Research and technology have been given a new boost in recent years. Basic technologies such as information and communication, material sciences and biotechnology have burst onto the economic and social scene and are now going from strength to strength.
The European Union (EU), quick to perceive this trend, has recognized the importance of joint collaboration and cooperation between European firms in activities with a high technological content (European Commission, 1995). Community policies are therefore propounding and supporting programmes of great strategic importance, both technological and commercial, in certain industrial sectors.
Since 1984, the Community has been implementing an autonomous research and technological development (R&D) policy by means of “framework programmes.” The framework programme is the legal and organizational instrument for their development, laying down the main lines of work that qualify for EU financing—see Table 1 and for further information, see http//europa.eu.int; or http//www.cordis.lu.
* Million Euros
The aim is to make R&D activities more complementary with the union’s other policies. The R&D projects involve several European countries—to foment transnational collaboration—several companies—to abide by competition laws—and, as far as possible, SMEs and universities —to stimulate technology transfer. European cooperation is thus a sine qua non condition for obtaining financing from framework programmes. Projects submitted have to meet the following terms:
That each team includes centers and/or companies from at least two community countries;
That participants include at least one industry and a university or research center, and;
That working teams are as interdisciplinary as possible.
In short, the European Union’s technology policy encourages transnational network initiatives; hence the importance and interest of analyzing them, as they make up the basic structure for technological development and in fact a key part of business competitiveness.
Our goal is to analyze the management in R&D networks. As a departure point, we consider that R&D networks are a complex phenomenon, a complexity that derives from the heterogeneity of agents who take part and multiple interactions, from the technological process (non-linear and non-sequential) and from the organizational form in which R&D projects are supported (the network of partners who move between the conflict and the cooperation). In this paper, we seek to address this complexity through general systems theory, which will allow us to model R&D networks by means of three subsystems—technological, structural, and governance—and to identify the variables that characterize each subsystem.
This paper is organized as follows. In section 2, we give a brief overview on complexity of R&D networks. Section 3 presents the model and characterizes variables to analyze the network as a complex system. In section 4, we describe the empirical study, the measurement of variables, and the main results. In section 5, we discuss the results and the interactions among the variables of each subsystem. The paper ends with some concluding remarks.
The sources of complexity in R&D networks
R&D networks are usually defined as the union of two or more parties, institutions or individuals, who pursue a distinct assignment together (Balachandra & Friar, 1997; Aronson et al., 2001). The development of R&D projects implies the execution of activities that will create interactions in the dynamic process for the accomplishment of objectives. Laredo and Mustar (1996) also indicate that an organizational form is necessary for the development of the project; Teece (1992) and Ring and Van de Ven (1992) conclude that R&D networks are a form of business organization.
The first stumbling block in studying technological networks is the lack of a single theoretical reference on which to base their study, since these types of agreements and structures take in a host of different aspects and hence different approaches (for a review see Auster, 1994). Networks and alliances constitute a new “ubiquitous phenomenon” (Gulati, 1998) expressing a wide-ranging field that incorporates many different denominations: agreements, coalitions, consortia and networks, strategic alliances or associations.
From the above it can be deduced that the phenomenon of collaboration between economic agents, and more specifically in technological areas, is characterized by the ambiguity of the terminology, the multiple analytic approaches, the diversity of objectives, the multiple organizational forms, and so forth. These aspects show that complexity is a feature of technological networks. In order to define principles for dealing with the above, it will be fundamental to take into account this aspect, and we will therefore proceed to describe the main factors that, in our view, make technological networks a complex phenomenon.
Interactions and structural heterogeneity in technological networks
The aim of technological networks is technological development, this being considered as a process ranging from generation to diffusion in the market. Since the 1980s, with the explosion of international collaboration in this area, technological processes have ceased to be considered as sequential and linear, among other factors because technological processes are developed in networks where multiple interactions and a great diversity of participating agents are introduced, in what has been described as an interactive process, non-sequential and non-linear (Rothwell, 1994). The network is not only made up of companies, but also involves customers, suppliers, universities, public research centers, and so forth. This circumstance derives from a complex interaction process between the different participants and a great structural heterogeneity through both the diversity of organization types and the different levels covered by the network (individuals, companies or company groupings, national innovation system, and so forth) as well through the environment in which they materialise—local, national, or supranational.
