Characteristics of knowledge and actors that influence (ii) whether the receivercan re-interpret the knowledge are the cumulative nature of knowledge or theknowledge relatedness agentsrsquo skills or cognitive distances and their absorptivecapacities New knowledge always is a (re-)combination of previous knowledge(Schumpeter 1912) The more complex and industry-specific the knowledge is thehigher the importance of the own knowledge stock and knowledge relatednessTo understand and integrate new knowledge agents need absorptive capacities(Cohen and Levinthal 1989 1990) The higher the cognitive distance between twoactors the more difficult it is for them to exchange and internalise knowledgeSo the lsquooptimalrsquo cognitive distance can be decisive for learning (Nooteboom et al2007)

Characteristics of knowledge that influence (iii) whether the knowledge canactually be used by the receiver are the context-specific or local character of knowl-edge Even if the knowledge is freely available public good features of knowledgemight not be decisive and the knowledge might be of little or no use to the receiverWe have to keep in mind that knowledge itself has no value it only becomes valu-able to someone if the knowledge can be used to eg solve specific problems(Potts 2001) Hence knowledge has different values to different actors and fol-lowing this assumption more knowledge is not always better Actors need theright knowledge in the right context and have to be able to combine this knowl-edge in the right way The valuable resource knowledge might only be relevantand of use in the narrow context for and in which it was developed (Galunic andRodan 1998a)

Making use of this model from communication theory quite impressivelyshows why an adequate comprehensive understanding and definition of knowl-edge and its characteristics is of utmost importance when analysing and manag-ing knowledge diffusion between actors and within networks With an inade-quate understanding and definition of knowledge policy makers will not be ableto manage and foster knowledge creation and diffusion This can for instance beseen by policies inspired by a somewhat linear understanding of knowledge andinnovation processes heavily supporting basic research but failing to transfer theresults into practice and producing successful innovations

It is safe to state that the improved understanding and definition of knowledgepositively influenced innovation policies (see eg study 3 in this thesis) By ap-plying a more realistic understanding of the characteristics of knowledge and howthese influence knowledge creation and diffusion researchers and policy makerswere able to shift their focus on systemic problems instead of the correction ofmarket failures (Chaminade and Esquist 2010) Understanding the importance ofnetworks and their structures for knowledge diffusion performance allows for cre-ating network structures that foster knowledge diffusion However researchersand policy makers alike so far mainly focus on the creation and diffusion of in-formation or mere techno-economic knowledge This intense focus neglects otherimportant characteristics and types of knowledge and therefore is not able to givevalid policy recommendations eg for innovation policies aiming at fosteringtransformation endeavours Especially the politically desired transformation to-wards a sustainable knowledge-based Bioeconomy (SKBBE) has been identified tobe in need of more than the production and diffusion of techno-economic knowl-edge (Urmetzer et al 2018) Inspired by sustainability literature three other typesof knowledge have been identified to be relevant for such transformations (Abson

4

Chapter 1 Introduction

et al 2014) These are systems knowledge normative knowledge and transfor-mative knowledge In chapter 4 the concept of so-called dedicated knowledge in-corporating besides mere techno-economic knowledge also systems normativeand transformative knowledge is presented

Since the effect of network structure on knowledge diffusion performance isstrongly affected by what actually diffuses analysing and incorporating differentkinds of knowledge is important Therefore the first step in this direction is donein study 3 (chapter 4) Study 3 puts special emphasis on different types of knowl-edge necessary to foster the transformation towards a sustainable knowledge-based Bioeconomy (SKBBE) It shows that innovation policies so far put no suffi-cient attention on what actually diffuses throughout innovation networks

As the collection and generation of (new) knowledge give such competi-tive advantages there is a keen interest of firms and policy makers on howto foster the creation and diffusion of (new) knowledge Firms can get accessto new knowledge either by internal knowledge generation processes as egintra-organisational learning or by access to external knowledge sources eginter-organisational cooperation (Malerba 1992) Nowadays strong focus on inter-organisational cooperation can be explained by the fact that invention activityis far from being the outcome of isolated agentsrsquo efforts but that of interactiveand collective processes (Carayol and Roux 2009 p 2) In nearly all industriesand technological areas we observe a pronounced intensity of RampD cooperationand the emergence of innovation networks with dynamically changing compo-sitions over time (Kudic 2015) At the same time we know that the economicactorsrsquo innovativeness is strongly affected by their strategic network positioningand the structural characteristics of the socio-economic environment in which theactors are embedded Networks provide a natural infrastructure for knowledgeexchange and provide the pipes and prisms of markets by enabling informationand knowledge flow (Podolny 2001)

It comes as no surprise that innovation networks and how knowledge diffu-sion within (and outside of) these networks takes place have become the centreof attention within the last decades (Valente 2006 Morone and Taylor 2010 Jack-son and Yariv 2011 Lamberson 2016) Diffusion literature in this context has beenidentified to mainly focus on four different areas or key questions

These area) What diffusesb) How does it diffusec) Where does it diffused) What are the effects (or performance) of the diffusion process and how are

they measured (see Schlaile et al 2018)In line with this the four studies presented in this thesis focus to different

extends on these four questions Study 1 and 2 focus on how different networkstructures influence knowledge diffusion performance (key question c) and d))for different diffusion mechanisms (key question b)) Study 3 focuses on whatactually diffuses within the network (key question a)) and study 4 combines find-ings from study 1 and 2 with those of study 3 namely how network structureinfluences knowledge diffusion in general and how network structure influencesknowledge diffusion of special types of knowledge in particular

These four key areas show that knowledge diffusion research can take variousforms with varying results Especially when analysing the effect of network prop-erties on knowledge diffusion performance as done in this thesis the identifiedeffects are manifold and even ambiguous This is why dependent on a) b) and c)

5

Chapter 1 Introduction

different studies identified different network characteristics and structures as fos-tering or being rsquooptimalrsquo for knowledge diffusion performance (see eg Moroneet al (2007) vs Lin and Li (2010))

While sometimes disagreeing in what actually is the best structure for diffu-sion performance agent-based simulation models within the last 15 years con-firmed that certain network characteristics and structures seem to be relativelyimportant for diffusion These are the networksrsquo average path lengths the net-worksrsquo average or global clustering coefficients the networksrsquo sizes and the net-worksrsquo degree distributions (Morone and Taylor 2004 Cowan and Jonard 20042007 Morone et al 2007 Cassi and Zirulia 2008 Kim and Park 2009 Choi et al2010 Lin and Li 2010 Zhuang et al 2011 Liu et al 2015 Luo et al 2015 Wang etal 2015 Yang et al 2015 Mueller et al 2017 Bogner et al 2018)

Concerning question a) what actually diffuses most studies focus on analysingknowledge (rather oversimplified represented as numbers or vectors) in differ-ent network structures (Morone and Taylor 2004 Cowan and Jonard 2004 2007Morone et al 2007 Cassi and Zirulia 2008 Kim and Park 2009 Lin and Li 2010Zhuang et al 2011 or study 1 and 2 in this thesis) Only a few authors investigatedifferent kinds of knowledge (Morone et al 2007 or study 3 in this thesis)

Regarding question b) how it diffuses researchers either assumed knowledgeexchange as a kind of barter trade in which agents only exchange knowledge ifthis is mutually beneficial (Cowan and Jonard 2004 2007 Morone and Taylor 2004Cassi and Zirulia 2008 Kim and Park 2009) Alternatively they assume knowledgeexchange as a gift transaction ie for some reason agents freely give away theirknowledge (Cowan and Jonard 2007 Morone et al 2007 Lin and Li 2010 Zhuanget al 2011 or study 2 and 4 in this thesis)

