Does country-level R&D efficiency benefit from the collaboration network structure?

In this study, we attempt to fill an important gap that literature has yet to investigate, that is, the influence of collaboration network structure on national research and development (R&D) efficiency. We not only provide country-level evidence that the collaboration network structure influences the R&D result performance measured by output quantity. We also prove that the collaboration network structure influences the R&D process performance measured by input–output efficiency score. The latter exploration presents the underlying explanations for the former conclusion. We construct a unique dataset that enables us to build seven scientific collaboration networks at the country level. Based on the collections of R&D data for each country in our networks, we have measured R&D efficiency scores by using the Malmquist productivity index associated with data envelopment analysis. The clustering coefficient (CC), structural holes (SH), degree centrality (DC), closeness centrality (CNC), and betweenness centrality (BC) of each country are jointly used to comprehensively measure the structural properties of collaboration networks. Panel data models are employed to explore the effect of the network properties on R&D efficiency. Our results not only reconfirm that collaboration network structure influences scientific publications at the country level, but also show that the higher SH, DC, CNC, and BC correlate positively with the better future efficiency.