We should point out that the actual term “interaction” or relationship between organizations is in itself diffuse, a plurality of interpretations existing derived from diverse study and analytic approaches (Balakrishnan & Koza 1993; Chiesa & Manzini,1998; Teece 1998).
Generally, however, we can confine the content of the relationship between organizations to four elements (Johanson & Mattson, 1987):
A common orientation or predisposition to act jointly, whether it be exploiting and sharing a good (generating economies of scale) or making use of complementary aspects in the participating agents.
Dependency, deriving from different organizations acting together.
Implementation of the connecting links, which in some ways is a form of uniting the interacting parts. These may present a series of characteristics, which Aldrich (1979) limits to four: formalization, intensity, reciprocity, and standardization.
Investment made by the parties involved, which will determine the future commitment to the relationship and which normally materializes in people and time.
|Sources of complexity of R&D networks|
|1. Partners’ heterogeneity||Partners (number and type)|
|2. Multiple interactions||Resources (number and type)|
|3. Conflicts of interest||Concordance of objectives|
All these elements mean that relations between the agents are very complex when taking into account the great importance of these interactions in technological processes.
The conflict of interests in technological networks
A wide variety of institutions and organizations participate in the networks with multiple objectives and preferences, with different information between themselves, with different capacities and decision-making criteria, and all of this in uncertain environments. Situations of conflict sometimes arise from structural network aspects (Thomas, 1976), such as incompatible objectives, or disagreement on decisions and contribution of resources (clear conflict); in others, they are related to attitudes and feelings among network members and are associated with disagreements about functions to be carried out and over expectations, perceptions, and communication (underlying conflicts).
We could therefore consider relations between agents in technological networks to be a game of dynamic equilibrium, between cooperation and conflict, which increases the level of management complexity due to the difficulty of predicting the behavior of participating agents. The main characteristics defining complexity as an economic phenomenon are structural heterogeneity and functional interdependence (Koppel et al., 1991). Table 2 summarizes these issues.
Approaches to the study of technological networks
For the purposes of this paper we define networks as the set of cooperation agreements reached between different independent organizations (companies, the government, universities, or other types of institutions) to carry out a communal technological project.
To understand the management of R&D networks that have arisen in the framework of community R&D programmes, the analysis, in our opinion, has to take in the following questions:
Structural aspects: in terms of the kind of members involved in the projects, their number, characteristics, and so on.
Technological aspects: in terms of the type of products generated, the activities carried out, the stages into which the process is divided, interrelations between partners, and so on.
Organizational aspects: in terms of governance network, hierarchical structure, the definition of objectives, and so on.
Strategic aspects: In terms of the possible competitive advantages deriving from participation in technological projects, with the type of resources sought in the cooperation, and so on.
We can therefore pinpoint four theoretical approaches for tackling the host of aspects involved in the study of technological networks and, in general, in the analysis of cooperation between organizations; see Table 3. In sum, technological networks have a markedly multidisciplinary character and embrace different approaches, dimensions that all need to be taken on board in our study, otherwise our account could end up being partial and biased.
R&D networks from a systemic approach
Along these lines, therefore, we can analyze R&D networks from a systemic perspective whereby networks are conceived as complex organizational forms responding to a behavioral scheme, endowed with control mechanisms and adapted to the environment (Bertalanffy, 1956). This approach accepts the construction of a logic to analyze and explain the problems of network management. In this way, we can move closer to the complex reality of the technological collaboration phenomenon and break down the system complexity into simpler subsystems, in which we can tackle analysis from study approaches already theorized.
The accomplishment of a technology project in an R&D network supposes on the one hand the development of a technological process (R&D process) and, on the other hand the existence of an organizational structure (R&D network) to develop it. In the development of an R&D process a set of partners (firms, universities, research centers, and so on) are involved that, through a series of stages (identification of needs, technological description, and so on) carry out technological activities (basic or applied research, prototype, etc.) to achieve diverse objectives (patents, new products, training, etc.).
