FROM INNOVATION PROJECTS TO KNOWLEDGE NETWORKS: CONTINGENCIES IN THE SECTORAL ORGANISATION OF

INNOVATION

Fernando Perini f.perini@sussex.ac.uk

SPRU, University of Sussex The Freeman Centre

Brighton BN19EQ UK DRAFT

Paper prepared for the IRNOP VIII conference – Brighton - UK

Abstract

This paper examines quantitatively three contingencies connecting the knowledge-base of innovation projects and the organisational characteristics of the network emerging among firms and their technological partners such as educational and research institutes. The following three propositions are tested empirically: (i) the boundaries between companies and technological partners are influenced by the type of knowledge base and the availability of disperse resources in the knowledge network; (ii) different types of knowledge-base require different governance mechanisms resulting in long-term specialisation in the knowledge network; and (iii) specific types of knowledge base require different types of inter-organisational linkages limiting the possible knowledge flows to specific communities of practice. The three propositions are respectively tested using (i) a longitudinal examination of the boundaries between firms and technological partners in different types of innovation projects followed by an ANOVA test, (ii) a project based index of revealed technological advantage (PRTA) and (iii) a social network correlation technique (QAP).This analysis uses an exclusive dataset developed from 10,088 innovation projects performed by companies under the Brazilian tax incentives to innovation activities in the sector (“ICT Law”) between 1997 and 2003.

Key-words: sectoral innovation systems, knowledge network, organisation of innovation, social networks, tax incentives to R&D, innovation projects

Acknowledgements: This paper benefited from insights provided by Joe Tidd, Martin Bell and Elisa Giuliani, Peter Gammeltoft,, Nick von Tunzelman and Carlos Sato. Usual disclaimers apply. The author is grateful for the financial supported provided by the Programme AlBan, European Union Programme of High Level Scholarships for Latin America, identification number E03D16012BR, and the institutional support of the ABDI and SEPIN/MCT that made this project possible.

2

Introduction

Knowledge networks became a fundamental way to research and understand inter-organisational learning in sectors (Tidd 1997; Malerba 2002; Acha and Cusmano 2005; Giuliani and Bell 2005; Owen-Smith and Powell 2005). The knowledge networks are considered (i) a new source of competitive advantage where relevant knowledge is created in the interface between firms, universities and research institutes (Arrow 1962; Gibbons 1994); (ii) a source of social cohesion and a way to unleash entrepreneurial capacity between established and new firms (Walker, Kogut et al. 1997; Burt 2000); and (iii) ultimately, these complex networks would be a requirement for those nations willing to develop dynamic comparative advantage in sectors inside the global economy (Nelson 1994). Scholars have used joint venture, surveys on university-industry links, patents citation and co-authorship in publications as ways to explain the process of specialization and formation of comparative advantages among firms, sectors and countries (see Meyer 2002 for review). However, despite its key role in the organisation of innovative activities, detailed studies on networks formed around innovation projects are particularly rare.

This paper contributes to the line of enquiry on the nature of the knowledge networks discussing how the knowledge base in innovation projects may influence in the patterns of horizontal collaboration in sectorsi. Although collaboration in projects is not the only mechanisms of organisational learning in sectorsii, they are increasing important for companies, governments and technological institutes giving its dynamic organisational nature that allows a balance between exploitation of technological niches and exploration of new opportunities (Davies and Hobday 2005). More and more, innovation projects are used as a key organisational unit for decision-making and mechanism of collaboration in knowledge-related activities by companies, technological institutions and governments. Nevertheless, comparing and codifying transactions and contracts involved in projects in sectors is very complex and resulting in a lack of empirical databases that would help our understanding on the creation and evolution of knowledge networks.

Inspired by evolutionary approaches to describe the co-evolution between technology, industrial structure and supporting institutions, we use detailed data on 10,088 innovation projects to operationalise the elusive concept of knowledge; in addition, economic transactions inside innovation projects are used to observe the even more ambiguous concept of inter-organisational knowledge flow inside sectors. Although using economic transactions in innovation projects as a proxy for knowledge flows in sectors is certainly a simplification, it certainly offers many advantages in relation to its counterpart measures, such as patents, citations and surveys. The use of innovation project level data to analyse the content of the flow in knowledge networks avoids any assumptions that knowledge flow (or leakage) among firms may happen in the air, that it is costless or that it may happen as a by-product of commercial transaction. It reinforces the idea that intentional flow of codified types of information as well as tacit knowledge embedded in people and constructed in organisational routines is a requirement for organisational learning in firms and sectors. Innovation projects may be seen as a key mechanism where interaction among organisations takes place and relevant knowledge for the parties is constructed and transferred, providing a useful way to examine how the governance of the innovative activities occurs in sectors.

3

In specific, this paper proposes and tests different ways how the type of innovative activities influence the (i) the boundaries between firms and technological partners in different knowledge related activities, (ii) the process of specialisation in different governance mechanisms and (iii) the nature of the knowledge flows occurring between different sets of actors. This paper distinguishes between 10 different innovation-related types of projects in the ICT sector: laboratory and equipment infrastructure, technological training, technological services, R&D quality systems, process technology, product development in software, middleware, hardware and semiconductors as well as research activities. These three propositions are respectively tested using (i) a longitudinal examination of the boundaries between firms and technological partners in different types of innovation projects, (ii) a project based index of revealed technological advantage (PRTA) and (iii) a social network correlation technique (QAP).

This paper draws upon an exclusive dataset containing details of the innovation projects conducted by national and multinational manufacturing companies as well as educational and research institutes in the Brazilian ICT sector and declared under the Brazilian ICT Law between 1997 and 2003. After decades of important substitution policies, the market was opened to foreign investments during the early nineties. The previous regulation in the Brazilian ICT sector was substituted by tax incentives to the commercialisation of a set of industrialised products in the internal market conditioned to local manufacturing and investments in R&D. R&D offset scheme implemented in the sector promoted an overall private investment of more than $2 billion dollars in innovation during the last decade involving more than 200 companies as well as 200 universities and research institutes. The ICT Law became one of the pioneering projects for the development of sectoral innovation systems in Latin America after its liberalisation policies. From the empirical perspective, the paper sheds new light upon the structure that co-evolved from technological opportunities, organisations and institutional changes in the Brazilian ICT sector in the period.

The structure of the document is as follows. The first section briefly defines the how the knowledge network is defined and used in this paper, followed by the description of the contingencies (propositions) between the different types of innovation projects and specific characteristics of the knowledge network, namely the boundaries of between firms and technological partners, governance mechanisms and inter-organisational knowledge flows (section 2). The third section describes the characteristics of the database of innovation projects in the Brazilian ICT sector, and the main characteristics of the sectoral knowledge networks. This is followed by the details of the methods used in the investigation of the hypothesis considering some of the limitations of the research design and the results of the tests. The last section summarises the empirical findings and some preliminary implications for firm strategy and institutional design of sectoral networks.

1. On project-based knowledge networks

Given the widespread use of the term ‘knowledge network’ to discuss different interactions between actors, the natural first step is the clarification of what is meant by the term and underlying assumptions. The term “knowledge network” can be used as a metaphor to represent the complexity of the innovation process (DeBresson and Amesse 1991; von Tunzelmann 2004), a middle of the way between market and hierarchy (Powell 1990) or even to describe the fundamental nature of the firm and all the economic activities (Coase 1937; Williamson 1985). In contrast, the same term “knowledge network” has also been used as an attempt to employ new methodological

4

tools for the analysis of the interactions among agents (Wasserman and Faust 1994; Powell, Koput et al. 1996; Pyka and Küppers 2002; Malerba 2005). Social network analysis has emerged recently as one of the most promising tool for the analysis of the knowledge flows in innovation studies where specific rules/ norms or institutions would allow defining the boundaries of the observable network, its participants and the scope of their activities.

First of all, as previously discussed, the term as used in this paper is limited to the quantitative examination of networks formed by innovation projects. Naturally, the evolution of industrial sectors under evolutionary principles embraces a wider range of activities, basing its foundations on the concepts of capabilities. By focusing on innovation projects, the analysis emphasises tacit knowledge creation and flows rather than traditional market-related transactions in direct acquisition of products and services (Bell and Pavitt 1993; Giuliani and Bell 2005). By delimiting the analysis to innovation projects, the analysis focuses on relatively dynamic capabilities, as projects are by definition characterised by their relative uniqueness and defined time span.