Concerning c) where it diffuses most studies focus on knowledge diffusionin different archetypical network structures as eg small-world structures ran-dom structures or regular structures Most of the time the investigated structuresare static (Cowan and Jonard 2004 2007 Morone et al 2007 Cassi and Zirulia2008 Kim and Park 2009 Lin and Li 2010 or study 1 and 2 in this thesis) whilesometimes researchers also investigate dynamic network structures or networkstructures evolving over time (Morone and Taylor 2004 Zhuang et al 2011 orstudy 4 in this thesis)

As the effects of network characteristics and structures on knowledge diffusionperformance depend on many different aspects researchers found ambiguous ef-fects of network characteristics and structures on diffusion Most studies measured) the effect of the diffusion process as efficiency and equity of knowledge diffu-sion ie the mean average knowledge stocks of all agents over time as well as thevariance of knowledge levels (Cowan and Jonard 2004 2007 Morone and Taylor2004 Morone et al 2007 Cassi and Zirulia 2008 Kim and Park 2009 Lin and Li2010 Zhuang et al 2011 and study 1 and 2 of this thesis) (sometimes comple-mented by an analysis of individual knowledge levels (Zhuang et al 2011 andstudy 1 and 2 of this thesis)) However d) the actual performance differs

Many researchers agree that small-world network structures (Watts and Stro-gatz 1998) are favourable (if not best) for knowledge diffusion performance(Cowan and Jonard 2004 Morone and Taylor 2004 Morone et al 2007 Cassiand Zirulia 2008 Kim and Park 2009 and study 1 of this thesis) Nonethelessthis is not always the case and depends on different aspects While Cowan andJonard found that small-world structures lead to the most efficient diffusion ofknowledge these structures at the same time lead to most unequal knowledge

6

Chapter 1 Introduction

distribution (Cowan and Jonard 2004) Morone and Taylor found that small-world networks only provide optimal patterns for diffusion if some lsquobarriers tocommunicationrsquo are removed (Morone and Taylor 2004) Cassi and Zirulia statethe small-world networks only foster diffusion performance for a certain level ofopportunity costs for using the network (Cassi and Zirulia 2008) Moreover otherstudies found that other network structures provide better patterns for knowledgediffusion in general These are for instance scale-free network structures (Lin andLi 2010) or even random network structures (Morone et al 2007) Besides manyresearchers agree that how network structures affect diffusion performance alsodepends on eg agentsrsquo absorptive capacities (Cowan and Jonard 2004) agentsrsquocognitive distances (Morone and Taylor 2004 and study 2 in this thesis) on thediffusion mechanism (Cowan and Jonard 2007 and study 1 and 2 in this thesis)on the special kind of knowledge and its relatedness (Morone et al 2007) on thecosts of using the network (Cassi and Zirulia 2008) and many more ldquoAnalysingthe structure of the network independently of the effective content of the relationcould be therefore misleadingrdquo (Cassi et al 2008 p 285) Summing up it isimpossible to make general statements and give valid policy recommendationsabout network structures lsquooptimalrsquo for knowledge diffusion

As knowledge diffusion performance is affected by so many different inter-dependent aspects Table 11 in the Appendix gives an overview of those studiesconducting experiments on knowledge diffusion used for and in my thesis Amore general comprehensive overview however goes beyond the scope of thisintroduction Such a rather general overview can for instance be found in Va-lente (2006) Morone and Taylor (2010) Jackson and Yariv (2011) or Lamberson(2016)

12 Analysing Knowledge Diffusion in Innovation Net-works

Analysing knowledge diffusion can be quite a challenging task As learningknowledge exchange and knowledge diffusion mainly take place between con-nected agents in different types of networks a method needed and used in studiesanalysing knowledge diffusion within networks is social network analysis (SNA)

Social network analysis aims at describing and exploring ldquopatterns apparentin the social relationships that individuals and groups form with each other () Itseeks to go beyond the visualisation of social relations to an examination of theirstructural properties and their implications for social actionrdquo (Scott 2017 p 2)Since knowledge diffuses throughout networks many studies analyse how theunderlying network structures or properties of agents within the network influ-ence knowledge diffusion (Ahuja 2000 Cowan and Jonard 2004 2007 Mueller etal 2017 Bogner et al 2018 to name just a few) Applying the network perspec-tive ldquoallows new leverage for answering standard social and behavioural scienceresearch questions by giving precise formal definition to aspects of the politicaleconomic or social structural environmentrdquo (Wasserman and Faust 1994 p 3)Hence the network perspective allows not only for analysing eg face-to-faceknowledge exchange between two agents but rather to focus on knowledge sys-tems and therefore to examine knowledge spread and diffusion throughout thewhole network

Due to the impossibility of actually measuring knowledge and its diffusionapproaches in diffusion literature either use empirical methods as measuring

7

Chapter 1 Introduction

patents (Ahuja 2000 Nelson 2009) or questionnaires (Reagans and McEvily 2003)Or they focus on using experiments ranging from laboratory experiments (Mo-rone et al 2007) to quasi-experiments in form of different models such as perco-lation models (Silverberg and Verspagen 2005 Bogner 2015) or other agent-basedmodels (ABMs) (Cowan and Jonard 2004 2007 Mueller et al 2017 Bogner et al2018) The studies presented in this doctoral thesis analyse knowledge diffusionthrough agent-based models (study 1 and 2) and by analysing network structuresand potential diffusion performance by deducing from theoretical considerationsin social network analysis (study 4)

The advantage of using agent-based models for analysing knowledge diffu-sion results from the fact that ABM can help to not only understand agentsrsquo orindividualsrsquo behaviours but also to understand aggregate behaviour and how thebehaviour and interaction of many individual agents lead to large-scale outcomes(Axelrod 1997) This helps understanding fundamental processes that take placeAgent-based modeling is a way of doing thought experiments Although the as-sumptions may be simple the consequences may not be at all obvious (Axelrod1997 p 4) According to Axelrod ABM [] is a third way of doing science(Axelrod 1997 p 3) in addition to deduction and induction

Many scientists argue that traditional modelling tools in economics are nolonger sufficient for analysing innovation processes especially if we want to anal-yse innovations in an increasingly complex and interdependent world (Dawid2006 Pyka and Fagiolo 2007) This is another reason why the method of agent-based modelling has become that famous Dawid (2006) explains this by thefact that ABM is able to incorporate the particularities of innovation processessuch as the unique nature of knowledge and the heterogeneity of actorsrsquo knowl-edge Agent-based simulation allows for modelling and explaining rsquotruersquo non-reversible path dependency feedbacks herding-behaviour and global phenom-ena as for instance the creation of knowledge and its diffusion (Pyka and Fagiolo2007) Therefore the method of agent-based modelling (ABM) is an approach thatbecomes more and more important especially in modelling and studying complexsystems with autonomous decision making agents that interact with each otherand with their environment (North and Macal 2007) as agents that repeatedly ex-change knowledge being arranged according to a particular network algorithm

13 Structure of this Thesis

Given the importance of understanding and even managing knowledge exchangebetween different actors this doctoral thesis aims to use methods as social net-work analysis and agent-based modelling in addition to theoretical considera-tions to explore how different network structures influence knowledge diffusionin different settings Therefore the overall research question of this doctoral the-sis is ldquoWhat are the effects of different network structures on knowledge diffusionperformance in RampD networks This research question can be subdivided in nineresearch questions covered in the four studies presented in this thesis These are

bull Covered in study 1 What are the effects of different network structures onknowledge diffusion performance in informal networks in which knowl-edge is exchanged as a barter trade How does the embeddedness of ac-tors and especially the existence of lsquostarsrsquo affect individual as well as overalldiffusion performance