According to the objectives of R&D projects, we will suppose that partners, activities, and stages will be different. Furthermore, to develop the R&D project it will be necessary to establish an organizational structure the objective of which is to manage the activities and relations among the partners in order to fulfill the goal of R&D networks. Depending on the technological objectives of the project, the networks will be more or less structured and more open or closed to external contacts. From the structural point of view, the networks can adopt forms nearer to the company structure, or forms with a lesser organizational structure, nearer to the market. From the viewpoint of contacts, the networks will be more open or closed according to the typology and frequency of the external contacts.
We will therefore define R&D networks as a complex socio-technical system in interaction with their environment, thus allowing us to define a series of interdependent subsystems related to their governance, their technological, and finally their structural outlook. In the framework that follows, we will discuss these subsystems.
The recognition of organizational structures in R&D alliances is an initial condition frequently mentioned in collaborative R&D literature. Ring and Van de Ven (1994) and Teece (1992) affirm that R&D networks are a form of business organization, although a great disparity of views exists on how to approach the different governance structures of R&D networks (Branstetter & Sakakibara, 2002). Gulati (1998) defines the governance structure as the formal contractual structure used to organize the partnership in strategic alliances. Williamson (2002), for his part, points out that the objective of a governance structure is to infuse order into a relation where potential conflict can arise, and where opportunities to make common gains exist. This author illustrates that three types of attributes to describe a mode of governance exist: 1) incentive intensity; 2) administrative controls; and 3) the regulatory regime.
From a transaction cost approach, the main explanatory variables for governance structures are specificity and appropriability (Williamson, 2002). Then, the objectives of governance structures are minimizing transaction costs and opportunist behavior, turning them into hierarchy and appropriation costs. From this point of view the greater the specificity of an alliance, the more structured the form of governance; and the greater the appropriation of technology, the greater the safeguard mechanisms (see Figure 1).
The question that emerges is how to determine the most efficient form of governance for R&D networks, as much in terms of the structure most adapted to manage the network as in relation to the control or safeguard mechanisms to mitigate conflictive situations. Thus we will consider that a governance form is efficient when it optimizes its functioning (solving conflicts, coordinating tasks, and distributing results).
As mentioned above, in order to achieve the technological objectives of the network, it is necessary to develop a number of activities, which can be grouped in stages that shape the technological process (Kline & Rosenberg, 1986). The research on technological process has explored, in the development of different stages, non-sequential and non-linear processes as well as multiple links as attributes of them (Rothwell, 1994). We will consider that the accomplishment of efficiency in an R&D network will consists in their optimization; that is, the maximization of the relation between the R&D project results and the inputs used in performance of the process. In the literature of project management, it is usual to consider that inputs are the activities developed and the time devoted by partners in their execution, although in the case of the time variable, many authors consider the stages by which the project is developed the variable that fits this concept better.
The practical questions that emerge are: What activities are developed in an R&D project? Which activities are more significant in this project? What stages are performed in an R&D project? Which are the most significant?
Networks are made up of a series of organizations (universities, public research centers, companies, the government, consulting firms, and so forth) that constitute their nodes, bonded by a series of links of interrelationship and interdependence. The development of these aspects sometimes implies important difficulties for the network. As has been indicated above, R&D networks are organizational forms that move between collaboration and conflict. This situation is derived from the heterogeneity of agents taking part in the network. The difficulty from a structural point of view arises when attempting to combine individual objectives of partners with the objectives of the R&D network. In this sense, Marschack and Radner (1974) established different types of economic organizations. Taking the distinguishing criteria to be the relationship that may exist between the objectives of the organization and those of its members, they note different situations: when an objective exists for the organization that coincides with the individual objectives of the participants (the team); and when participants have individual objectives that do not concur, but an incentive to cooperate exists (the coalition). Between these two cases, diverse combinations of degrees of concordance will exist. Relations between agents in networks can be described as a dynamic game in search of equilibrium, between cooperation and conflict; that is, between teams and coalitions. In cooperation, an incentive to cooperate exists following the purpose for which the network was organized, so joint action will yield more results than individual action; in conflict, non-concordance between the individual objectives (individual utility) and common objectives (joint utility) can occur.