A second aspect is a necessary distinction of the level of aggregation inside the observed networks. Even the term “project-based network” could refer to different levels of aggregation such as: individuals in the labour market (Granovetter 1973); among groups inside an organisation; as well as in vertical (Hardstone 2004) or horizontal relations (Acha and Cusmano 2005) inside an industryiii. In this paper, there is focus on the latter, and in specific, the innovation activities occurring inside and among multinational and national manufacturing firms and technological counterpart such as educational and research institutes in the sectoriv. However, as in any study in social sciences, any bounded network these interactions are not self-contained (Giddens 1979) and the behaviour of agents inside the bounded network is always influenced by its relations with other actors outside the delimited network. For instance, the interaction with foreign organisations and institutions, clients and a wider range of stakeholders would influence the behaviour and decision –making of agents in specific sector. Therefore, the analysis of the behaviour of the bounded-network, in terms of locus, sector technology and/or institution, should not ignore the possible influence of the unobserved networks over the network examined (Malerba 2005). This is especially important in decentralised networks as the usual coherence provided by a leading organisation (a company or other form of associations) is missing or tenuous.

Third, it is necessary to distinguish between normative and operational aims inside of a knowledge network. While most times a knowledge network would involve some sort of formal aims that would define the network (e.g. an institutional set associated to regional/sectoral development, etc), individual agents may operate under a different set of strategic aims and under a different set of organisational principles. Therefore, knowledge networks are not necessarily formed by relatively homogenous entities that are coordinated together and, in fact, heterogeneity will be a key characteristic of a complex sectoral network, resulting in a wider diversity of aims in individual organisations (Dosi, Winter et al. 2000). Rather than profit maximising, the organisational development is influenced uncertainty and path-dependence - result of the endogenous formation of the bounded networks defined by previous interactions in innovation projects as well as the wider range of organizational and social structures (Simon 1979). The underlying complexity of interactions would mean that individual organisations would pursue specific operative aims that are not necessarily aligned with wider formal sectoral aims proposed for the network.

However, despite accepting historical and sectoral heterogeneity as a fact, this paper does not assume that there is no possibility of some generalisations in the evolutionary

5

dynamic happening in knowledge networks. Careful consideration should allow extracting common emerging configurations inside these project-base knowledge networks (Meyer, Tsui et al. 1993) and further research on these configurations is fundamental to identify and examine patterns of similarities and important differences in the organisational development.

2. Relationships between the knowledge-base in innovation projects and firm boundaries, governance and inter-organisational knowledge flows

This paper investigates how different types of innovative projects performed by actors in a sector and the organisation of the sectoral knowledge network co-evolve. This section discuss three propositions on how the relationship between these dimensions.

Proposition 1- The boundaries between companies and technological partners are influenced by the type of knowledge base and the availability of disperse resources in the knowledge network.

Networks are a way to avoid the dichotomy between market and hierarchy in the analysis of industrial organisation as well as key method used by firms to learn and incorporate external knowledge. However, networks do not emerge naturally. They are a result of the companies’ needs to balance their internal growth with the resources available outside the firm (Pavitt 2001). Although some network theorists tend to focus on the structure of agents inside the network structure, the agency and aspects of its decision making is the crucial factor in determining many of network attributes (see Mizruchi 1994 for review). In specific to the case examined in this paper, the “make-or-buy” decision in innovative projects would determine fundamentally the characteristics of the knowledge network formed. An investigation of the type of activities internalised and outsourced in innovation projects therefore provide a necessary first step in the examination of the bottom-up evolution of the project-base knowledge networks and its characteristics.

The first proposition argues that two key elements would significantly influence the boundaries found between firms and possible technological partners in project based knowledge networks: the type of the knowledge activity and the availability of external resources. Following a transaction cost perspective (Coase 1937; Williamson 1985), firms will be especially interested in developing in-house certain type of innovative activities where the costs of searching and identifying appropriate partners and developing and enforcing appropriate contracts are high.

According to the traditional market failure argument, firms tends to use external sources of knowledge in innovative activities when they cannot fully appropriate from their investments. Activities such as long-term research, training and other infra-structure elements would have the attributes of public goods, as several companies may access the qualified human resources using the labour market and infrastructure or information services provided by universities and research institutes. The social benefits derived from these activities would provide a fundamental rationale for government intervention as the use of these widespread resources would not have a significant diminishing effect on its value (non-rivalry argument) and individual agents should not prevent other companies from using these public goods (non-excludability argument).

The market failure rationale is however the focus of considerable criticism. Given the increasing interactive nature of knowledge creation in sectors the distinctions between producers and users of knowledge are increasingly fuzzy (Geuna, Salter et al. 2003). In

6

many sectors, and in particular ICT, the boundaries between different agents are blurring given the extensive need for inter-organisational linkages (Antonelli, Geuna et al. 2000). The governance of the innovative activities becomes increasing diverse and complex in order to coordinate the knowledge flows between different public and private agents as well as firms and technological institutes. Project based knowledge networks would need to mix the ability to accumulate internal capabilities inside firms with governance structures that would allow a wider reconfiguration of the capabilities inside a network of organisations.

Resource-based theorists complement this discussion arguing that the internal importance of the accumulation of technological capabilities inside organisations are a necessary way of identifying possible technological opportunities and possible technological sources (Teece 1994; Penrose 1995). Capabilities outside the firm would not be a substitute for internal capabilities as organisational learning would allow for economies of repetition and the formation of comparative advantages (Brusoni, Prencipe et al. 2001). Balancing accumulation of internal capabilities and exploitation of external sources is at the centre of the firm technological renewal and diversification.

An adequate portfolio of innovation projects is an essential organisational way to achieve this balance (Lam 2005; Manning 2005). Although companies will tend to integrate vertically the activities that provide them with comparative technological advantage, they also will need to integrate with external sources as new opportunities arise, interacting with other groups and firms. The formation of developed network result from the need to adapt to new technological opportunities and integrate new sources of knowledge, forcing companies not just to compete and cooperate in the industrial structure (Miles and Snow 1992). Large agents (both private and public) would create the availability of resources that could be used by entrepreneurs in new innovation projects (Audretsch 1995; Cantwell 2001). While incumbents will tend to explore their dominant position in existing network, changes in technological opportunities will mean that managers need to adapt and integrate new sources of knowledge technologies, inter alia, because of the cost and time necessary to develop new capabilities in-house.

The proposition points out that the capabilities in the project-based knowledge network are not only driven by exogenous factors (e.g. technology created in/absorbed from universities and research institutes), but also (and possibly primarily) by the endogenous differentiation the capabilities accumulated by firms in the industrial structure (Nelson 1994; Gulati 1999; Gulati and Gargiulo 1999).

Proposition 2 - Different types of knowledge-base require different governance mechanisms resulting in long-term specialisation in the knowledge network.

This proposition is based on the assumption that sectoral knowledge networks are formed from a decentralised pursue of innovation projects, rather than a top down/formal definition of the role that should be performed by individual actors in a sectoral innovation system. Different organisations identify potential technological niches and promoted by the economic imperatives such as the costs of specific types of knowledge-related activities (Nelson 1993; Edquist 1997).v In contrast to a normative definition of the functions/technological areas of individual organisations inside the sector, specialisation happens mainly as result of interaction among actors that compete in the marketplace as well as collaborate in different forms of governance structures.

Organisations in modern sectors specialise and compete in order to develop comparative advantages in innovative activities (Pavitt 1998). In specific to the case of project organisations and networks, economies of repetition and recombination are key in developing long-term comparative advantage (Davies and Brady 2000). Intra and inter

7

organisational governance would emerge as organisations draw on knowledge developed in previous innovation projects and integrate available capabilities for new opportunities (Prencipe and Tell 2001; Engwall 2003). The variety of case studies shows that these governance structures are necessarily and increasingly complex and diverse (Leonard-Barton 1987; Hobday 2000). Strong hierarchical ties inside firms are required to appropriate from the technological capabilities in the exploitation of technological opportunities (Gann and Salter 2000). Subsidiaries of multinational companies would make use of a different set of organisational attributes to facilitate the adoption, creation and diffusion of knowledge inside the organisation (Ghoshal and Bartlett 1988; Chiesa 2000). Meanwhile, a wider governance, connection universities and research institutes, would be necessary to share common infra-structure and identify long-term technological opportunities in regions (Fombrun 1986; Powell, Koput et al. 1996).

Different governance structures emerge in the knowledge network allowing for the formation of positive externalities and structures with specific aims co-evolve with the technological basis and develop relative comparative advantages inside the sector (Grabher 2004; Grabher 2004). This specialisation process would be then be reinforced, as other agents identify productivity gains in collaborating with more efficient organisations in specific activities. At the same time, as specific agents developed their position inside the project-based network, their exploitation of a dominant position may limit the opportunities to new actors to position themselves in similar niches.

Empirical analysis of the process of specialisation in project-based knowledge networks becomes important given complementarities between different activities in the innovation process. Identifying the process of specialisation in the distributed governance structures in sectors could shed light into the source of leadership or look-ins in the evolving technological trajectories.

Proposition 3 - Different types of knowledge base require different types of inter-organisational linkages limiting the possible knowledge flow to specific communities of practice.

In order to investigate the nature of knowledge networks, the identification of the type of knowledge flowing among actors is paramount. The position of the agent in different networks would provide them with a strategic advantage and the knowledge flow is influenced by the costs of codification and diffusion of knowledge as well as the conflict of interests in the process of searching and following technological opportunities (Teece 1989; Cummings and Teng 2003).