8

Chapter 1 Introduction

bull Covered in study 2 What are the effects of different network structures onknowledge diffusion performance in formal RampD networks in which knowl-edge diffuses freely between the actors however is limited by the differentcognitive distances of these actors Which cognitive distance between actorsis best in the respective network structure

bull Covered in study 3 Which types and characteristics of knowledge are in-strumental creating policies fostering a transformation towards a sustain-able knowledge-based Bioeconomy (SKBBE) What are the policy-relevantimplications of this extended perspective on the characteristics of knowl-edge

bull Covered in study 4 What is the structure of the publicly funded RampD net-work in the German Bioeconomy From a theoretical point of view howdoes this network structure influence knowledge diffusion of both meretechno-economic knowledge as well as dedicated knowledge Does policymaking allow for or even create network structures that adequately supportthe creation and diffusion of dedicated knowledge in RampD networks

To answer the named research questions the structure of this doctoral thesisis as follows (see also Figure 11)

FIGURE 11 Structure of my doctoral thesis

First chapter 1 gives an introduction to the topics covered the research ques-tions answered the and methods used

In chapter 2 study 1 analyses the effect of different structural disparities onknowledge diffusion by using an agent-based simulation model It focuses onhow different network structures influence knowledge diffusion performanceThis study especially emphasises the effect of an asymmetric degree distributionon knowledge diffusion performance Study 1 complements previous research onknowledge diffusion by showing that (i) besides or even instead of the averagepath length and the average clustering coefficient the (asymmetry of) degreedistribution influences knowledge diffusion In addition (ii) especially smallinadequately embedded agents seem to be a bottleneck for knowledge diffusion

9

Chapter 1 Introduction

in this setting and iii) the identified somewhat negative network structures onthe macro level appear to result from the myopic linking strategies of the actors atthe micro level indicating a trade-off between lsquooptimalrsquo structures at the networkand the actor level

In chapter 3 study 2 uses an agent-based simulation model to analyse theeffect of different network properties on knowledge diffusion performance Asstudy 1 study 2 also focuses on the diffusion of knowledge however in contrastto study 1 this study analyses this relationship in a setting in which knowledge isdiffusing freely throughout an empirical formal RampD network as well as throughthe four benchmark networks already investigated in study 1 Besides the con-cept of cognitive distance and differences in learning between agents in the net-work are taken into account Study 2 complements study 1 and further previousresearch on knowledge diffusion by showing that (i) the (asymmetry of) degreedistribution and the distribution of links between actors in the network indeedinfluence knowledge diffusion performance to a large extent In addition (ii) theextent to which a skewed degree distribution dominates other network character-istics varies depending on the respective cognitive distances between agents

In chapter 4 study 3 analyses how so-called dedicated knowledge can contributeto the transformation towards a sustainable knowledge-based Bioeconomy Inthis study the concept of dedicated knowledge ie besides mere-techno economicknowledge also systems knowledge normative knowledge and transformativeknowledge is first introduced Moreover the characteristics of dedicated knowl-edge which are influencing knowledge diffusion performance are analysed andevaluated according to their importance and potential role for knowledge diffu-sion In addition it is analysed if and how current Bioeconomy innovation policiesaccount for dedicated knowledge This study complements previous research onknowledge and knowledge diffusion by taking a strong focus on different typesof knowledge besides techno-economic knowledge (often overemphasised in pol-icy approaches) It shows that i) different types of knowledge necessarily needto be taken into account when creating policies for knowledge creation and diffu-sion and ii) that especially systems knowledge so far has been only insufficientlyconsidered by current Bioeconomy policy approaches

In chapter 5 study 4 analyses the effect of different structural disparities onknowledge diffusion by deducing from theoretical considerations on networkstructures and diffusion performance This study focuses on how different net-work structures influence knowledge diffusion performance by analysing anempirical RampD network in the German Bioeconomy over the last 30 years bymeans of descriptive statistics and social network analysis The first part of thestudy presents the descriptive statistics of the publicly funded RampD projects andthe research conducting actors within the last 30 years The second part of thestudy shows the network characteristics and structures of the network in six dif-ferent observation periods These network statistics and structures are evaluatedfrom a knowledge diffusion point of view The study tries to answer whether theartificially generated network structures seem favourable for the diffusion of bothmere techno-economic knowledge as well as dedicated knowledge Study 4 es-pecially complements previous research on knowledge diffusion by (i) analysingan empirical network over such a long period of time (i) and (ii) by showing thateven though a network and its structure might be favourable for the diffusionof information or mere techno-economic knowledge this does not imply it alsofosters the creation and diffusion of other types of knowledge (ie dedicated

10

Chapter 1 Introduction

knowledge) necessary for the transformation towards a sustainable knowledge-based Bioeconomy (SKBBE)

In chapter 6 the main results of this doctoral thesis are summarised and dis-cussed and an outlook on possible future research avenues is given

References

Abson David J Baumgaumlrtner Stefan Fischer Joumlrn Hanspach Jan Haumlrdtle WHeinrichs H et al (2014) Ecosystem services as a boundary object for sus-tainability In Ecological Economics 103 pp 29ndash37

Ahuja Gautam (2000) Collaboration Networks Structural Holes and Innova-tion A Longitudinal Study In Administrative science quarterly

Arrow Kenneth Joseph (1972) Economic welfare and the allocation of resourcesfor invention In Readings in Industrial Economics Springer pp 219ndash236

Axelrod Robert (1997) The complexity of cooperation Agent-based models of compe-tition and collaboration Princeton University Press

Barabaacutesi Albert-Laacuteszloacute (2016) Network science Cambridge University Press

Barabaacutesi Albert-Laacuteszloacute Albert Reacuteka (1999) Emergence of Scaling in RandomNetworks In Science 286 (5439) pp 509ndash512

Bjoumlrk Jennie Magnusson Mats (2009) Where Do Good Innovation Ideas ComeFrom Exploring the Influence of Network Connectivity on Innovation IdeaQuality In Journal of Product Innovation Management 26 (6) pp 662ndash670

Bogner Kristina (2015) The Effect of Project Funding on Innovative Performance- An Agent-Based Simulation Model In Hohenheimer Discussion Papers inBusiness Economics and Social Sciences

Bogner Kristina Mueller Matthias Schlaile Michael P (2018) Knowledge dif-fusion in formal networks the roles of degree distribution and cognitivedistance In Int J Computational Economics and Econometrics pp 388ndash407

Boschma Ron A (2005) Proximity and innovation a critical assessment InRegional studies 39 (1) pp 61ndash74

Carayol Nicolas Roux Pascale (2009) Knowledge flows and the geography ofnetworks A strategic model of small world formation In Journal of Eco-nomic Behavior amp Organization 71 (2) pp 414-427

Cassi Lorenzo Corrocher Nicoletta Malerba Franco Vonortas Nicholas (2008)The impact of EU-funded research networks on knowledge diffusion at theregional level In Res Eval 17 (4) pp 283ndash293

Cassi Lorenzo Zirulia Lorenzo (2008) The opportunity cost of social relationsOn the effectiveness of small worlds In J Evol Econ 18 (1) pp 77ndash101

Chaminade Cristina Esquist Charles (2010) Rationales for public policy inter-vention in the innovation process Systems of innovation approach In Thetheory and practice of innovation policy Edward Elgar Publishing