Therefore, in order to reach network objectives, partners contribute different resources to maximize the joint utility (network efficiency). Several characteristics that stem from the network structure must be studied. The first characteristic is the heterogeneity of the agents involved in the network, because a variable and diverse number of agents form part of it. This question makes us wonder about the number and type of partners that need to participate in the network for it to be efficient. The second characteristic we have mentioned is the multiplicity of interactions, the purpose of which is the exchange of resources among partners. Interaction may be technological, financial, informative, and so on. This makes us wonder which class and quantity of resources it is necessary to contribute so that network can be efficient.
On the other hand, each partner in the network tries to maximize its individual utility, contributing different resources. This question makes us wonder if different efficiency points exist for networks and for individual partners. Table 4 shows an overall scheme.
The data were collected in 1999 in joint projects developed under the III—IV R&D Framework Programme, between 1990 and 1998. The need for a representative sample (with experience and high participation in European technological projects) urged us to choose a set of institutions at a European level, the ILOs, institutions under the aegis of the Community European Technology Program (III and IV Framework Programme). These constitute a very dynamic set of specialized institutions with great professional experience in running European R&D projects. The sample had a population of 202 institutions, which was selected for a mail survey.
We pre-tested the survey instrument with a small group of ILOs from different countries before sending out the final version. The final questionnaire was then sent to the whole set. A total of 189 valid surveys were returned, a response rate of 93.5 percent.
Since few empirical precedents existed to develop these measurements, we relied on extensive literature and fieldwork to select individual items for our scales. Table 5, 6, and 7 provide a synthesis from a review of extensive joint research projects that refer to representative studies analyzing these issues.
In relation to the governance subsystem, the selected variables refer to the mechanisms that govern the R&D network and are explained from a transaction costs perspective. The first variable is the degree of hierarchy, which can vary depending upon the similarity of the structural elements of the network with the market or the firm. The hierarchical mechanisms used are derived from the need to plan, decide, and organize the activities to be developed. In relation to the planning of the network, diverse criteria are noted in the literature (Brockhoff, 1992; Sakakibara, 1997; Trott, 1998). The first is linked to the equilibrium among partners and in European transnational projects includes the country factor, which seeks a certain equilibrium in the distribution of tasks. The second criterion considers the scientific and technological specialization of partners. The last one refers to the special requirements of the project, mainly in sponsored projects. Regarding decision making, the specific literature shows that two centers of decision making exist: the coordinator of the network and the consensus between the partners. Concerning the organization of activities between partners, two situations are seen as most common in the development of the project: the teams within the network and the independent development of activities.
The second group of variables is related to the degree of control or safeguard mechanisms that govern the R&D network in order to avoid opportunistic behavior (Tidd & Trewhella, 1997; Williamson, 2002; Bierly & Coombs, 2004). The specific literature on networks asserts that the selection of a partner, based on previous experiences and confidence, serves as an important factor to minimize opportunistic behavior. Furthermore, the definition of responsibilities (both in the inputs and the sharing of benefits as well as in the definition of tasks) and the monitoring mechanisms (reports and meetings among partners, informal communication, and the role of coordinators) are frequently used as safeguard mechanisms (Ring & Van de Ven, 1992, 1994; Mohr & Sperkman, 1994).
Note: Respondents used a 5-point Likert scale to provide responses on each item, such that “1 is strongly disagree or low frequency and 5 is strongly agree or high frequency”.
The next group of variables tries to analyze the technological subsystem. The variables that measure this subsystem are the type of activity (Activities), the objectives of the project (Products), and the stages developed in the project (Stages). The combination of these three groups of variables will determine the type of R&D project developed. The degree of applicability of the R&D project will be determined, as Trott (1998) indicates, by its proximity to basic research projects (less applied) or to more applied R&D projects such as technological development, prototype, or detail engineering (Kline & Rosenberg, 1986; Dosi, 1988; Brockhoff, 1992; Rothwell, 1994; Sakakibara, 1997; Savioz & Sannemann, 1999; Husain & Pathak, 2002; O’Sullivan, 2003).