The reused of knowledge will be influenced by norms and practices developed in previous relations as well as the power of receptors to bending rules in their advantage along the dynamic development of the network (Dyer and Singh 1998; Lane and Lubatkin 1998; Knight 2002). As technological opportunities change over time, organisations need to recombine the accumulated knowledge or “transfer” new types of knowledge inside and outside the organisational boundary (Hansen, Mors et al. 2005). Ultimately this will result on diverse communities where specific types of knowledge is diffused (Brown and Duguid 2001; Swan, Scarbrough et al. 2002).

Different stakeholders will be involved in influencing existing practices and developing knowledge flows in specific directions. Given the existence of imperfect information, uncertainty and transactional costs, new innovation projects will be developed where agents perceive technological opportunities inside their existing social network, influenced by the specific industry dynamics, locations and previous capabilities (Dosi 1988; Cimoli and Dosi 1995).

8

Therefore, knowledge is not diffused evenly throughout the project-based network. Identifying the overlapping structures of specific communities of practice in different types of innovative activities allows detecting the number of actors participating in multiple networks (gatekeepers) (Tushman and Katz 1980) and which kind of organisational channels are used for the transfer of similar types of knowledge. The role played by the actors involved in different networks can determine significantly the knowledge flows in the sector. Their behaviour becomes fundamental on explaining the evolution of the network and translation of knowledge along different steps in the decentralised innovation process.

3. The knowledge networks under the Brazilian ICT Law: general characteristics of the database

A possible way to define the network is in terms of the broad institutions, the actors, ties and the content of the interactions (Malerba 2005). This paper delimits its analysis to the network formed by innovation projects declared under the tax scheme developed in the Brazilian ICT sector, called ‘ICT Law’. The tax scheme defined R&D obligations proportional to sales in the national market in exchange to different types of tax exemptions/waves in products of manufacturing companies. In order to be entitled to the tax scheme the companies were obliged to invest approximately 5% of their national turnover in innovative activitiesvi.Ex-post, these activities conducted should be described into projects and in turn audited by the regulatory governmental agency (SEPIN) connected to the Brazilian Ministry of Science and Technology.

Through a collaboration agreement with the Brazilian Ministry for Science and Technology, the database of projects used for administrative purposes was codified for this research. While keeping the confidentiality requirements of the contract, this research uses the normalised procedure for collecting data from the companies as a way to explore the the relations between the types of projects and the organisation of the knowledge networkvii.

In this unique database of projects, we use some information of the executor of individual projects to identify the process of knowledge creation in different organisations and transactions among firms and technological partners discussed to identify the process of inter-organisational knowledge flow. In terms of projects, the dataset contains 10,088 projects executed under the Brazilian ICT Law between 1997 and 2003 (an average of 1261 per year). The projects sum up an amount of R$ 1.6 billions executed internally to the companies and R$ 1,1 billion executed in partnership with universities and technological institutes (annual average of R$358.1m) (see Table 1) .

Table 1- Longitudinal distribution of the projects

Total 1997 1998 1999 2000 2001 2002 2003 Average Total

Investments (''R$) 304.3 346.8 389.5 560.4 249.6 349.4 306.3 358.1 2864.4

Number of projects 1194 1381 1439 1741 783 1235 1055 1261 10088

Average project size ('R$) 2421.5 2738.8 2907.5 3868.1 4555.0 4818.0 8799.7 3665.5 33774.1

Equiv. Staff/FT * 2637.2 2823.0 2666.2 3582.1 1535.3 2090.1 1563.6 2355.2 19252.6

* Estimate number of full-time staff (direct + indirect HR costs)/(Average Cost Man/Hour*2000)

In terms of actors, the dataset involves 211 companies and 181 educational and research institutes inside the Brazilian ICT Law for the period 1997-2003. These actors are located in the entire Brazilian territory with the exception of the Manaus Free-Trade Zone, which receives specific incentives to manufacture and for R&D activities.

9

In terms of ties, the knowledge flows are developed based on more than 35,000 transactions inside the projects, creating 948 ties between these 392 nodes. In terms of activities involved in the ties, the network could be divided according to the classification used inside the procedure, namely investments in laboratory and infrastructure for S&T, quality systems for R&D, training in S&T, technological services, development of products in hardware, software, semiconductors, middleware, production process, as well as research activities. Table 2 shows the distribution of ties according to type of activities. It also summarises some basic statistics about the network in terms of investments in projects, the number of firms, the number of ties, and the strength and density of the network divided by the different activities. The table also contains some details about the density and concentration in the different network.

Table 2 - Descriptive Statistics about the ‘ICT Law’ Knowledge Network - 1997-2003

Dimension

Infr

astr

uctu

re

Qua

lity

Tech

nolo

gica

l Se

rvic

es

Trai

ning

in S

&T

Sem

icon

duct

ors

Prod

uctio

n P

roce

ss

Har

dwar

e

Syst

em

Softw

are

Res

earc

h

Sum of Investments ('' R$)

169.7 118.2 84.7 159.5 44.7

108.9 203.4 621.7 838.3 121

(with partners) 103.7 27 65.8

100.4 4 13.5 46.3 212.4 385 97.2

Number of firms 142 170 104 177 30 140 191 234 271 195

(with partners) 64 67 76 87 15

44 81 127 157 111

Number of Partners

96 52 71

117 18 54 71 92 140 121

Number of Ties 174 120 162 240 22 90 141 230 425 304

(>R$ 1M) 18 5 12 20 1 3 8 31 56 23

Tie strength (' R$) Average

570 174 387 388 189 145 304 799 830 309

Tie strength (' R$) Maximum

11584

3349 20957

28565

1427 1818 7300 28188

58622

9229

Density 407 105 258 394 17 53 181 834 1512 381

Concentration – (10-Firm Ratio)

73% 53% 72% 70% 99% 63% 63% 64% 70% 65%

Concentration -(5-Firm Ratio)

42% 36% 51% 51% 97% 47% 48% 49% 45% 48%

Concentration – (3-Firm Ratio)

26% 26% 40% 41% 95% 38% 34% 36% 29% 37%

Concentration – (1-Firm Ratio)

9% 10% 25% 25% 72% 23% 13% 16% 12% 18%

Foreign companies represented 72% of the total investments, while domestic companies sum 28%. The concentration is especially high among the top 20 companies. The top 20 represent 73% of the total investments. From these 20 companies, 16 are subsidiaries of foreign multinational companies, responding for 64%. Similar concentration can be observed among the receivers of investments. In relation to the proportion of the resources allocated to technological partners by the firms, approximately 60% of the total investments went to Private Research Institutes, followed by private educational institutes (18%), public research institutes (12%) and public educational institutes (9%).

10

Table 3 - The knowledge network under the Brazilian ICT Law

Dimensions Description

Institutions

Perform-conditioned tax incentive - it conditioned tax benefits on the products commercialized by the company in Brazil to minimum investments in innovation. Companies were required to invest approximately 5% of the national sales in innovative activities (2.3% need to involve a research and/or educational institute).

Actors 211 domestic and foreign firms manufacturing products under the incentives

180 technological partners - educational and technological institutes that attended the regulation requirements

Ties More than 30000 transactions inside innovation projects conducted inside collaborative agreements between firms and technological partners

Content Type of innovation activities allowed under the incentives. Internal costs and transactions among partner were classified using the following categories: laboratory and infrastructure for S&T, quality systems for R&D, training in S&T, technological services, development of products in hardware, software, semiconductors, middleware, production process, as well as research activities. viii

The definition of the type of activities are connected to the definition used in the standard procedures, namely investments in laboratory and infrastructure for S&T, quality systems for R&D, training in S&T, technological services, development of products in hardware, software, semiconductors, middlewareix, production process, as well as research activities. This categorisation at project level represents an advantage in terms of defining the knowledge base independently from the final product classification (.e.g. Pavitt taxonomy, most of the sectoral system studies). In other words, it allows the existence of multi-technology firms and avoids ambiguity in the definition of the boundaries of the sectoral systems under analysis.

The nodes of the network are companies and their ‘technological partners’. The companies are restricted to national and multinational companies with local manufacturing of products under the incentives (usually products that integrate advanced electronics, such as computers, mobiles and telecommunication equipments). The regulation also defined that part of the investments (approximately 40%) should be conducted with educational and/or research institutes (henceforth, ‘technological partners’) in an explicit attempt to promote university-industry linkages. These partners were especially important in the regulation that aimed to reinforce these organisations as the key nodes in the sector. The database of projects contains details on the costs in innovative activities both inside companies and in technological partners. As the regulation does not define the type of activities that should be conducted inside the firm boundaries or with partners, this database provides a useful source for investigating the firm “make-or-buy” decision-making.