11

Chapter 1 Introduction

Chiu Yen-Ting Helena (2008) How network competence and network locationinfluence innovation performance In Jnl of BusampIndus Marketing 24 (1) pp46ndash55

Choi Hanool Kim Sang-Hoon Lee Jeho (2010) Role of network structure andnetwork effects in diffusion of innovations In Industrial Marketing Manage-ment 39 (1) pp 170ndash177

Cohen Wesley M Levinthal Daniel A (1989) Innovation and learning the twofaces of R amp D In The economic journal 99 (397) pp 569ndash596

Cohen Wesley M Levinthal Daniel A (1990) Absorptive Capacity A NewPerspective on Learning and Innovation In Administrative science quarterly

Cowan Robin Jonard Nicolas (2004) Network structure and the diffusionof knowledge In Journal of Economic Dynamics and Control 28 (8) pp1557ndash1575

Cowan Robin Jonard Nicolas (2007) Structural holes innovation and the dis-tribution of ideas In J Econ Interac Coord 2 (2) pp 93ndash110

Dawid Herbert (2006) Agent-based models of innovation and technologicalchange In Handbook of computational economics 2 pp 1235ndash1272

Dosi Giovanni (1982) Technological paradigms and technological trajectoriesa suggested interpretation of the determinants and directions of technicalchange In Research Policy 11 (3) pp 147ndash162

Erdos Paul Reacutenyi Alfreacuted (1959) On random graphs Publicationes Mathemati-cate pp 290-297

Erdos Paul Reacutenyi Alfreacuted (1960) On the evolution of random graphs In PublMath Inst Hung Acad Sci 5 (1) pp 17ndash60

Foray Dominique Lundvall Bengt-Aringke (1998) The knowledge-based economyfrom the economics of knowledge to the learning economy In The economicimpact of knowledge pp 115ndash121

Foray Dominique Mairesse Jacques (2002) The knowledge dilemma and thegeography of innovation In Institutions and Systems in the Geography of In-novation Springer pp 35ndash54

Fraunhofer IMW (2018) Knowledge - the most valuable resource of the future EdFraunhofer Center for International Management and Knowledge EconomyIMW Available via httpswwwimwfraunhoferdeenpressinterview-annual-reporthtml (accessed on 16 April 2018)

Galunic Charles Rodan Simon (1998) Resource recombinations in the firmKnowledge structures and the potential for Schumpeterian innovation InStrat Mgmt J 19 (12) pp 1193ndash1201

Gilsing Victor Nooteboom Bart Vanhaverbeke Wim Duysters Geert van denOord Ad (2008) Network embeddedness and the exploration of novel tech-nologies Technological distance betweenness centrality and density InResearch Policy 37 (10) pp 1717ndash1731

12

422 How Knowledge Concepts have Inspired Innovation PolicyMaking 67

423 Knowledge Concepts in Transformative Sustainability Science 6743 Dedicated Knowledge for an SKBBE Transformation 68

431 Systems Knowledge 68432 Normative Knowledge 70433 Transformative Knowledge 71

44 Policy-Relevant Implications 73441 Knowledge-Related Gaps in Current Bioeconomy Policies 73442 Promising (But Fragmented) Building Blocks for Improved

SKBBE Policies 7545 Conclusions 76

5 Knowledge Networks in the German Bioeconomy Network Structure ofPublicly Funded RampD Networks 9351 Introduction 9452 Knowledge and its Diffusion in RampD Networks 95

521 Knowledge Diffusion in Different Network Structures 96522 Knowledge for a Sustainable Knowledge-Based Bioeconomy 100523 On Particularities of Publicly Funded Project Networks 102

53 The RampD Network in the German Bioeconomy 104531 Subsidised RampD Projects in the German Bioeconomy in the

Past 30 Years 105532 Network Structure of the RampD Network 108

54 The RampD Network in the German Bioeconomy and its PotentialPerformance 115

55 Conclusion and Future Research Avenues 119

6 Discussion and Conclusion 12861 Summary and Discussion of Results 12862 Conclusions Limitations and Avenues for Future Research 133

iv

List of Figures

11 Structure of my doctoral thesis 9

21 Visualisation of network topologies created in NetLogo 2522 Mean average knowledge levels of agents 2623 Mean average knowledge levels of agents 2724 Average degree distribution of the agents 2925 Relationship between the variance of the degree distribution and

the mean average knowledge level 2926 Relationship between the time the agents in the network stop trad-

ing and their degree 3027 Relationship between the agentsrsquo mean knowledge levels their de-

gree and the reason why agents eventually stopped trading 3128 Relationship between agentsrsquo mean knowledge levels and the de-

gree difference between them and their partners 3329 Effect of different policy interventions on mean average knowledge

levels 34210 Effect of different policy interventions on the mean average knowl-

edge levels of the 10 smallest agents 35

31 Knowledge gain for different max cognitive distances 4632 Flow-chart of the evolutionary network algorithm 4833 Average degree distribution of the actors 5034 Mean knowledge levels and variance of the average knowledge lev-

els for different max cognitive distances 5135 Mean knowledge levels and variance of the average knowledge lev-

els at different points in time 5236 Histogram of knowledge levels 5437 Relationship between actorsrsquo mean knowledge levels and their de-

gree centralities 55

51 Type and number of actors projects and moneyyear 10852 Visualisation of the knowledge network 10953 Number of nodesedges and network density as well as average

degree 11154 Average path length and average clustering coefficients of the net-

works 11355 Degree distribution of the RampD network in the different observa-

tion periods 11456 Top-15 actors in most joint projects (lhs) and most single projects

(rhs) 126

v

List of Tables

11 Overview of agent-based models on knowledge diffusion 16

21 Average path length and global clustering coefficient (cliquishness)of the four network topologies 27

31 Standard parameter setting 4932 Network characteristics of our five networks 50

41 The elements of dedicated knowledge in the context of SKBBE poli-cies 74

51 Descriptive statistics of both joint and single research projects 10652 Network characteristics of the RampD network in different observa-

tion periods 110

vi

This doctoral thesis is dedicated to three people in particularIt is dedicated to my precious sister for always supporting

and believing in meIt is dedicated to my beloved husband for lighting up my life

and being my best friendIt is dedicated to my dearest son for proving me that I have

forces I never knew I had

vii

Chapter 1

Introduction

Chapter 1

Introduction

Knowledge is the most valuable resource of the futurerdquo (Fraunhofer IMW 2018)Researchers and policy makers alike agree upon the fact that knowledge is a cru-cial economic resource both as an input and output of innovation processes (Lund-vall and Johnson 1994 Foray and Lundvall 1998) However it is not only the inputand output of innovation processes but moreover the solution to problems (Potts2001) This resource is allowing firms to innovate and keep pace with nationaland international competitors It is helping to generate technological progresseconomic growth and prosperity Therefore understanding and managing thecreation and diffusion of knowledge within and outside of firms and innovationnetworks is of utmost importance to make full use of this precious resource