The last group of variables tries to measure the structural subsystem. The degree of concordance will be analyzed through different items. R&D networks are made up of a set of organizations (Partners and Size) that constitute their nodes, bonded by a series of links of interrelationship and interdependence with an incentive to participate (Incentives). In addition to these variables, Bierly and Coombs (2004) point out that the ranking of individual objectives of partners and the common objectives of the network (Ranking) is a variable that indicates the degree of concordance between partners. As Hagedoorn (1993) and Auster (1994) indicate, the greater the diversity of objectives, the size of the network, and the heterogeneity of partners are factors that increase the probability of discrepancies arising inside the network, and therefore the degree of concordance in objectives will be lower (Dyer & Nobeoka, 2000; Rowley et al., 2000; Hagedoorn et al., 2000; O’Sullivan, 2003; Johansen et al. 2005).
The following subsections show the results from the study. Table 8 presents a set of descriptive data from European R&D networks such as scientific areas, geographic distribution, and typology of networks. Afterwards we analyze the main characteristics of the three subsystems marked governance, technological, and structural.
We proceeded to treat the variables, homogenizing and simplifying them, with the aim of obtaining a set of factors that represent the different subsystems of the R&D network. Because multiple-item scales have been used to measure factors and a composite score based on these items is used in further analyses, it is important to assess the validity and reliability of the scales proposed (Bagozzi & Yi, 1988). Selection of scale items based on prior literature, fieldwork, and pre-testing of the survey instrument helped ensure content or face validity. To assess reliability we computed Cronbach alphas for each multiple-scale item and found this to be well above the cut-off value of 0.7 in each case (Hair et al., 1998).
Table 9 shows the reliability of the measures, descriptive statistics, and the group of different variables that we obtained through the factor analysis.
Table 10 provides the correlation matrix of the factors (Hair et al., 1998). We get satisfactory results for validity and reliability from the factors and, thus, we can accept their validity.
On the other hand, as is mentioned above, the research on technological process has explored, in the development of different stages, non-sequential and non-linear processes as well as the multiple links as attributes of them (Rothwell, 1994). The second question raised in our study is whether some interrelation exists between the different technological R&D projects identified previously in the exploratory analysis. We proceeded to determine if interactions between the different kinds of R&D projects exist, using a multiple regression analysis (see Table 11).
For a better understanding of the interaction between projects, we have created a dynamic simulation through the application of a neural network. Neural networks are analytic techniques modeled after the (hypothesized) processes of learning in the cognitive system and the neurological functions of the brain. They are capable of predicting new observations (on specific variables) from other observations (on the same or other variables) after executing a process of so-called learning from existing data. Artificial neural networks (ANNs) have a long tradition of application in the fields of marketing and management due to their usefulness in solving problems where non-linearity exists. ANNs are considered non-linear regression models or extensions of multinomial logit models.
We have utilized Multilayer Perception (MLP—Neural Connection 2.1), which is the most common neural computing technique. In the training phase, we obtained the best adjustment of neural networks, determining the number of hidden nodes as well as the transfer function for the three possible combinations of relations among the types of R&D projects (invention, innovation, and diffusion). We opted for the dynamic representation, which allows a better viewing of the results (see Figures 2, 3, and 4).
Correlation is significant at level ***p<0.10 **p<0.05 *p<0.01
The framework of analysis developed allows us to characterize and explain R&D networks from a systemic point of view.
Regarding the results, they show the existence of a technological subsystem, the factor loadings of which reveal three types of R&D projects: innovation, diffusion, and invention projects, with diverse degrees of applicability, as a function of the project’s objective. As is shown in Figures 2, 3, and 4, linearity between these projects does not exist. Non-linearity is characterized because increases in the input variable do not correspond with equal increases in the output variable. Moreover, as can be seen in the three graphs, there is a higher ceiling to functions, which implies that greater increases in the input variable do not generate more output or results, which allows us to discard progressiveness.
