Abstract

We propose a model that reflects two important processes in R&D activities of firms, the formation of R&D alliances and the exchange of knowledge as a result of these collaborations. In a data-driven approach, we analyze two large-scale data sets extracting unique information about 7500 R&D alliances and 5200 patent portfolios of firms. This data is used to calibrate the model parameters for network formation and knowledge exchange. We obtain probabilities for incumbent and newcomer firms to link to other incumbents or newcomers which are able to reproduce the topology of the empirical R&D network. The position of firms in a knowledge space is obtained from their patents using two different classification schemes, IPC in 8 dimensions and ISI-OST-INPI in 35 dimensions. Our dynamics of knowledge exchange assumes that collaborating firms approach each other in knowledge space at a rate $μ$ for an alliance duration $τ$. Both parameters are obtained in two different ways, by comparing knowledge distances from simulations and empirics and by analyzing the collaboration efficiency $\hat{C}_n$. This is a new measure, that takes also in account the effort of firms to maintain concurrent alliances, and is evaluated via extensive computer simulations. We find that R&D alliances have a duration of around two years and that the subsequent knowledge exchange occurs at a very low rate. Hence, a firm's position in the knowledge space is rather a determinant than a consequence of its R&D alliances. From our data-driven approach we also find model configurations that can be both realistic and optimized with respect to the collaboration efficiency $\hat{C}_n$. Effective policies, as suggested by our model, would incentivize shorter R&D alliances and higher knowledge exchange rates.

Abstract

The authors develop an agent-based model to reproduce the size distribution of R&D alliances of firms. Agents are uniformly selected to initiate an alliance and to invite collaboration partners. These decide about acceptance based on an individual threshold that is compared with the utility expected from joining the current alliance. The benefit of alliances results from the fitness of the agents involved. Fitness is obtained from an empirical distribution of agent's activities. The cost of an alliance reflects its coordination effort. Two free parameters ac and a1 scale the costs and the individual threshold. If initiators receive R rejections of invitations, the alliance formation stops and another initiator is selected. The three free parameters (ac; a1; R) are calibrated against a large scale data set of about 15,000 firms engaging in about 15,000 R&D alliances over 26 years. For the validation of the model the authors compare the empirical size distribution with the theoretical one, using confidence bands, to find a very good agreement. As an asset of our agent-based model, they provide an analytical solution that allows to reduce the simulation effort considerably. The analytical solution applies to general forms of the utility of alliances. Hence, the model can be extended to other cases of alliance formation. While no information about the initiators of an alliance is available, the results indicate that mostly firms with high fitness are able to attract newcomers and to establish larger alliances.

Abstract

We study the role of geography in R&D networks by means of a quantitative, micro-geographic approach. Using a large database that covers international R&D collaborations from 1984 to 2009, we localize each actor precisely in space through its latitude and longitude. This allows us to analyze the R&D network at all geographic scales simultaneously. Our empirical results show that despite the high importance of the city level, transnational R&D collaborations at large distances are much more frequent than expected from similar networks. This provides evidence for the ambiguity of distance in economic cooperation which is also suggested by the existing literature. In addition we test whether the hypothesis of local buzz and global pipelines applies to the observed R&D network by calculating well-defined metrics from network theory.

Abstract

We analyze large-scale data sets about collaborations from two different domains: economics, specifically 22.000 R&D alliances between 14.500 firms, and science, specifically 300.000 co-authorship relations between 95.000 scientists. Considering the different domains of the data sets, we address two questions: (a) to what extent do the collaboration networks reconstructed from the data share common structural features, and (b) can their structure be reproduced by the same agent-based model. In our data-driven modeling approach we use aggregated network data to calibrate the probabilities at which agents establish collaborations with either newcomers or established agents. The model is then validated by its ability to reproduce network features not used for calibration, including distributions of degrees, path lengths, local clustering coefficients and sizes of disconnected components. Emphasis is put on comparing domains, but also sub-domains (economic sectors, scientific specializations). Interpreting the link probabilities as strategies for link formation, we find that in R&D collaborations newcomers prefer links with established agents, while in co-authorship relations newcomers prefer links with other newcomers. Our results shed new light on the long-standing question about the role of endogenous and exogenous factors (i.e., different information available to the initiator of a collaboration) in network formation.

Abstract

This paper studies the selection of partners for R&D collaborations of firms both empirically, by analyzing a large data set of R&D alliances over 25 years, and theoretically, by utilizing an agent-based model of alliance formation. We quantify the topological position of a firm in the R&D network by means of the weighted k-core decomposition which assigns a coreness value to each firm. The evolution of these coreness values over time reconstructs the career path of individual firms, where lower coreness indicates a better integration of firms in an evolving R&D network. Using a large patent dataset, we demonstrate that coreness values strongly correlate with the number of patents of a firm. Analyzing coreness differences between firms and their partners, we identify a change in selecting partners: less integrated firms choose partners of similar coreness until they reach their best network position. After that, well integrated firms (with low coreness) choose preferably partners with high coreness, either newcomers or firms from the periphery. We use the agent-based model to test whether this change in behavior needs to be explained by means of strategic considerations, i.e. firms switching their strategy in choosing partners dependent on their network position. We find that the observed behavior can be well reproduced without such strategic considerations, this way challenging the role of strategies in explaining macro patterns of collaborations.