11 Knowledge and its Diffusion in Innovation Networks

How knowledge has been created and managed by entrepreneurs firms and pol-icy makers within the last decades strongly depends on the underlying under-standing and definition of knowledge In mainstream neo-classical economicsknowledge has been understood and treated as an intangible good exhibitingpublic good features as non-excludability and non-rivalry in consumption (Solow1956 1957 Arrow 1972) These (alleged) public good features of knowledge havebeen assumed to cause different problems As other actors cannot be excludedfrom the consumption of a public good new knowledge freely flows from oneactor to another causing spillover effects In this situation some actors can bene-fit from the new knowledge without having invested in its creation ie a typicalfree-riding problem emerges (Pyka et al 2009) Therefore the knowledge creat-ing actor cannot fully appropriate the returns resulting from his research activities(Arrow 1972) leading to market failure ie a situation in which the investment inRampD is below the social optimum In this mainstream neo-classical world knowl-edge falls like manna from heaven (Solow 1956 1957) Knowledge instantly flowsfrom one actor to another (at no costs) and therefore there is no need for learningThis understanding and definition of knowledge has influenced policy makingfor a long period Policies inspired by the concept of knowledge as a public goodhave mainly focused on the mitigation of potential externalities for instance byincentive creation through RampD subsidies and knowledge production by the pub-lic sector (Smith 1994 Chaminade and Esquist 2010) While policies changed toa certain extent when the understanding of knowledge changed nowadays in-novation policies as RampD subsidies (especially for the public sector) in sociallydesirable fields are still common practice (see chapter 5 on this topic)

Challenged by the fact that the mainstream neo-classical understanding lacksin giving an adequate and comprehensive definition of knowledge (evolutionaryor neo-Schumpeterian) innovation economists and management scientists tried

2

Chapter 1 Introduction

providing a much more appropriate analysis of knowledge creation and diffu-sion by considering different features of knowledge affecting its creation anddiffusion Following this understanding knowledge can instead be seen as a la-tent public good (Nelson 1989) exhibiting many non-public good features Thesefeatures have a substantial impact on how to best manage the valuable resourceknowledge Nowadays we know that knowledge is far from being a pure publicgood Knowledge is characterised by cumulativeness (Foray and Mairesse 2002Boschma 2005) relatedness (Morone and Taylor 2010) path dependency (Dosi1982 Rizzello 2004) tacitness (Polanyi 2009) stickiness (von Hippel 1994 Szulan-ski 2002) dispersion (Galunic and Rodan 1998b) context specificity and locality(Galunic and Rodan 1998b) Therefore the neo-Schumpeterian approach espe-cially emphasises the role of knowledge and learning (Hanusch and Pyka 2007Malerba 2007) In line with the neo-Schumpeterian approach anyone that hasever prepared for an exam written a doctoral thesis or even tried to cook a dishfrom a famous cook would agree upon the fact that knowledge does not fall likemanna from heaven and does not freely flow from one actor to another without theother actor actively acquiring the knowledge (learning) Instead it is the case thatthe named non-public good characteristics of knowledge substantially influencelearning and the exchange and diffusion of knowledge between actors and withina network

Inspired by communication theory how the characteristics of knowledgeinfluence its exchange and diffusion can be understood by using the famousShannon-Weaver-Model of communication (Shannon 1948 Shannon and Weaver1964) Following the idea of this model not only information transfer but alsoknowledge transfer can (at least to a certain extent) be seen as a systemic processin which knowledge is transmitted from a sender to a receiver (see eg Schwartz2005) How knowledge can be transferred from one actor to another dependsboth on the characteristics of the sender and the receiver and on the characteris-tics of knowledge itself When talking about knowledge and the impacts of itscharacteristics on knowledge exchange the questions are (i) is it possible to sendthe knowledge (ii) can the receiver re-interpret the knowledge and (iii) can theknowledge be used by the receiver

Characteristics of knowledge that influence knowledge diffusion and (i)whether it is possible to send the knowledge are tacitness stickiness and dis-persion Knowledge is not equal to information Essential parts of knowledgeare tacit ie very difficult to be codified and to be transferred (Galunic and Ro-dan 1998a) Tacit knowledge is rival and excludable and therefore not public(Polanyi 1959) Even if the sender of the knowledge is willing to share the tacit-ness makes it sometimes impossible to send this knowledge Besides knowledgeand its transfer can be sticky (von Hippel 1994 Szulanski 2002) ie the transferof this knowledge requires significantly more effort than the transfer of otherknowledge According to Szulanski (2002) both the knowledge and the processof knowledge exchange might be sticky Reasons can be the kind and amount ofknowledge itself but also attributes of the sender or receiver of knowledge Thedispersion of knowledge also influences the possibility of sending knowledgeGalunic and Rodan (1998a) explain dispersed knowledge by using the exampleof a jigsaw puzzle The authors state that knowledge is distributed if all actorsreceive a photocopy of the picture of the jigsaw puzzle At the same time theknowledge is dispersed if every actor receives one piece of the jigsaw puzzlemeaning that everybody only holds pieces of the knowledge but not the lsquowholersquopicture Dispersed knowledge (or systems-embedded knowledge) is difficult to

3

Chapter 1 Introduction

be sent as detecting dispersed knowledge can be quite problematic (Galunic andRodan 1998a)

Characteristics of knowledge and actors that influence (ii) whether the receivercan re-interpret the knowledge are the cumulative nature of knowledge or theknowledge relatedness agentsrsquo skills or cognitive distances and their absorptivecapacities New knowledge always is a (re-)combination of previous knowledge(Schumpeter 1912) The more complex and industry-specific the knowledge is thehigher the importance of the own knowledge stock and knowledge relatednessTo understand and integrate new knowledge agents need absorptive capacities(Cohen and Levinthal 1989 1990) The higher the cognitive distance between twoactors the more difficult it is for them to exchange and internalise knowledgeSo the lsquooptimalrsquo cognitive distance can be decisive for learning (Nooteboom et al2007)

Characteristics of knowledge that influence (iii) whether the knowledge canactually be used by the receiver are the context-specific or local character of knowl-edge Even if the knowledge is freely available public good features of knowledgemight not be decisive and the knowledge might be of little or no use to the receiverWe have to keep in mind that knowledge itself has no value it only becomes valu-able to someone if the knowledge can be used to eg solve specific problems(Potts 2001) Hence knowledge has different values to different actors and fol-lowing this assumption more knowledge is not always better Actors need theright knowledge in the right context and have to be able to combine this knowl-edge in the right way The valuable resource knowledge might only be relevantand of use in the narrow context for and in which it was developed (Galunic andRodan 1998a)

Making use of this model from communication theory quite impressivelyshows why an adequate comprehensive understanding and definition of knowl-edge and its characteristics is of utmost importance when analysing and manag-ing knowledge diffusion between actors and within networks With an inade-quate understanding and definition of knowledge policy makers will not be ableto manage and foster knowledge creation and diffusion This can for instance beseen by policies inspired by a somewhat linear understanding of knowledge andinnovation processes heavily supporting basic research but failing to transfer theresults into practice and producing successful innovations

It is safe to state that the improved understanding and definition of knowledgepositively influenced innovation policies (see eg study 3 in this thesis) By ap-plying a more realistic understanding of the characteristics of knowledge and howthese influence knowledge creation and diffusion researchers and policy makerswere able to shift their focus on systemic problems instead of the correction ofmarket failures (Chaminade and Esquist 2010) Understanding the importance ofnetworks and their structures for knowledge diffusion performance allows for cre-ating network structures that foster knowledge diffusion However researchersand policy makers alike so far mainly focus on the creation and diffusion of in-formation or mere techno-economic knowledge This intense focus neglects otherimportant characteristics and types of knowledge and therefore is not able to givevalid policy recommendations eg for innovation policies aiming at fosteringtransformation endeavours Especially the politically desired transformation to-wards a sustainable knowledge-based Bioeconomy (SKBBE) has been identified tobe in need of more than the production and diffusion of techno-economic knowl-edge (Urmetzer et al 2018) Inspired by sustainability literature three other typesof knowledge have been identified to be relevant for such transformations (Abson

