Quantifying knowledge exchange in R&D networks: a data-driven model

Abstract

We develop an agent-based model to reproduce the process of link formation and to understand the effect of knowledge exchange in collaborative inter-firm networks of Research and Development (R&D) alliances. In our model, agents form links based on their previous alliance history and then exchange knowledge with their partners, thus approaching in a knowledge space.We validate our model against real data using a two-step approach. Through an inter-firm alliance dataset, we estimate the model parameters related to the alliance formation, at the same time reproducing the topology of the resulting collaboration network. Subsequently, using a dataset on firm patents, we estimate the parameters related to the process of knowledge exchange.The underlying knowledge space that we consider in our study is defined by real patent classes, allowing for a precise quantification of every firm's knowledge position. We find that real R&D alliances have a duration of around two years, and that the subsequent knowledge exchange occurs at an extremely low rate - a firm's position is rather a determinant than a consequence of its R&D alliances. Finally, we propose an indicator of collaboration performance for the whole network and, remarkably, we find that the empirical R&D network extracted from our data does not maximize such an indicator. However, we find that there exist configurations that can be both realistic and optimized with respect to the collaboration performance. Effective policies, as suggested by our model, would incentivize shorter R&D alliances and higher knowledge exchange rates.