The ties of knowledge network are formed by the transactions inside the innovation projects. In order to operationalise knowledge flows, this paper uses 35000 transactions that involved the collaboration agreements between firms and educational/technological institutes inside the scope of the framework. They formed more than 948 ties among the nodes. There were also transactions with other companies creating a wider, open network (commercial software companies, suppliers of equipments and training abroad

11

and other organisations not classified as ‘technological partners’ inside the network). However the analysis of these transactions would add another layer of complexity and therefore they are beyond the scope of this paper.

The Figure 1 provides a visual representation of the knowledge networks divided by the different activities. Companies are represented as circles and technological partners as squares. Domestic companies are represented in white, foreign companies are represented in blue, Educational institutes are in red and research institutes in black. The diameter is proportional to the sum of innovation projects conducted by the specific organisation during the period between 1997 and 2003.

Figure 1 – The knowledge networks in the Brazilian ICT sector divided by type of activity - 1997-2003 - complete

Research Networks

Laboratorial Infrastructure Network

Training in S&T Network

Quality Systems Network Technological Services Network

12

13

Semiconductors Network

Production Process Network

Hardware Network

Middleware Network

Software Network

Source: Based on MCT/SEPIN data using NetDraw 2.37. (Borgatti 2002)

14

Before entering into the characteristics of these networks, it is relevant to observe the relevance of the innovation projects database compared to the total investments in innovation in the Brazilian Telecommunications and Computers sector. One way to proceed is to compare the results with an external measurement of the total investments in R&D conducted by these two sectors. The total investments in R&D in the telecommunications sector and the computer sector by private companies as assessed by the PINTEC (Brazilian innovation survey) were R$627m in 2000 and R$637m in 2003 according to the two innovations surveys conducted in the Brazilian ICT sector (MCT, 2006). In addition, the innovation survey estimated that the total outsourcing of R&D was R$153.9m in 2000 and R$184.2m in 2003. From these numbers, it is possible to estimate that the SEPIN database contains in average more than 55% of the investments in R&D in the computer and telecommunications sector (the average annual investments inside the ICT Law was R$386m for the entire period). In addition, more than 85% of the innovation projects outsourced occurred inside the regulatory frameworkx.

Although there are some differences in the concept used to classify R&D in the two databases, the general number obtained through these two different databases show two general remarks about the dataset: (i) There is possibly more R&D activities inside companies in the sector (that contain a much larger sample, such as software companies and services that do not have a manufacturing production systems with products/ minimum standards required by the regulation), although the amount of projects in the dataset is a important proportion. (ii) almost the totality of the outsourced R&D in the computer and telecommunication sector was conducted inside the regulation. Therefore, in general, we assume that the project and ties pointed here do provide an important measurement of the investments that the companies would conduct inside the limits of the sector under analysis.

4. Methods and empirical results

The methods used for the investigation of each one of the proposition are presented in Table 4.

15

Table 4 - Proposition and description of the variables and procedures

Proposition 1- The boundaries between companies and technological partners are influenced by the type of knowledge base and the availability of disperse resources in the knowledge network.

The different networks are analysed using the trend (two years average) for the investments in the different types of knowledge-related activity and the locus of execution of the projects (firms or technological partners). The significance of the differences between the vertical integration in different activities is verified using an ANOVA test (Annex 1).

The type knowledge-related activity is defined as the total investments in specific innovative activity in relation to the total investment. Managers classified individual projects among infrastructure to R&D, technological services, training, hardware, middleware, software, semiconductors, process technology, other types of product development, or research.

Vertical integration is defined as the total investments in internal projects in contrast to projects assigned to technological partners. In each project, managers assigned the organisation responsible for the project management.

Proposition 2 - Different types of knowledge-base require different governance mechanisms resulting in long-term specialisation in the knowledge network.

A specialisation index was adapted from the revealed technology advantage (RTA) indexxi. In our case, we use the value of projects conducted by the organisation to arrive to the project-based revealed technological advantage (PRTA) index calculated for the different types of agents ( “types of governance mechanism”) for the different types of knowledge activities. The project-based specialisation index (PRTA) could be defined as:

Type of governance mechanisms (g) are divided among foreign companies, domestic companies, public and private research institutes, public and private educational institutes. Pij is the costs of project executed by organisational type i in knowledge related activity j. As in the traditional RTA, values greater than one suggest that a organizational type is comparatively specialised in the innovative activity in question relative to other organizational types (as it conducted more projects in this activity than the general average for the group), while values less than one are indicative of a position of comparative disadvantage. This procedure would allow to control for the general concentration of specific organisations as well as the rules that define broader proportions that should be spent in companies and technological partners.

Proposition 3 - Different types of knowledge base require different types of inter-organisational linkages limiting the possible knowledge flow to specific communities of practice.

In order to text this hypothesis, a correlation is used to analyse the interdependence between the structures of the different networks. In specific, the Quadratic Assignment Procedure (QAP)xii is used for all the possible combinations between the 10 knowledge networks. Each network structure is represented by a valued matrix (Aikk), where i is the type of activity and k is the number of organisations in the network . In this case, k is contant and equal to 398 as there are 211 firms and 180 technological partners in the network. The values of these networks are the sum of the transactions among partners (i.e. valued network). These values were normalised using natural logarithm in order to obtain a normal distribution among the technological partners (Annex 1). The mathematical procedure could be defined as:

xiii

The result of the correlation is a matrix (Xkk) containing the strength of the overlapping between each pair of networks.

5. Results

16

Proposition 1- The boundaries between companies and technological partners are

influenced by the type of knowledge base and the availability of disperse resources in the

knowledge network

The first proposition is related to make-or-buy decision making process in innovative activities. In order to explore the dynamic of the formation and interaction among the knowledge networks, Figure 2 shows the trends in the accumulated technological capabilities in the different technologies (estimated based on the percentage of the total investments) and the degree of vertical integration of the innovative activities (based on the number of projects controlled by the company compared to those outsourced to technological partners). In the development of new products, the technological opportunities identified by the companies have changed considerably during the period, mainly from middleware to software. Clearly companies inside the framework identified limited opportunities in microelectronics, hardware and production process.

First of all, Figure 2 reinforces the visual inspection of the networks in Figure 1: (i) there are basically no/or very incipience networks related to semiconductors, production process and hardware. (ii) wider networks with relative weak ties were formed in activities such as training, technological services and research. (iii) there are strong-ties networks in to middleware and, most of all software, where considerable governance mechanisms could be expected through the technological partners. Secondly, these three groups are significantly different in terms of boundaries of the firm in innovative activities and accumulated capabilities, as supported by the an ANOVA test (Annex 1). These three groups could be further described.

Enabling networks (Low levels of investments, low vertical integration) – Following the “market failure” argument, the group of points in the right bottom refers to activities such as training in science and technology, technological services (e.g. metrology, certification) and research activities. Investments inside innovation projects tend to be smaller and close to market mediated. Only a smaller part of the investments in technological services (36%), training in S&T (45%) and research activities (22%) were conducted internally. To some extent, investments in infrastructure and laboratories also could be associated with this group although they have a less significant difference in terms of firm boundaries (46%).

These-ties with external organisations are also weaker. In fact, if the total investments in these activities and the number of ties, we would have of 1 tie per million in Infrastructure and laboratories, 1.6 ties/million for training, 2.1 for technological services and 2.5 for research activities. These numbers contrast significantly with averages of 0.4 to 0.6 tie per million invested in the other “product development” networks (Figure 3).

17

Figure 3 - Number of ties per million in different technologies

Although weak ties are usually assumed to be related to research projects (as companies would look for technologies opportunities based on these type of activities), it is clear that these other networks also have a fundamental connection between firms and a large number of different supporting organisations. These ties would be important for the development of human resources, technological information, etc. Some of these supporting organisations (providers of training, technological services, infrastructure, etc), sometimes neglected inside the innovation studies, need to be understood in more detail as they could not be assumed as available in most developing countries.

Developing networks (Low levels of investments, high vertical integration)- The points at the bottom left are mainly composed of three groups of innovation activities in different time periods: product development using hardware, semiconductors, production process technology as well as quality systems. An important aspect of these three arrows is the strong internalisation of the technological investments inside firms. Although linkages could exist with international partners or other stakeholders, there are very limited horizontal connections with partner technological institutions. This indicates that in these networks the companies show resilience to use external sources of technology. This analysis reinforces the visual analysis (Figure 1) that the formation of disperse governance mechanisms have been limited in these activities.

