Mesoscale Structures in World City Networks

Abstract

This study conceptualizes variant mesoscale structures (e.g., core–periphery, community, flat-world, or hybrid) in world city networks (WCNs) that enable us to understand the grouping features of cities with similar roles and positions, which are defined by distinctive relational patterns between cities as well as groups. A new analytical framework relying on a Bayesian-inference weighted stochastic block model (WSBM) is proposed to infer and compare latent mesoscale structures. This framework is superior to existing city-clustering approaches, which either ex ante postulate a mesoscale structure before clustering (e.g., the community detection method) or ex post explain the roles and positions of cities after clustering based on their similarities in attributes and thus avoid the traps of methodological determinism and territorialism. For substantiating mesoscale structures, we study a WCN of 126 cities, between which the relational strength is measured by the pairwise coreference frequency of cities on massive Internet Web pages collected by a webometrics approach. Modeling results show that the WCN reflected on Web pages has a distinctive multicore–periphery structure mixed with communities. We also develop a likelihood-based estimation to compare alternative mesoscale structures and find that the WCN configuration differs from many geographical imaginaries of globalization, such as a homogenous world, a ranking hierarchy, or a territorialist regional geography. This study asks for more comparative analyses of mesoscale structures in the future research agenda of macroregional or WCN studies.