Homophily, Preferential Attachment or Triadic Closure? How the Topology of Acquaintance Networks is Determined by Processes of Link Creation

Abstract

The proposed model integrates several approaches to explain the evolution of social networks. These approaches -- preferential attachment, homophily, spatial constraints and triadic closure -- are found independently to reproduce network characteristics which can be found in empirical social networks. ERGMs and SAOMs in turn comprehensively fit all these effects to empirical network structures to precisely estimate the strength of all these processes during network evolution. Nonetheless, the latter models have problems reproducing global network topologies as found in empirical data. Here, a model is described which focalizes global network topologies to estimate the role of different network evolution processes. While being far from such precision as ERGMs and SAOMs, we came to some important conclusions: (1) While the skewness of degree distributions can be reproduced without preferential attachment, its inclusion always leads to bad fits. (2) Triadic closure alone is not capable of reproducing high clustering coefficients. (3) The main effects appear to be homophily and spatial constraints while the dimensions of these effects need to be limited. These results should be taken into account to improve overall fits of SAOMs and ERGMs.