Figu

re 2

- T

he S

ize

and

Bou

ndar

ies

of th

e K

now

ledg

e N

etw

orks

in th

e B

razi

lian

ICT

sec

tor

Sy, 1

997

Sy,

1998

.199

9

Sy,

2000

.200

1

Sy,

2002

.200

3

SW,

1997

SW,

1998

.199

9

SW,

2000

.200

1

SW,

2002

.200

3

IL,

1997

IL,

1998

.199

9

IL,

2000

.200

1IL

, 20

02.2

003

Re, 1

997

Re,

1998

.199

9

Re,

2000

.200

1Re

, 20

02.2

003

TS,

1997

TS,

1998

.199

9

TS,

2000

.200

1TS

, 20

02.2

003

PP, 1

997

PP,

1998

.199

9PP

, 20

00.2

001

PP,

2002

.200

3

HW,

1997

HW

, 19

98.1

999

HW

, 20

00.2

001

HW

, 20

02.2

003

SE, 1

997

SE,

1998

.199

9SE

, 20

00.2

001

SE,

2002

.200

3Tr,

1997

Tr,

1998

.199

9Tr

, 20

00.2

001

Tr,

2002

.200

3

QS,

199

7Q

S, 1

998.

1999

QS,

200

0.20

01

QS,

200

2.20

030%5%10

%

15%

20%

25%

30%

35%

40%

45%

0%10

%20

%30

%40

%50

%60

%70

%80

%90

%10

0%

Ver

tica

l In

tegr

atio

n

Sy SW

IL Re TS

PP

HW

SE Tr QS

2 pe

r. M

ov. A

vg. (

Sy)

2 pe

r. M

ov. A

vg. (

SW)

2 pe

r. M

ov. A

vg. (

IL)

2 pe

r. M

ov. A

vg. (

Re)

2 pe

r. M

ov. A

vg. (

TS)

2 pe

r. M

ov. A

vg. (

PP)

2 pe

r. M

ov. A

vg. (

HW)

2 pe

r. M

ov. A

vg. (

SE)

2 pe

r. M

ov. A

vg. (

Tr)

2 pe

r. M

ov. A

vg. (

QS)

Hie

rarc

hy

Ma

rke

t

Syst

em (

Dev)

Soft

war

e (D

ev)

Infra

& L

abs

Rese

arch

Tech

. Ser

vice

sPr

oduc

tion

Hard

war

e (D

ev)

Sem

icon

duct

ors(

Dev

)Te

ch T

rain

ing

(De

v)Q

ualit

y Sy

stem

(De

v)

There are also very different trends as shown by the arrows. The arrow related to semiconductors show incipient, but increasing, initiatives to accumulate technological capabilities inside the companies. An opposite trend is observed in relation to production technology that has decreased and outsourced activities. The arrow and the dots related to hardware show that there is an upward movement, although it has been turbulent throughout the period, probably given the instability in the initiatives undertaken by different companies in this type of technology.

Developed networks (High levels of investments, intermediate vertical integration) - A different portrait can be developed around the dynamic involving in the two largest networks: the networks formed by product development projects using middleware and software technology. They both are characterised by higher levels of investments and an intermediate level of desegregation of the activities between hierarchies and partnerships.

The analysis of these trends over time shows that the development of the network evolved in opposite directions. From this trend, we can imply that the established and newcomer companies have shifted their investments from middleware to software along the period. In the middleware network, while the investments were reducing in middleware technology, companies tended to retain internal projects, rather than consolidate. At the same time, the companies which were increasing their investments in software identified existing capabilities available in partners and the general vertical integration decreased.

In the two large areas of investment (software and middleware), the data implies that there was a considerable scope for governance structures with strong ties among partners. These lower level of vertical integration in relation to other types of product development knowledge networks support the hypothesis that as there is increasing resources available in the network, governance mechanisms would tend to emerge and integrate disperse resources. When these capabilities decrease - as in the case of middleware -the level of vertical integration tends to increase simultaneously. This support a resource-base view, where the evolution of the knowledge networks in product development are mainly connected with endogenous differentiation among companies. As these investments are especially important in project-based knowledge networks giving their higher magnitude, the accumulation of capabilities inside firms seems to be indeed the key driver of the network evolution.

The number of ties per total investment is significantly lower in product development when compared to training, technological services and research activities. It suggests that firms have fewer, but strong, ties in product development, while, companies will also tend to have more and weaker ties in relation to technological services, training and research activities.

Proposition 2 - Different types of knowledge-base require different governance

mechanisms resulting in long-term specialisation in the knowledge network.

A next step is to expand the analysis from the simply bilateral relation in terms of vertical integration, to the analysis of the division of innovative labour in the entire network. In order to understand how the process of specialisation happened inside the network, Table 5 shows the measurement of the project-based technological revealed advantage (PRTA), for the different organisational mechanisms.

20

Table 5 - Revealed Technology Advantage of the different organisational mechanisms

Specialisation Index

Count Research

Hardware

Software

Semicond

Middleware

Process

Training

Tech Services

Quality

Infrastructure

Foreign companies

51 0.25 0.96 1.08 1.77 1.06 1.79 0.72 0.42 1.26 0.83

Domestic companies

163 0.58 2.18 0.69 0.97 1.38 0.77 0.61 0.49 1.64 0.53

Private Research Institute

47 1.33 0.41 1.17 0.14 0.77 0.22 1.46 2.39 0.43 1.31

Public Research Institute

20 3.08 0.49 0.89 0.29 0.73 0.45 0.65 1.57 0.21 0.51

Private Educational Institute

75 2.15 0.17 1.02 0.02 0.97 0.20 1.25 0.98 0.67 2.26

Public Educational Institute

40 4.98 1.85 0.51 0.56 0.35 0.31 3.39 0.32 0.20 1.55

In the table it is possible to observe patterns of specialisation in the different nodes, therefore identifying how knowledge were associated with different governance mechanisms in the network (i.e PRTA>1). Analysing the results, the following patterns of specialisation emerge in the interrelation among the different networks:

While domestic companies focused their investments in middleware and hardware (as well relatively higher investments in quality systems), foreign companies where predominant in the emerging software. The latter also had important initiatives in the smaller semiconductors and production process network of projects. This provides a strong indication that while domestic companies tend to be more connected to their manufacturing basis, multinational companies tend to be more capable of diversifying into distinct competences in software projects (therefore disconnected from product manufacturing) operating in product development inside the international division of labour.

Among the technological partners, private research institutes became key governance structures in the research activities, software, training, technological services and development of labs and technological infrastructure, while public research institutes became highly specialised in research and technological services. It is possible to speculate on the organisational characteristics that define these differences. This indicates that public funds tended to complement private investments in long term infra-structure and research personnel required for these activities, creating a relative comparative advantage of these organisations. Meanwhile, the governance of these organisations and their policies could be too rigid to adapt to short-term requirements of companies, as private research insitutes became fundamental inter-organisational linkages in the software project-based networks.

Finally, as possibly expected, both private and public educational institutes were involved particularly in research, training and infra-structure. Public educational institutes, a group composed mainly of federal and state universities, were particularly specialised in the research and training areas. Clearly, the public organizations developed their comparative advantage from their traditional role inside the structured national educational system (acting on the traditional market failures related to research and training). Interestingly, these traditional organisations did not tend to diversify into collaborative activities in the new technological areas. The relatively new private educational institutes have a participation in software very close to the average, but it could not be considered that they were “specialised” inside the sector in these activities. Although the public

21

educational institutes kept their relative strength in hardware (a relatively small network, mainly connected to domestic companies), they clearly lag behind in the emerging software trajectory.

This pattern of specialisation sheds new light into the distributed innovation process after the liberalisation of the sector. Recent academic discussions in the sector were heated as authors investigate different patterns. For instance, that the process of liberalisation resulted in decreasing capabilities on the cluster previously concentrated in domestic firms in Campinas (Szapiro and Cassiolato 2003), the active role of policy and multinational equipment manufacturers in the sector (Mani 2004), the dependence of the innovation system in software from multinational companies in Brazil (Stefanuto 2004).The pattern of specialisation described above shows how these different governance structures co-evolved as a result of the mixture of technical change, foreign direct investment and sectoral policies. It could be observed from the table that foreign companies, private research institutes (and to a small degree) private educational institutes could be considered key nodes on the development of the expanding software network in Brazil. Meanwhile, the, mainly led by domestic companies and public research and educational institutes where key nodes in the decreasing middleware network and hardware.

The following proposition examines how these organisations may in fact result in distinct communities in the sector, restricting the knowledge flows among specific agents.

Proposition 3 - Different types of knowledge base require different types of inter-

organisational linkages limiting the possible knowledge flow to specific communities of

practice.

The next level of our analysis is to examnine how the different knowledge flows occur inside and among different project-base knowledge network. Using the payments inside innovation projects as a proxy for knowledge flows among actors in different activities (see Figure 1 for the visualisation of the networks), Table 6 shows the result of correlation among the different valued networks.