4

Chapter 1 Introduction

et al 2014) These are systems knowledge normative knowledge and transfor-mative knowledge In chapter 4 the concept of so-called dedicated knowledge in-corporating besides mere techno-economic knowledge also systems normativeand transformative knowledge is presented

Since the effect of network structure on knowledge diffusion performance isstrongly affected by what actually diffuses analysing and incorporating differentkinds of knowledge is important Therefore the first step in this direction is donein study 3 (chapter 4) Study 3 puts special emphasis on different types of knowl-edge necessary to foster the transformation towards a sustainable knowledge-based Bioeconomy (SKBBE) It shows that innovation policies so far put no suffi-cient attention on what actually diffuses throughout innovation networks

As the collection and generation of (new) knowledge give such competi-tive advantages there is a keen interest of firms and policy makers on howto foster the creation and diffusion of (new) knowledge Firms can get accessto new knowledge either by internal knowledge generation processes as egintra-organisational learning or by access to external knowledge sources eginter-organisational cooperation (Malerba 1992) Nowadays strong focus on inter-organisational cooperation can be explained by the fact that invention activityis far from being the outcome of isolated agentsrsquo efforts but that of interactiveand collective processes (Carayol and Roux 2009 p 2) In nearly all industriesand technological areas we observe a pronounced intensity of RampD cooperationand the emergence of innovation networks with dynamically changing compo-sitions over time (Kudic 2015) At the same time we know that the economicactorsrsquo innovativeness is strongly affected by their strategic network positioningand the structural characteristics of the socio-economic environment in which theactors are embedded Networks provide a natural infrastructure for knowledgeexchange and provide the pipes and prisms of markets by enabling informationand knowledge flow (Podolny 2001)

It comes as no surprise that innovation networks and how knowledge diffu-sion within (and outside of) these networks takes place have become the centreof attention within the last decades (Valente 2006 Morone and Taylor 2010 Jack-son and Yariv 2011 Lamberson 2016) Diffusion literature in this context has beenidentified to mainly focus on four different areas or key questions

These area) What diffusesb) How does it diffusec) Where does it diffused) What are the effects (or performance) of the diffusion process and how are

they measured (see Schlaile et al 2018)In line with this the four studies presented in this thesis focus to different

extends on these four questions Study 1 and 2 focus on how different networkstructures influence knowledge diffusion performance (key question c) and d))for different diffusion mechanisms (key question b)) Study 3 focuses on whatactually diffuses within the network (key question a)) and study 4 combines find-ings from study 1 and 2 with those of study 3 namely how network structureinfluences knowledge diffusion in general and how network structure influencesknowledge diffusion of special types of knowledge in particular

These four key areas show that knowledge diffusion research can take variousforms with varying results Especially when analysing the effect of network prop-erties on knowledge diffusion performance as done in this thesis the identifiedeffects are manifold and even ambiguous This is why dependent on a) b) and c)

5

Chapter 1 Introduction

different studies identified different network characteristics and structures as fos-tering or being rsquooptimalrsquo for knowledge diffusion performance (see eg Moroneet al (2007) vs Lin and Li (2010))

While sometimes disagreeing in what actually is the best structure for diffu-sion performance agent-based simulation models within the last 15 years con-firmed that certain network characteristics and structures seem to be relativelyimportant for diffusion These are the networksrsquo average path lengths the net-worksrsquo average or global clustering coefficients the networksrsquo sizes and the net-worksrsquo degree distributions (Morone and Taylor 2004 Cowan and Jonard 20042007 Morone et al 2007 Cassi and Zirulia 2008 Kim and Park 2009 Choi et al2010 Lin and Li 2010 Zhuang et al 2011 Liu et al 2015 Luo et al 2015 Wang etal 2015 Yang et al 2015 Mueller et al 2017 Bogner et al 2018)

Concerning question a) what actually diffuses most studies focus on analysingknowledge (rather oversimplified represented as numbers or vectors) in differ-ent network structures (Morone and Taylor 2004 Cowan and Jonard 2004 2007Morone et al 2007 Cassi and Zirulia 2008 Kim and Park 2009 Lin and Li 2010Zhuang et al 2011 or study 1 and 2 in this thesis) Only a few authors investigatedifferent kinds of knowledge (Morone et al 2007 or study 3 in this thesis)

Regarding question b) how it diffuses researchers either assumed knowledgeexchange as a kind of barter trade in which agents only exchange knowledge ifthis is mutually beneficial (Cowan and Jonard 2004 2007 Morone and Taylor 2004Cassi and Zirulia 2008 Kim and Park 2009) Alternatively they assume knowledgeexchange as a gift transaction ie for some reason agents freely give away theirknowledge (Cowan and Jonard 2007 Morone et al 2007 Lin and Li 2010 Zhuanget al 2011 or study 2 and 4 in this thesis)

Concerning c) where it diffuses most studies focus on knowledge diffusionin different archetypical network structures as eg small-world structures ran-dom structures or regular structures Most of the time the investigated structuresare static (Cowan and Jonard 2004 2007 Morone et al 2007 Cassi and Zirulia2008 Kim and Park 2009 Lin and Li 2010 or study 1 and 2 in this thesis) whilesometimes researchers also investigate dynamic network structures or networkstructures evolving over time (Morone and Taylor 2004 Zhuang et al 2011 orstudy 4 in this thesis)

As the effects of network characteristics and structures on knowledge diffusionperformance depend on many different aspects researchers found ambiguous ef-fects of network characteristics and structures on diffusion Most studies measured) the effect of the diffusion process as efficiency and equity of knowledge diffu-sion ie the mean average knowledge stocks of all agents over time as well as thevariance of knowledge levels (Cowan and Jonard 2004 2007 Morone and Taylor2004 Morone et al 2007 Cassi and Zirulia 2008 Kim and Park 2009 Lin and Li2010 Zhuang et al 2011 and study 1 and 2 of this thesis) (sometimes comple-mented by an analysis of individual knowledge levels (Zhuang et al 2011 andstudy 1 and 2 of this thesis)) However d) the actual performance differs

Many researchers agree that small-world network structures (Watts and Stro-gatz 1998) are favourable (if not best) for knowledge diffusion performance(Cowan and Jonard 2004 Morone and Taylor 2004 Morone et al 2007 Cassiand Zirulia 2008 Kim and Park 2009 and study 1 of this thesis) Nonethelessthis is not always the case and depends on different aspects While Cowan andJonard found that small-world structures lead to the most efficient diffusion ofknowledge these structures at the same time lead to most unequal knowledge

6

Chapter 1 Introduction

distribution (Cowan and Jonard 2004) Morone and Taylor found that small-world networks only provide optimal patterns for diffusion if some lsquobarriers tocommunicationrsquo are removed (Morone and Taylor 2004) Cassi and Zirulia statethe small-world networks only foster diffusion performance for a certain level ofopportunity costs for using the network (Cassi and Zirulia 2008) Moreover otherstudies found that other network structures provide better patterns for knowledgediffusion in general These are for instance scale-free network structures (Lin andLi 2010) or even random network structures (Morone et al 2007) Besides manyresearchers agree that how network structures affect diffusion performance alsodepends on eg agentsrsquo absorptive capacities (Cowan and Jonard 2004) agentsrsquocognitive distances (Morone and Taylor 2004 and study 2 in this thesis) on thediffusion mechanism (Cowan and Jonard 2007 and study 1 and 2 in this thesis)on the special kind of knowledge and its relatedness (Morone et al 2007) on thecosts of using the network (Cassi and Zirulia 2008) and many more ldquoAnalysingthe structure of the network independently of the effective content of the relationcould be therefore misleadingrdquo (Cassi et al 2008 p 285) Summing up it isimpossible to make general statements and give valid policy recommendationsabout network structures lsquooptimalrsquo for knowledge diffusion

