Proximity and self-organizing mechanisms underlying scientific collaboration of cities in the Yangtze River Delta

With further research on intercity knowledge networks, the underlying influencing factors and mechanisms have become important issues in urban geography and regional studies. This study constructed an intercity scientific collaboration network of the Yangtze River Delta based on the co-publication data derived from the Web of Science during 2019-2020. After an exploratory analysis of spatial patterns and topological characteristics of the intercity scientific collaboration network, valued exponential random graph models were designed to quantitatively explore the effects of variables at the city, intercity-relation, and network-structure levels on the formation of the network, and then unravel the underlying self-organizing and proximity mechanisms. The results show that: (1) The intercity scientific collaboration network of the study area results from the joint effects of endogenous forces and exogenous forces. Exogenous forces include conventional urban knowledge endowments and multi-dimensional proximities between cities, while endogenous forces are self-organizing and self-evolving forces from local structures of the network per se which is relatively under-reported. In terms of urban endowment variables, cities with more universities, more R&D investment, and larger GDP per capita are more likely to develop scientific collaboration with other cities, among which the number of universities plays the most important role. (2) In terms of intercity relational variables, organizational proximity contributes most to the formation of the intercity scientific collaboration network. The probability of scientific collaboration between cities in the same province is 3.157 times the collaboration between cities in different provinces. For every 0.1 unit increase of cognitive proximity between cities, the probability of scientific collaboration between them would be 1.981 times the previous probability. Geographical proximity and social proximity contribute little to facilitating the intercity scientific collaboration. In contrast, the impacts of institutional proximity and cultural proximity are negative due to the stronger effects of preferential attachment and weaker barriers of regional dialects. (3) In terms of network structural variables, the intercity scientific collaboration network presents significant self-organizing and self-evolving properties. The contribution of local structures, i.e., star configuration and triangle configuration, to the formation of new intercity scientific collaboration is respectively 0.875 and 0.540, suggesting that the preferential attachment effect is stronger than the triadic closure effect.