Therefore, to the weak interrelation between projects we can add non-linearity and non-progressiveness as characteristics of interrelation. This allows us to consider the independence between projects and non-sequentiality linking invention—innovation—diffusion.
In the structural subsystem the characteristic variables analyzed were the degree of concordance. In this regard, the results show a diversity of incentives between partners, which makes the probability of agreement among them low, particularly if we consider that the number of partners is also high (8 to 10 partners, with loading 0.543; >10 partners, with loading 0.529). This circumstance also makes it clear when comparing the results of the variables’ individual objectives (whose item loading is 0.669), and the concordance of objectives (item loading is 0.325). Therefore we can affirm, from the structural subsystem point of view, that there is a low coincidence of objectives, despite the existence of a common incentive. The network, in Marshak’s terminology, can be defined as a coalition of institutions.
In the case of the governance subsystem, the characteristic variables are the degree of hierarchy and the degree of control in the development of joint R&D projects. Concerning the degree of hierarchy, the analysis shows that the existence of a coordinator in the network as well as the opinion of partners are perceived as similar variables (Opinions 1 and 2, the factor loadings of which are 0.811 and 0.823 respectively) when problems arise. This result confirms one of the characteristics of networks, which is the scarce (low) hierarchy in their governance. Another characteristic of networks is the existence of different criteria regarding planning, the principal one being the equal distribution between partners and countries (the factor loading of which is 0.837), and with low loadings, the requirements of the program (0.756), the scientific and technological knowledge of partners
Regarding the existence of safeguard mechanisms—or the degree of control in the network—results show that a series of ex-ante and ex-post mechanisms exist. Among the former, a very important factor is the existence of previous experiences that, along with the technological requirements of the program, constitute the ex-ante control mechanisms for partner selection. Another ex-ante mechanism is the definition of responsibilities, both in contributing and sharing results between partners, and in the boundary between tasks. Ex-post mechanisms are based fundamentally on meetings between partners, making clear again the importance of consensus in the management of a project. We can conclude that the degree of control is higher and is based on partner selection (as result of previous experiences, the average of which is 4.0 and the loading factor 0.794) and regular meetings (the average is 4.1 and the loading factor 0.765).
The purpose of our study has been to approach the analysis of three subsystems for the management of R&D networks.
The governance subsystem takes in the analysis of the organizational structure and of the decision-making systems, in addition to the control and information systems that flesh out the network. Related to this subsystem, the technological networks analyzed are founded on a simple consensus-based structure with a certain ad hoc character, similar to the adhocratic structures proposed by Mintzberg (1979). The results of the study show that the organizational structure created for running the network is very simple. The network promoter, whose capacity for decision taking and supervision is very limited, and is linked to the consensus between partners, generally takes on the coordination. There is not a marked hierarchical structure, where lateral links serve as the main mechanism for coordination between partners. Summarizing the above, we can say that network management is based on a priori planning determined by the programme requirements. The searching for consensus is a constant feature, above all in important decisions, which are taken in meetings attended by all partners.
Regarding the technological subsystem, results show that a low interrelation between the projects exists. Furthermore, to the weak interrelation between projects we can add non-linearity and non-progressiveness as characteristics of interrelation. This allows us to consider the independence between projects and the non-sequentially linking invention—innovation—diffusion.
About the structural subsystem, taking into account both the kind of organizations that participate and the relations and interactions set up between the agents involved in the technological process, we can affirm that there is a low coincidence of objectives, despite the existence of a common incentive. The network therefore could be defined as a coalition of institutions. This situation might lead to conflicts of interests between partners, both in the contribution (resources) and in the sharing out of results. The networks analyzed in the sample do not correspond strictly speaking to cooperation networks, but rather to coalitions of institutions.
Regarding the methodological sphere, the most important factor here is the consideration of R&D networks as a multidisciplinary and complex phenomenon. In our opinion, therefore, it should be tackled from a general viewpoint, from which to generate a methodology for its study and management. We believe that this methodology could serve as the starting point for the analysis of efficiency in the development of technological processes in networks. This analysis would be an extremely useful tool for organizations, weighing up the pros and cons of carrying out cooperative projects with other companies.