Table 6 – QAP Correlation among the knowledge networks developed in different activities

1 2 3 4 5 6 7 8 9 10 11

1- Infrastructure

2 - Technological Services 0.283

3 – Training 0.580 0.173

4 - Quality Systems 0.276 0.197 0.204

5 – Semiconductors 0.035 0.064 0.033 0.057

6 - Production technology 0.403 0.132 0.606 0.181 0.081

7 – Hardware 0.155 0.035 0.101 0.207 0.097 0.217

8 – Middleware 0.262 0.210 0.265 0.416 0.049 0.145 0.137

9 – Software 0.309 0.330 0.449 0.577 0.047 0.274 0.158 0.600

10 – Others 0.473 0.034 0.392 0.410 0.012 0.274 0.314 0.173 0.319

11 – Research 0.334 0.617 0.183 0.358 0.209 0.181 0.242 0.293 0.368 0.217

22

*All the correlations are significant at 0.01.

QAP procedure developed in UCINET 6(Borgatti, Everett et al. 2002)

The first clear result of the correlation is that the knowledge flows inside the project-based networks are not homogeneous. Most of the networks presented in Figure 1 are significantly different from each other as demonstrated by the relative small correlation between the different networks in most of the cases. Different knowledge activities would create significantly different communities of practice that would co-evolve in the sector.

The second set of results with empirical relevance refers to those networks that do have a relative strong correlation. Establishing 0.5 as a threshold to a strong relationship, we could point out just 5 intertwined networks. These intertwined networks could be further grouped into three distinct communities of practice:

(i) Training and Infrastructure/ Training and Production Technology – The analysis suggests companies used the incentives to improve the infra-structure of partners which could also provide training in new technologies. In addition, production technology was also specially related to training in new technologies.

(ii) Research and Technological Services – Other channels became specialised in providing research activities and technological services (metrology) for the companies. It is interesting to note that, in general, research and technological services (possibly centres of excellence in different technologies) were not strongly related with the linkages involved in product and process development.

(iii) Product development in Software and Quality Systems/ Software and Middleware – specific channels became related to the improvement of quality systems in R&D (e.g. CMM certification) and the development of products in software. Here, it is also possible to observe a strong relation between the formation of the capabilities in middleware and software. Although this test does not allow us to attribute causality, the dynamic changes shown in Figure 1 may allow to reinforce the interrelation between the lost its relative importance of middleware-knowledge and the growth of the software network. While the newcomers (specially multinational companies), invested in new opportunities in software, private research institutes became key integrators between ‘old’ and ‘new’ capabilities.

Previous public research institutes and universities were isolated from most of the product development activities. They specialised mainly on research and training activities (weaker ties) while private research institutes developed preferential attachment with large multinational companies while integrating dispersed capabilities in the sectoral network.

Private research institutes became organisational forms interconnecting the large part of the capabilities in the network. However, the extent to which these organisations operate as private organisations or policy networks is certainly a question that needs to be investigated in detail. Although the diffusion of knowledge beyond the initial partners do not necessarily occur for cognitive and/or strategic reasons, these vertically disintegrated technological networks retained capabilities that are certainly important for entrepreneurial companies into new ventures.

23

6. Analysis and implications

The understanding of the roles of different agents and their knowledge-related interactions has been the key challenge for innovation management and policy. This paper contributed to the empirical literature on the organisation of innovation systems exploring the three contingencies between the knowledge-base of innovation projects and the structure and dynamics of the knowledge network. This paper uses longitudinal analysis of the boundaries between firms and technological partners in different activities, a project based index of revealed technological advantage for different governance structures (PRTA) and social network correlation techniques (QAP) to identify some principles on the co-evolution between innovation projects and knowledge networks. The three distinct propositions show that rather than cumulative (same actors going from basic training, to product development, to research activities), the development of the system is multi-dimensional, where different governance structures are involved in specific activities in a process of specialisation and differentiation.

The first proposition states that the knowledge-base influences significantly the boundaries of the innovative activities as well as the endogenous availability of resources inside the network. The results support the proposition that project-based networks emerge more naturally from activities such as research, training, technological services and infrastructure. In these activities, the natural trend would the formation of weak ties with capabilities accumulating mainly outside the firms in universities and public research institutes. (enabling networks). Companies would be willing to retain weak ties with a plethora of organisations in these activities. In the other hand, following resource-based theories, the accumulation of capabilities inside firms, and further endogenous reconfiguration of these capabilities in networks, seems to be the key process occurring in emerging technological trajectories (developing and developed networks). Innovation projects in all developing networks were associated with higher internalisation levels. The vertical integration of the project-based knowledge networks both in middleware and software were inversely related to the resources available inside the network (using the percentage of the total investments as proxy).

The second hypothesis examines the process of specialisation of governance structures inside the project-based knowledge network in different activities. A wider number of organisations coordinated available resources in new opportunities forming complex inter-organisational sectoral governance structures. Multinational companies, connected to private research institutes, were important in the emerging software technology while domestic companies remained connected to hardware and public educational institutes. Public research centres and educational institutions became central in widespread network of weaker ties. The inclusion of a specific companies in the knowledge network (for instance through the association of their products to the tax incentives) and other organisations (through regulatory requirements, for instance) involves trade-offs that will influence the diversity creation and the selection process.

The third proposition examines the patterns in the knowledge flows formed between companies and technological partners. The low correlation between the networks in different activities shows that the different types of knowledge tends flow in distinct communities of practice. There were though some connections between the types of activities developed between companies and technological partners in specific activities. Specially, the three distinct strong correlations emerged: (i) production process and laboratorial infra-structure/equipments were connected with training, (ii) research was

24

correlated with the same partners involved in technological services, and (iii) product development activities in middleware and software were strongly correlated.

These contingencies have implications for those involved organisational studies and institutional design of sectoral policies. The results show that the breath and depth of the definition for innovation project influences significantly the kind of knowledge networks that will emerge. Weak ties or weighty capabilities in strong ties are trade-offs. No type of activity alone will possibly bring the required dynamic in emerging sectors. The limits between private and public investments that will promote the most adequate level of learning are a contingent to the specificities of the existing capabilities in the sector and the technological opportunities open to it. The desirable sustainable relation between public and private investments on these networks should be considered in reference to the level of general development and open technological opportunities.

The analysis also shows the diversity and multi-level nature of the governance mechanisms need in modern catching-up. The research provides a glimpse on how the institutional framework can allow space for organisational learning and the decentralised interaction between stakeholders with very different interests. While, most of the justification on science and technology policy relies heavily on econometric measures to understand knowledge spillovers measuring it in terms of the possible economic outcomes, it is necessary to recognize that the knowledge spillovers is ultimately the result of the underlying organisational and political structures. The formation of these from group level to sectoral and international networks determines the occurrence and direction of the possible knowledge flows. Along the time, compromise and adjustment between different interests inside and outside the network should allow the identification of endogenous growth opportunities.

The balance between competition and collaboration diversity and selection needs to be addressed in the context of the institutional design. The regulation of those organisations that will be involved in the knowledge network for participation is susceptible to political influence. Political strength of specific stakeholders may block the participation of other groups, resulting in reinforcing path-dependency. Different groups will have naturally disagreements about their relative past and potential contribution to the general performance of the sector.

In this direction, a project-level analysis of sectoral system raises lines of theoretical and practical lines of enquiry that remain largely open. For instance, how do individual actors contribute to the vitality of these networks? How could changes in the rules promote a better allocation of resources in the decentralised network? Would a rule-based allocation, a discretionary allocation or a combination of both improve the long term allocation of resources in the decentralised network? Which kind of interventions (or non- interventions) should be carried out in different stages of the development of the knowledge networks? Which mechanisms should be combined in order to promote wider knowledge spillovers, stronger growth and sustainability in the networks? Further detailed analysis of the behavioural aspects of these networks as well as the developments in methodological tools need to be considered in order to attempt some answers to these questions.

As other sources of funding emerge (e.g. sectoral funds, local agencies, etc), the analysis of these networks could help developing more adequate sectoral strategies in the different

25

developmental aims. Although the network is deeply influenced by the context of the tax regime as present between 1997 and 2003, there is no reason to constrain the methods used to this unique source of funding. A project level analysis of the knowledge networks could encompass other forms of funding organised by projects, as is more and more common in grants or other forms of support to innovation. Although this level of analysis needs to be considered in terms of the costs of accountability, they certainly could have a long-run pay-off as long as they guide organisational and institutional learning.

Finally, if we expect to understand the formation of the knowledge systems in developing countries, and especially the interaction between the knowledge networks in multinational and domestic networks in sectors, one must go beyond the debate between nationalistic charged statements or ideologically influenced liberalizing reforms. This integration between networks requires complex alignment in multi-level governance structures (Kim and Tunzelmann 1998; von Tunzelmann 2004). At the same time, these organisational changes, learning and adjustments between conflicting interests are made project by project (Nonaka and Takeuchi 1995; Lam 2000). Thus it is only examining these underlying, more complex set, the creation of knowledge in companies and the interactions between private and public, domestic and foreign economic agents that define the characteristics of the system that these knowledge networks can be understood. A quantitative analysis of the innovation projects and knowledge networks provides useful insights into this direction.