As knowledge diffusion performance is affected by so many different inter-dependent aspects Table 11 in the Appendix gives an overview of those studiesconducting experiments on knowledge diffusion used for and in my thesis Amore general comprehensive overview however goes beyond the scope of thisintroduction Such a rather general overview can for instance be found in Va-lente (2006) Morone and Taylor (2010) Jackson and Yariv (2011) or Lamberson(2016)

12 Analysing Knowledge Diffusion in Innovation Net-works

Analysing knowledge diffusion can be quite a challenging task As learningknowledge exchange and knowledge diffusion mainly take place between con-nected agents in different types of networks a method needed and used in studiesanalysing knowledge diffusion within networks is social network analysis (SNA)

Social network analysis aims at describing and exploring ldquopatterns apparentin the social relationships that individuals and groups form with each other () Itseeks to go beyond the visualisation of social relations to an examination of theirstructural properties and their implications for social actionrdquo (Scott 2017 p 2)Since knowledge diffuses throughout networks many studies analyse how theunderlying network structures or properties of agents within the network influ-ence knowledge diffusion (Ahuja 2000 Cowan and Jonard 2004 2007 Mueller etal 2017 Bogner et al 2018 to name just a few) Applying the network perspec-tive ldquoallows new leverage for answering standard social and behavioural scienceresearch questions by giving precise formal definition to aspects of the politicaleconomic or social structural environmentrdquo (Wasserman and Faust 1994 p 3)Hence the network perspective allows not only for analysing eg face-to-faceknowledge exchange between two agents but rather to focus on knowledge sys-tems and therefore to examine knowledge spread and diffusion throughout thewhole network

Due to the impossibility of actually measuring knowledge and its diffusionapproaches in diffusion literature either use empirical methods as measuring

7

Chapter 1 Introduction

patents (Ahuja 2000 Nelson 2009) or questionnaires (Reagans and McEvily 2003)Or they focus on using experiments ranging from laboratory experiments (Mo-rone et al 2007) to quasi-experiments in form of different models such as perco-lation models (Silverberg and Verspagen 2005 Bogner 2015) or other agent-basedmodels (ABMs) (Cowan and Jonard 2004 2007 Mueller et al 2017 Bogner et al2018) The studies presented in this doctoral thesis analyse knowledge diffusionthrough agent-based models (study 1 and 2) and by analysing network structuresand potential diffusion performance by deducing from theoretical considerationsin social network analysis (study 4)

The advantage of using agent-based models for analysing knowledge diffu-sion results from the fact that ABM can help to not only understand agentsrsquo orindividualsrsquo behaviours but also to understand aggregate behaviour and how thebehaviour and interaction of many individual agents lead to large-scale outcomes(Axelrod 1997) This helps understanding fundamental processes that take placeAgent-based modeling is a way of doing thought experiments Although the as-sumptions may be simple the consequences may not be at all obvious (Axelrod1997 p 4) According to Axelrod ABM [] is a third way of doing science(Axelrod 1997 p 3) in addition to deduction and induction

Many scientists argue that traditional modelling tools in economics are nolonger sufficient for analysing innovation processes especially if we want to anal-yse innovations in an increasingly complex and interdependent world (Dawid2006 Pyka and Fagiolo 2007) This is another reason why the method of agent-based modelling has become that famous Dawid (2006) explains this by thefact that ABM is able to incorporate the particularities of innovation processessuch as the unique nature of knowledge and the heterogeneity of actorsrsquo knowl-edge Agent-based simulation allows for modelling and explaining rsquotruersquo non-reversible path dependency feedbacks herding-behaviour and global phenom-ena as for instance the creation of knowledge and its diffusion (Pyka and Fagiolo2007) Therefore the method of agent-based modelling (ABM) is an approach thatbecomes more and more important especially in modelling and studying complexsystems with autonomous decision making agents that interact with each otherand with their environment (North and Macal 2007) as agents that repeatedly ex-change knowledge being arranged according to a particular network algorithm

13 Structure of this Thesis

Given the importance of understanding and even managing knowledge exchangebetween different actors this doctoral thesis aims to use methods as social net-work analysis and agent-based modelling in addition to theoretical considera-tions to explore how different network structures influence knowledge diffusionin different settings Therefore the overall research question of this doctoral the-sis is ldquoWhat are the effects of different network structures on knowledge diffusionperformance in RampD networks This research question can be subdivided in nineresearch questions covered in the four studies presented in this thesis These are

bull Covered in study 1 What are the effects of different network structures onknowledge diffusion performance in informal networks in which knowl-edge is exchanged as a barter trade How does the embeddedness of ac-tors and especially the existence of lsquostarsrsquo affect individual as well as overalldiffusion performance

8

Chapter 1 Introduction

bull Covered in study 2 What are the effects of different network structures onknowledge diffusion performance in formal RampD networks in which knowl-edge diffuses freely between the actors however is limited by the differentcognitive distances of these actors Which cognitive distance between actorsis best in the respective network structure

bull Covered in study 3 Which types and characteristics of knowledge are in-strumental creating policies fostering a transformation towards a sustain-able knowledge-based Bioeconomy (SKBBE) What are the policy-relevantimplications of this extended perspective on the characteristics of knowl-edge

bull Covered in study 4 What is the structure of the publicly funded RampD net-work in the German Bioeconomy From a theoretical point of view howdoes this network structure influence knowledge diffusion of both meretechno-economic knowledge as well as dedicated knowledge Does policymaking allow for or even create network structures that adequately supportthe creation and diffusion of dedicated knowledge in RampD networks

To answer the named research questions the structure of this doctoral thesisis as follows (see also Figure 11)

FIGURE 11 Structure of my doctoral thesis

First chapter 1 gives an introduction to the topics covered the research ques-tions answered the and methods used

In chapter 2 study 1 analyses the effect of different structural disparities onknowledge diffusion by using an agent-based simulation model It focuses onhow different network structures influence knowledge diffusion performanceThis study especially emphasises the effect of an asymmetric degree distributionon knowledge diffusion performance Study 1 complements previous research onknowledge diffusion by showing that (i) besides or even instead of the averagepath length and the average clustering coefficient the (asymmetry of) degreedistribution influences knowledge diffusion In addition (ii) especially smallinadequately embedded agents seem to be a bottleneck for knowledge diffusion

9

Chapter 1 Introduction

in this setting and iii) the identified somewhat negative network structures onthe macro level appear to result from the myopic linking strategies of the actors atthe micro level indicating a trade-off between lsquooptimalrsquo structures at the networkand the actor level

In chapter 3 study 2 uses an agent-based simulation model to analyse theeffect of different network properties on knowledge diffusion performance Asstudy 1 study 2 also focuses on the diffusion of knowledge however in contrastto study 1 this study analyses this relationship in a setting in which knowledge isdiffusing freely throughout an empirical formal RampD network as well as throughthe four benchmark networks already investigated in study 1 Besides the con-cept of cognitive distance and differences in learning between agents in the net-work are taken into account Study 2 complements study 1 and further previousresearch on knowledge diffusion by showing that (i) the (asymmetry of) degreedistribution and the distribution of links between actors in the network indeedinfluence knowledge diffusion performance to a large extent In addition (ii) theextent to which a skewed degree distribution dominates other network character-istics varies depending on the respective cognitive distances between agents

