MARS: A framework for modelling register-based social networks

Abstract

Register-based social networks have become of increasing interest in countries where formal government-curated microdata is available. Due to the non-trivial generative process of register-based networks, existing random graph models fail to facilitate effective structural analysis, hindering the discovery of meaningful insights in the underlying social system. In this paper we introduce the Multiplex Affiliation-based Random Spatially-embedded (MARS) graph framework, which replicates the construction method of register-based social networks. We derive fundamental statistical properties of MARS ensembles in general and special cases. To demonstrate the applicability of the framework, we implement a simple model under the MARS framework and show that it recovers similar properties to those exhibited by the population-scale register-based social network of the Netherlands. Furthermore, we analyse the effect of spatial tie strength on closure in the network and compare our results with existing empirical findings, showing that increased spatial freedom is correlated with decreased social cohesion.