26

ANNEX 1 – Descriptive Statistics, Tests and complementary tables

Figure a - Vertical Integration in different innovative activities conducted by companies under the Brazilian ICT Law. 1997-2003

Vertical Integration

0% 20% 40% 60% 80% 100%

Infrastructure

Technological Services

Training in S&T

Hardware

Production

Semiconductors

Software

System

Others

Research

Internal costs External Activities

Figure c - Mean for different technological groups defined in H1

Group

3.002.001.00

Vert

ical I

nte

gra

tion

.9

.8

.7

.6

.5

.4

.3

Group

3.002.001.00

Siz

e (

perc

enta

ge o

f in

vestm

ents

)

.4

.3

.2

.1

0.0

Table a - ANOVA test for different technological groups defined in H1

ANOVA Sum of

Squares df Mean

Square F Sig.

INT Between Groups

1.557 2 .779 55.394 .000

Within Groups

.520 37 .014

Total 2.077 39 IMP Between

Groups .398 2 .199 134.455 .000

Within Groups

.055 37 .001

Total .453 39

27

Figure d - distribution of investments per technological activity and type of organisation (used in H2 for calculating PRTA)

Row Labels Total Research Hardware Software Semicond. Middleware Process Training Tech Services Quality Infrastructure

Foreign companies 42.71% 10.81% 41.16% 46.12% 75.67% 45.12% 76.27% 30.63% 17.73% 53.76% 35.54%

Domestic companies 17.95% 10.34% 39.21% 12.45% 17.40% 24.74% 13.78% 10.94% 8.78% 29.49% 9.56% Private Research Institute 24.18% 32.23% 9.86% 28.29% 3.48% 18.71% 5.28% 35.38% 57.87% 10.49% 31.77%

Public Research Institute 5.01% 15.44% 2.48% 4.49% 1.48% 3.66% 2.28% 3.27% 7.89% 1.05% 2.56%

Private Educational Institute 6.83% 14.67% 1.14% 6.97% 0.11% 6.60% 1.35% 8.53% 6.67% 4.54% 15.42%

Public Educational Institute 3.32% 16.52% 6.15% 1.69% 1.87% 1.18% 1.04% 11.25% 1.06% 0.67% 5.16%

Grand Total 100.00% 100.00% 100.00% 100.00% 100.00% 100.00% 100.00% 100.00% 100.00% 100.00% 100.00%

Table b - Normality test for ties in the knowledge network – (H3)

Tests of N o r m alit y

.096 22.200 *.973 22.782

.045 141.200 *.995 141.892

.047 174.200 *.990 174.239

.100 60 .200* .974 60 .221

.092 90 .056 .946 90 .001

.065 120 .200* .986 120 .229

.025 304 .200* .997 304 .847

.078 162 .018 .983 162 .043

.043 425 .055 .991 425 .010

.049 230 .200* .990 230 .094

.043 240 .200* .996 240 .757

Tie S t rength (ln)

Tie S t rength (ln )

Tie S t rength (ln)

Tie S t rength (ln)

Tie S t rength (ln )

Tie Strength (ln)

Tie S t re n gt h ( l n)

Tie Stren gt h (l n)

Tie S t r engt h (l n)

Tie Strength (ln)

Tie S t re n gt h ( ln )

KNOW

Semic ondu ctors

Har dw ar e

Infras tr u c ture

Othe r s

Proce s s Techno log y

Quali ty S ystem

Res ea r ch

Ser vi c es

Sof tw ar e

Syst e m

Tra in ing

Stat istic df Sig.St at isticdf Sig.

Kolmo gorov-Smirno va

Shapi r o- Wilk

This is a low er b ound of the t rue significance.*.

Lilli e fors Signif i ca n ce C or re cti ona.

References

Acha, V. and L. Cusmano (2005). "Governance and Co-ordination of Distributed Innovation processes: patterns of R&D co-operation in the upstream pretroleum industry." Econ. Innov. New. Techn. 14(1-2): 1-21.

Antonelli, C., A. Geuna, et al. (2000). "Information and communication technologies and the production, distribution and use of knowledge." International Journal of Technology Management 20(1): 72-94.

Arrow, K. J. (1962). Economic Welfare and the Allocation of Resources for Innovation. The Rate and Direction of Inventive Activity. L. K. R. R. Nelson.

Audretsch, D. B. (1995). Innovation and Industry Evolution, The MIT Press. Bell, M. and M. Albu (1999). "Knowledge systems and technological dynamism in industrial clusters

in developing countries." World Development 27(9): 1715-1734. Bell, M. and K. Pavitt (1993). "Technological accumulation and industrial growth: contrasts between

developed and developing countries." Industrial and Corporate Change 2(2): 157-210. Borgatti, S. P. (2002). NetDraw: Graph Visualization Software, Harvard: Analytic Technologies. Borgatti, S. P., M. G. Everett, et al. (2002). Ucinet 6 for Windows: Software for Social Network

Analysis, Harvard: Analytic Technologies. Brown, J. S. and P. Duguid (2001). "Knowledge and Organization: A Social-Practice Perspective."

Organization Science 12(2): 198. Brusoni, S., A. Prencipe, et al. (2001). "Knowledge Specialization, Organizational Coupling, and the

Boundaries of the Firm: Why Do Firms Know More Than They Make?" Administrative Science Quarterly 46(4): 597-621.

28

Burt, R. S. (2000). "The network structure of social capital." Research in Organizational Behavior 22(2): 345-423.

Cantwell, J. (2001). Innovation, Profits and Growth: Schumpeter and Penrose. Economics and Management Series. Reading, University of Reading.

Chiesa, V. (2000). "Global R&D Project Management and Organization: A Taxonomy." Journal of Product Innovation Management 17(5 %R doi:10.1111/1540-5885.1750341): 341-359.

Cimoli, M. and G. Dosi (1995). "Technological Paradigms, Patterns of Learning and Development: An Introductory Roadmap." Journal of Evolutionary Economics 5(3): 243-68.

Coase, R. H. (1937). The Nature of the Firm. The Nature of the Firm, 1993. O. E. Williamson and S. G. Winter. Oxford, University Press.

Cummings, J. L. and B. S. Teng (2003). "Transferring R&D knowledge: the key factors affecting knowledge transfer success." Journal of Engineering and Technology Management 20(1-2): 39-68.

Davies, A. and T. Brady (2000). "Organisational capabilities and learning in complex product systems: towards repeatable solutions." Research Policy 29(7): 931-953.

Davies, A. and M. Hobday (2005). The business of projects : managing innovation in complex products and systems. Cambridge ; New York, Cambridge University Press.

DeBresson, C. and F. Amesse (1991). "Networks of innovators: Introduction." Research Policy 20: 363-379.

DeFillippi, R. J. and M. B. Arthur (1998). "Paradox in project-based enterprise: the case of film making." California Management Review 40(2): 125-139.

Dosi, G. (1988). Technical change and economic theory. London ; New York, Pinter Publishers. Dosi, G., S. G. Winter, et al. (2000). The Nature and Dynamics of Organizational Capabilities, Oxford

University Press. Dyer, J. H. and H. Singh (1998). "The Relational View: Cooperative Strategy and Sources of

Interorganizational Competitive Advantage." The Academy of Management Review 23(4): 660-679.

Edquist, C. (1997). Systems of innovation : technologies, institutions, and organizations. London, Pinter.

Engwall, M. (2003). "No project is an island: linking projects to history and context." Research Policy 32(5): 789-808.

Fombrun, C. J. (1986). "Structural Dynamics within and between Organizations." Administrative Science Quarterly 31(3): 403-421.

Gann, D. M. and A. J. Salter (2000). "Innovation in project-based, service-enhanced firms: the construction of complex products and systems." Research Policy 29(7): 955-972.

Geuna, A., A. J. Salter, et al. (2003). "Science and Innovation: Rethinking the Rationales for Funding and Governance."

Ghoshal, S. and C. A. Bartlett (1988). "Creation, Adoption, and Diffusion of Innovations by Subsidiaries of Multinational Corporations." Journal of International Business Studies 19(3): 365-388.

Gibbons, M. (1994). The new production of knowledge : the dynamics of science and research in contemporary societies. London ; Thousand Oaks, Calif., SAGE Publications.

Giddens, A. (1979). Central problems in social theory : action, structure, and contradiction in social analysis. Berkeley, University of California Press.

Giuliani, E. and M. Bell (2005). "The micro-determinants of meso-level learning and innovation: evidence from a Chilean wine cluster." Research Policy 34 47-68.

Grabher, G. (2004). "Learning in projects, remembering in networks? Communality, sociality, and connectivity in project ecologies." European Urban and Regional Studies 11(2): 103-123.

Grabher, G. (2004). "Temporary Architectures of Learning: Knowledge Governance in Project Ecologies." Organization Studies 25(9): 1491.

Gulati, R. (1999). "Network location and learning: the influence of network resources and firm capabilities on alliance formation." Strategic Management Journal 20(5): 397-420.