Abstract

Drawing on a large database of publicly announced R&D alliances, we empirically investigate the evolution of R&D networks and the process of alliance formation in several manufacturing sectors over a 24-year period (1986-2009). Our goal is to empirically evaluate the temporal and sectoral robustness of a large set of network indicators, thus providing a more complete description of R&D networks with respect to the existing literature. We find that most network properties are not only invariant across sectors, but also independent of the scale of aggregation at which they are observed, and we highlight the presence of core-periphery architectures in explaining some properties emphasized in previous empirical studies (e.g. asymmetric degree distributions and small worlds). In addition, we show that many properties of R&D networks are characterized by a rise-and-fall dynamics with a peak in the mid-nineties. We find that such dynamics is driven by mechanisms of accumulative advantage, structural homophily and multiconnectivity. In particular, the change from the "rise" to the "fall" phase is associated to a structural break in the importance of multiconnectivity.

Abstract

This paper investigates the process of knowledge exchange in inter-firm Research and Development (R&D) alliances by means of an agent-based model. Extant research has pointed out that firms select alliance partners considering both network-related and network-unrelated features (e.g., social capital versus complementary knowledge stocks). In our agent-based model, firms are located in a metric knowledge space. The interaction rules incorporate an exploration phase and a knowledge transfer phase, during which firms search for a new partner and then evaluate whether they can establish an alliance to exchange their knowledge stocks. The model parameters determining the overall system properties are the rate at which alliances form and dissolve and the agents' interaction radius. Next, we define a novel indicator of performance, based on the distance traveled by the firms in the knowledge space. Remarkably, we find that - depending on the alliance formation rate and the interaction radius - firms tend to cluster around one or more attractors in the knowledge space, whose position is an emergent property of the system. And, more importantly, we find that there exists an inverted U-shaped dependence of the network performance on both model parameters.

Abstract

We propose a model that reflects two important processes in R&D activities of firms, the formation of R&D alliances and the exchange of knowledge as a result of these collaborations. In a data-driven approach, we analyze two large-scale data sets extracting unique information about 7500 R&D alliances and 5200 patent portfolios of firms. This data is used to calibrate the model parameters for network formation and knowledge exchange. We obtain probabilities for incumbent and newcomer firms to link to other incumbents or newcomers which are able to reproduce the topology of the empirical R&D network. The position of firms in a knowledge space is obtained from their patents using two different classification schemes, IPC in 8 dimensions and ISI-OST-INPI in 35 dimensions. Our dynamics of knowledge exchange assumes that collaborating firms approach each other in knowledge space at a rate $μ$ for an alliance duration $τ$. Both parameters are obtained in two different ways, by comparing knowledge distances from simulations and empirics and by analyzing the collaboration efficiency $\hat{C}_n$. This is a new measure, that takes also in account the effort of firms to maintain concurrent alliances, and is evaluated via extensive computer simulations. We find that R&D alliances have a duration of around two years and that the subsequent knowledge exchange occurs at a very low rate. Hence, a firm's position in the knowledge space is rather a determinant than a consequence of its R&D alliances. From our data-driven approach we also find model configurations that can be both realistic and optimized with respect to the collaboration efficiency $\hat{C}_n$. Effective policies, as suggested by our model, would incentivize shorter R&D alliances and higher knowledge exchange rates.

Abstract

We develop an agent-based model to reproduce the processes of link formation and knowledge exchange in a Research and Development (R&D) inter-organizational network. In our model, agents form links based on their network features, i.e. their belonging to one of the network's circles of influence and their previous alliance history, and then exchange knowledge with their partners, thus modifying their positions in a metric knowledge space. Furthermore, we validate the model against real data using a two-step approach. Through the Thomson Reuters SDC alliance dataset, we estimate the model parameters related to the link formation, thus reproducing the topology of the resulting R&D network. Subsequently, using the NBER data on firm patents, we estimate the parameters related to the knowledge exchange process, thus evaluating the rate at which firms exchange knowledge and the duration of the R&D alliances themselves. The underlying knowledge space that we consider in our real example is defined by IPC patent classes, allowing for a precise quantification of every firm's knowledge position. Our novel data-driven approach allows us to unveil the complex interdependencies between the firms' network embeddedness and their technological positions. Through the validation of our model, we find that real R&D alliances have a duration of around two years, and that the subsequent knowledge exchange occurs at a very low rate. Most of the alliances, indeed, have no consequence on the partners' knowledge positions: this suggests that a firm's position - evaluated through its patents - is rather a determinant than a consequence of its R&D alliances. Finally, we propose an indicator of collaboration performance for the whole network. We find that the real R&D network does not maximize such an indicator. Our study shows that there exist configurations that can be both realistic and optimized with respect to the collaboration performance. Effective policies to obtain an optimized collaboration network - as suggested by our model - would incentivize shorter R&D alliances and higher knowledge exchange rates, for instance including rewards for quick co-patenting by allied firms.

