Research on the formation mechanism of research leadership relations: An exponential random graph model analysis approach

Abstract

The research leadership relations capture the directed and critical relations from leading authors to participating authors in research collaborations. Studying the formation mechanisms of research leadership relations helps us better understand research collaborations. When studying the formation mechanisms of research collaboration networks (RCN), existing literature primarily focuses on collaboration relations among all coauthors, ignoring the aspect of research leadership, and concentrates on cognitive proximity, ignoring other key proximities. To study the formation mechanism of research leadership relations in a comprehensive manner, we construct a research leadership network (RLN), composed of research leadership relations. We apply the Exponential Random Graph Model to model RLN, taking into account the influence of network structure and researchers’ attributes, and make a comparison between RLN and RCN. Our dataset consists of publications in Pharmaceutical Sciences, Computer Sciences, and Library & Information Sciences from 2011 to 2019. The results indicate that research leadership relations tend to be reciprocal and based on a local hierarchy. The out-degree has a significant preferential attachment effect. The homophily effects of cognitive, and institutional proximity play a significant role in shaping tie formation. Regarding the comparison between RLN and RCN, both the triadic closure and the preferential attachment play important roles but drive network formation differently. These results generally remain robust across the three research fields and provide new insights into the understanding of research collaborations.