In chapter 4 study 3 analyses how so-called dedicated knowledge can contributeto the transformation towards a sustainable knowledge-based Bioeconomy Inthis study the concept of dedicated knowledge ie besides mere-techno economicknowledge also systems knowledge normative knowledge and transformativeknowledge is first introduced Moreover the characteristics of dedicated knowl-edge which are influencing knowledge diffusion performance are analysed andevaluated according to their importance and potential role for knowledge diffu-sion In addition it is analysed if and how current Bioeconomy innovation policiesaccount for dedicated knowledge This study complements previous research onknowledge and knowledge diffusion by taking a strong focus on different typesof knowledge besides techno-economic knowledge (often overemphasised in pol-icy approaches) It shows that i) different types of knowledge necessarily needto be taken into account when creating policies for knowledge creation and diffu-sion and ii) that especially systems knowledge so far has been only insufficientlyconsidered by current Bioeconomy policy approaches

In chapter 5 study 4 analyses the effect of different structural disparities onknowledge diffusion by deducing from theoretical considerations on networkstructures and diffusion performance This study focuses on how different net-work structures influence knowledge diffusion performance by analysing anempirical RampD network in the German Bioeconomy over the last 30 years bymeans of descriptive statistics and social network analysis The first part of thestudy presents the descriptive statistics of the publicly funded RampD projects andthe research conducting actors within the last 30 years The second part of thestudy shows the network characteristics and structures of the network in six dif-ferent observation periods These network statistics and structures are evaluatedfrom a knowledge diffusion point of view The study tries to answer whether theartificially generated network structures seem favourable for the diffusion of bothmere techno-economic knowledge as well as dedicated knowledge Study 4 es-pecially complements previous research on knowledge diffusion by (i) analysingan empirical network over such a long period of time (i) and (ii) by showing thateven though a network and its structure might be favourable for the diffusionof information or mere techno-economic knowledge this does not imply it alsofosters the creation and diffusion of other types of knowledge (ie dedicated

10

Chapter 1 Introduction

knowledge) necessary for the transformation towards a sustainable knowledge-based Bioeconomy (SKBBE)

In chapter 6 the main results of this doctoral thesis are summarised and dis-cussed and an outlook on possible future research avenues is given

References

Abson David J Baumgaumlrtner Stefan Fischer Joumlrn Hanspach Jan Haumlrdtle WHeinrichs H et al (2014) Ecosystem services as a boundary object for sus-tainability In Ecological Economics 103 pp 29ndash37

Ahuja Gautam (2000) Collaboration Networks Structural Holes and Innova-tion A Longitudinal Study In Administrative science quarterly

Arrow Kenneth Joseph (1972) Economic welfare and the allocation of resourcesfor invention In Readings in Industrial Economics Springer pp 219ndash236

Axelrod Robert (1997) The complexity of cooperation Agent-based models of compe-tition and collaboration Princeton University Press

Barabaacutesi Albert-Laacuteszloacute (2016) Network science Cambridge University Press

Barabaacutesi Albert-Laacuteszloacute Albert Reacuteka (1999) Emergence of Scaling in RandomNetworks In Science 286 (5439) pp 509ndash512

Bjoumlrk Jennie Magnusson Mats (2009) Where Do Good Innovation Ideas ComeFrom Exploring the Influence of Network Connectivity on Innovation IdeaQuality In Journal of Product Innovation Management 26 (6) pp 662ndash670

Bogner Kristina (2015) The Effect of Project Funding on Innovative Performance- An Agent-Based Simulation Model In Hohenheimer Discussion Papers inBusiness Economics and Social Sciences

Bogner Kristina Mueller Matthias Schlaile Michael P (2018) Knowledge dif-fusion in formal networks the roles of degree distribution and cognitivedistance In Int J Computational Economics and Econometrics pp 388ndash407

Boschma Ron A (2005) Proximity and innovation a critical assessment InRegional studies 39 (1) pp 61ndash74

Carayol Nicolas Roux Pascale (2009) Knowledge flows and the geography ofnetworks A strategic model of small world formation In Journal of Eco-nomic Behavior amp Organization 71 (2) pp 414-427

Cassi Lorenzo Corrocher Nicoletta Malerba Franco Vonortas Nicholas (2008)The impact of EU-funded research networks on knowledge diffusion at theregional level In Res Eval 17 (4) pp 283ndash293

Cassi Lorenzo Zirulia Lorenzo (2008) The opportunity cost of social relationsOn the effectiveness of small worlds In J Evol Econ 18 (1) pp 77ndash101

Chaminade Cristina Esquist Charles (2010) Rationales for public policy inter-vention in the innovation process Systems of innovation approach In Thetheory and practice of innovation policy Edward Elgar Publishing

11

Chapter 1 Introduction

Chiu Yen-Ting Helena (2008) How network competence and network locationinfluence innovation performance In Jnl of BusampIndus Marketing 24 (1) pp46ndash55

Choi Hanool Kim Sang-Hoon Lee Jeho (2010) Role of network structure andnetwork effects in diffusion of innovations In Industrial Marketing Manage-ment 39 (1) pp 170ndash177

Cohen Wesley M Levinthal Daniel A (1989) Innovation and learning the twofaces of R amp D In The economic journal 99 (397) pp 569ndash596

Cohen Wesley M Levinthal Daniel A (1990) Absorptive Capacity A NewPerspective on Learning and Innovation In Administrative science quarterly

Cowan Robin Jonard Nicolas (2004) Network structure and the diffusionof knowledge In Journal of Economic Dynamics and Control 28 (8) pp1557ndash1575

Cowan Robin Jonard Nicolas (2007) Structural holes innovation and the dis-tribution of ideas In J Econ Interac Coord 2 (2) pp 93ndash110

Dawid Herbert (2006) Agent-based models of innovation and technologicalchange In Handbook of computational economics 2 pp 1235ndash1272

Dosi Giovanni (1982) Technological paradigms and technological trajectoriesa suggested interpretation of the determinants and directions of technicalchange In Research Policy 11 (3) pp 147ndash162

Erdos Paul Reacutenyi Alfreacuted (1959) On random graphs Publicationes Mathemati-cate pp 290-297

Erdos Paul Reacutenyi Alfreacuted (1960) On the evolution of random graphs In PublMath Inst Hung Acad Sci 5 (1) pp 17ndash60

Foray Dominique Lundvall Bengt-Aringke (1998) The knowledge-based economyfrom the economics of knowledge to the learning economy In The economicimpact of knowledge pp 115ndash121

Foray Dominique Mairesse Jacques (2002) The knowledge dilemma and thegeography of innovation In Institutions and Systems in the Geography of In-novation Springer pp 35ndash54

Fraunhofer IMW (2018) Knowledge - the most valuable resource of the future EdFraunhofer Center for International Management and Knowledge EconomyIMW Available via httpswwwimwfraunhoferdeenpressinterview-annual-reporthtml (accessed on 16 April 2018)

Galunic Charles Rodan Simon (1998) Resource recombinations in the firmKnowledge structures and the potential for Schumpeterian innovation InStrat Mgmt J 19 (12) pp 1193ndash1201

Gilsing Victor Nooteboom Bart Vanhaverbeke Wim Duysters Geert van denOord Ad (2008) Network embeddedness and the exploration of novel tech-nologies Technological distance betweenness centrality and density InResearch Policy 37 (10) pp 1717ndash1731

12