Gulati, R. and M. Gargiulo (1999). "Where Do Interorganizational Networks Come from?" American Journal of Sociology(March ): 177-231.

Hansen, M. T., M. L. Mors, et al. (2005). "Knowledge sharing in organisations: multiple networks, multiple phases." The Academy of Management Journal 48(5): 776-793.

Hobday, M. (2000). "The project-based organisation: an ideal form for managing complex products and systems." Research Policy 29(7-8): 871-893.

Hobday, M. (2000). "The project-based organisation: An ideal form for managing complex products and systems?" Research Policy 29(7,8): 871.

Kim, S.-R. and N. v. Tunzelmann (1998). Aligning internal and external networks: Taiwan’s specialization in IT. Electronic Working Papers Series. SPRU. Brighton, SPRU: 41.

29

Knight, L. (2002). "Network learning: Exploring learning by interorganizational networks." Human Relations 55(4): 427.

Lam, A. (2000). "Tacit Knowledge, Organizational Learning and Societal Institutions: An Integrated Framework." Organization Studies 21(3): 487.

Lam, A. (2005). "Organizational Innovation." The Oxford Handbook of Innovation, Oxford University Press, Oxford and New York.

Lane, P. J. and M. Lubatkin (1998). "Relative absorptive capacity and interorganizational learning." Strategic Management Journal 19(5): 461-477.

Leonard-Barton, D. (1987). Implementation as Mutual Adaptation of Technology and Organization, Division of Research, Harvard Business School.

Malerba, F. (2002). "Sectoral Systems of Innovation and Production." Research Policy 31(2). Malerba, F. (2005). "Sectoral systems of innovation: a framework for linking innovation to knowledge

base, structure and dynamics of sectors." Econ. Innov. New. Techn. 14(1-2): 63-82. Mani, S. (2004). Coping with Globalisation: an analysis of innovation capabilities in Brazilian

telecommunications equipment industry. Discussion Paper Series. U. N. University. Maastricht, Intech.

Manning, S. (2005). "Managing project networks as dynamic organizational forms: Learning from the TV movie industry." International Journal of Project Management 23: 410–414.

Meyer, A. D., A. S. Tsui, et al. (1993). "Configurational Approaches to Organizational Analysis." The Academy of Management Journal 36(6): 1175-1195.

Meyer, M. (2002). "Tracing knowledge flows in innovation systems." Scientometrics 54(2): 193-212. Miles, R. E. and C. C. Snow (1992). "Causes of failure in network organizations." California

Management Review 34(4): 53-72. Mizruchi, M. S. (1994). "Social Network Analysis: Recent Achievements and Current Controversies."

Acta Sociologica 37(4): 329. Nelson, R. R., Ed. (1993). National Innovation Systems. New York, Oxford University Press. Nelson, R. R. (1994). "The Co-evolution of Technology, Industrial Structure, and Supporting

Institutions." Ind Corp Change 3(1): 47-63. Nonaka, I. and H. Takeuchi (1995). The Knowledge-Creating Company. Oxford, Oxford University

Press. Owen-Smith, J. and W. W. Powell (2005). "Knowledge Networks as Channels and Conduits: The

Effects of Spillovers in the Boston Biotechnology Community." Organisation Science. Pavitt, K. (1998). "Technologies, products and organization in the innovating firm: what Adam Smith

tells us and Joseph Schumpeter doesn't." Industrial and Corporate Change 7(3): 433-451. Pavitt, K. (2001). Can the large Penrosian Firm cope with the Dynamics of Technology? Electronic

Working Paper Series. SPRU. Brighton, SPRU. Penrose, E. T. (1995). The theory of the growth of the firm. Oxford ; New York, Oxford University

Press. Powell, W. W. (1990). Neither market nor hierarchy: Network forms of organization. Research in

Organization Behavior. Barry and M. S. a. L. L. Cummings, JAY. 12: 295-336. Powell, W. W., K. W. Koput, et al. (1996). "Interorganizational Collaboration and the Locus of

Innovation: Networks of Learning in Biotechnology." Administrative Science Quarterly 41(1).

Prencipe, A. and F. Tell (2001). "Inter-project learning: processes and outcomes of knowledge codification in project-based firms." Research Policy 30(9): 1373-1394.

Pyka, A. and G. Küppers (2002). Innovation networks : theory and practice. Cheltenham, UK ; Northhampton, MA, Edward Elgar Publishing.

Simon, H. A. (1979). "Rational Decision Making in Business Organizations." The American Economic Review 69(4): 493-513.

Stefanuto, G. N. (2004). O Programa Softex e a indústria de Software no Brasil. Instituto de Geociências. Campinas, Unicamp. Phd: 182.

Swan, J. A., H. Scarbrough, et al. (2002). "The Construction of'Communities of Practice'in the Management of Innovation." Management Learning 33(4): 477.

Szapiro, M. and J. Cassiolato (2003). Telecommunications System of Innovation in Brazil: Development and Recent Challenges. The First Globelics Conference Innovation Systems and Development Strategies for the Third Millennium, Rio de Janeiro.

Teece, D. (1994). "The dynamic capabilities of firms: an introduction." Industrial & Corporate Change 3: 537-556.

Teece, D. J. (1989). "Inter-Organizational Requirements of the Innovation Process." Managerial and Decision Economics 10: 35-42.

Tidd, J. (1997). "Complexity, Networks & Learning: Integrative themes for research of innovation management." International Journal of Innovation Management 1(1).

30

Tushman, M. L. and R. Katz (1980). "External Communication and Project Performance: An Investigation into the Role of Gatekeepers." Management Science 26(11): 1071-1085.

von Tunzelmann, N. (2004). Network alignment in the catching-up economies of Europe. The Emerging Industrial Structure of the Wider Europe. F. McGowan, S. Radosevic and N. v. Tunzelmann. London - New York, Routledge.

Walker, G., B. Kogut, et al. (1997). "Social Capital, Structural Holes and the Formation of an Industry Network." Organization Science 8(2): 109-125.

Wasserman, S. and K. Faust (1994). Social network analysis : methods and applications. Cambridge ; New York, Cambridge University Press.

Williamson, O. E. (1985). The economic institutions of capitalism : firms, markets, relational contracting. New York, London, Free Press ;Collier Macmillan.

i Although there are many other mechanisms of knowledge spillover such as labour market and informal relationships, we limit our analysis to possible spillovers derived from the collaborations in innovation projects in specific. ii Other evolutionary mechanisms to transfer tacit knowledge are labour mobility and firm spin-offs. iii (for some examples of individuals, groups, and industrial vertical and horizontal relations, see Hobday, M. (2000). "The project-based organisation: An ideal form for managing complex products and systems?" Research Policy 29(7,8): 871.,DeFillippi, R. J. and M. B. Arthur (1998). "Paradox in project-based enterprise: the case of film making." California Management Review 40(2): 125-139. ,Acha, V. and L. Cusmano (2005). "Governance and Co-ordination of Distributed Innovation processes: patterns of R&D co-operation in the upstream pretroleum industry." Econ. Innov. New. Techn. 14(1-2): 1-21., {Hardstone, 2004 #735}.Manning, S. (2005). "Managing project networks as dynamic organizational forms: Learning from the TV movie industry." International Journal of Project Management 23: 410–414. iv Although these specific linkages, sometimes identified in the literature as “university industry linkages”, have a considerable amount of literature on its own, they rarely allow the examination of the networks and governance structures formed by the aggregation of individual ties. v This is supported by the idea that the normative definition of functions inside the innovation in developing countries is usually relatively weak, as has been shown by many empirical qualitative studies Bell, M. and M. Albu (1999). "Knowledge systems and technological dynamism in industrial clusters in developing countries." World Development 27(9): 1715-1734.. vi This percentage has decreased slightly during the last three years of the analysis. See www.mct.gov.br/sepin for more details about the regulatory framework. vii Further detail about the database will be provided in the forthcoming thesis. Three datasets were accessed in the Brazilian Ministry for Science and Technology in Brasilia for three different periods under a non disclosure agreement and for academic purposes only. The dataset was cleaned and integrated into the different levels of analysis. The consolidated data about the network was based on the dataset of the innovative projects developed by companies for the period 1997-2003, declared under the Brazilian ICT policy. viii The procedure used to collect the project by the ministry allowed managers to allocate individual projects into more than one category using an estimate percentage. When more than one category was provided for individual projects, the resources and transactions with partners were included into the different network weighted according to this estimate. ix The original classification was “System (hardware + software)” caracterising projects in the interface. The term system was substituted here for middleware to avoid confusion with sectoral systems. x It is supported by anecdotal information that highlights that R&D projects outsourced and the projects under the regulation are usually considered synonyms by interviewed people. xi Index is usually used with patent and scientific publications xii QAP correlation (# of Permutations: 5000,Random seed: 24322)

xiii Or