Abstract

From the perspective of innovation economics evolving institutions, innovating entrepreneurs, technological change, and creative destruction are the driving force of economic growth (Schumpeter, 1942). To mitigate the uncertainty involved in the creation of new processes, products, or business models, innovation exhibits an intrinsic collaborative nature. Innovator networks form through formal and informal collaborations between different agents, including firms, institutions, universities, state agencies, inventors, and other stakeholders of the innovation system. Being embedded in a network enables these agents to coordinate innovative efforts, as well as to pool and jointly create knowledge.

Abstract

We develop an agent-based model of strategic link formation in Research and Development (R&D) networks. Empirical evidence has shown that the growth of these networks is driven by mechanisms which are both endogenous to the system (that is, depending on existing alliances patterns) and exogenous (that is, driven by an exploratory search for newcomer firms). Extant research to date has not investigated both mechanisms simultaneously in a comparative manner. To overcome this limitation, we develop a general modeling framework to shed light on the relative importance of these two mechanisms. We test our model against a comprehensive dataset, listing cross-country and cross-sectoral R&D alliances from 1984 to 2009. Our results show that by fitting only three macroscopic properties of the network topology, this framework is able to reproduce a number of micro-level measures, including the distributions of degree, local clustering, path length and component size, and the emergence of network clusters. Furthermore, by estimating the link probabilities towards newcomers and established firms from the data, we find that endogenous mechanisms are predominant over the exogenous ones in the network formation, thus quantifying the importance of existing structures in selecting partner firms.

Abstract

We investigate the efficiency and stability of R&D networks in a model with network-dependent indirect spillovers. We show that the efficient network structure critically depends on the marginal cost of R&D collaborations. When the marginal cost is low, the complete graph is efficient, while high marginal costs imply that the efficient network is asymmetric and has a nested structure. Regarding the stability of network structures, we show the existence of both symmetric and asymmetric equilibria. The efficient network is stable for small industry size and small cost. In contrast, for large industry size, there is a wide region of cost in which the efficient network is not stable. This implies a divergence between efficiency and stability in large industries.

Abstract

We introduce a new model for the evolution of networks of firms exchanging knowledge in R&D partnerships. Innovation is assumed to result from the recombination of knowledge among firms in an R&D intensive industry. The decision of two firms to establish a new partnerships or to terminate an existing one, is based on their marginal revenues and costs, which in turn depend on the position they occupy in the network. Moreover, the formation of a collaboration has significant external effects on the other firms in the same connected component of the network. We show that this decentralized partner selection process leads to the existence of multiple equilibrium structures. Finally, by means of computer simulations, we study the properties of the emerging equilibrium networks and we show that they reproduce the stylized facts of R&D networks.

Abstract

We consider a dynamic model of network formation where agents form and sever links based on the centrality of their potential partners.We show that the existence of capacity constrains in the amount of links an agent can maintain introduces a transition from dissortative to assortative networks. This effect can shed light on the distinction between technological and social networks as it gives a simple mechanism explaining how and why this transition occurs.

Abstract

We develop a new framework for modeling innovation networks which evolve over time. The nodes in the network represent firms, whereas the directed links represent unilateral interactions between the firms. Both nodes and links evolve according to their own dynamics and on different time scales. The model assumes that firms produce knowledge based on the knowledge exchange with other firms, which involves both costs and benefits for the participating firms. In order to increase their knowledge production, firms follow different strategies to create and/or to delete links with other firms. Dependent on the information firms take into account for their decision, we find the emergence of different network structures. We analyze the conditions for the existence of these structures within a mathematical approach and underpin our findings by extensive computer simulations which show the evolution of the networks and their equilibrium state. In the discussion of the results, particular attention is given to the emergence of direct and indirect reciprocity in knowledge exchange, which refers to the emergence of cycles in the network structure. In order to motivate our modeling framework, in the first part of the chapter we give a broad overview of existing literature from economics and physics. This shows that our framework bridges and extends two different lines of research, namely the study of equilibrium networks with simple topologies and the dynamic approach of hypercycle models.

Abstract

We investigate some of the properties and extensions of a dynamic innovation network model recently introduced in [Koenig et al. Games and Eco. Beh. 75-2 p694-p713 (2012)]. In the model, the set of efficient graphs ranges, depending on the cost for maintaining a link, from the complete graph to the (quasi-) star, varying within a well defined class of graphs. However, the interplay between dynamics on the nodes and topology of the network leads to equilibrium networks which are typically not efficient and are characterized, as observed in empirical studies of R&D networks, by sparseness, presence of clusters and heterogeneity of degree. In this paper, we analyze the relation between the growth rate of the knowledge stock of the agents from R&D collaborations and the properties of the adjacency matrix associated with the network of collaborations. By means of computer simulations we further investigate how the equilibrium network is affected by increasing the evaluation time $ tau$ over which agents evaluate whether to maintain a link or not. We show that only if $ tau$ is long enough, efficient networks can be obtained by the selfish link formation process of agents, otherwise the equilibrium network is inefficient. This work should assist in building a theoretical framework of R&D networks from which policies can be derived that aim at fostering efficient innovation networks.