MARS framework models register-based social networks with spatial accuracy

Register-based social networks are built from government-curated microdata, such as tax records, addresses, and family ties, and they offer a powerful lens on entire populations. But existing random graph models fail to capture the complex generative process that produces these networks, making structural analysis difficult. To fill that gap, researchers Katherine Hamilton, Irina Epure, and Frank Takes propose MARS—the Multiplex Affiliation-based Random Spatially-embedded graph framework. MARS mimics how register-based networks are actually constructed, layering affiliations and spatial embedding into a single multiplex model. The team derives fundamental statistical properties for both general and special cases, then implements a simple model under the framework and shows that it closely matches the properties observed in the Netherlands' population-scale register-based social network.

In addition to validating MARS, the researchers analyzed how spatial tie strength affects network closure, a measure of how tightly connected triads are. Their finding—that increased spatial freedom correlates with decreased social cohesion—aligns with existing empirical studies and offers a concrete hypothesis for why geographically dispersed populations show weaker local ties. With 22 pages, 9 figures, and a DOI now registered via arXiv, this work provides a reusable framework for any country with register-based microdata. For social scientists and policymakers, MARS unlocks the ability to simulate and interrogate these networks at scale, revealing structural patterns that were previously inaccessible.