Origins of Structure in Social Networks

For more than a century, social scientists have documented recurring patterns in human relations: people cluster in tightly knit groups, surprisingly short acquaintance-chains link across society, and ties form disproportionately among similar individuals. These findings are foundational, yet explanations have often remained either descriptive or confined to specific contexts.
The rise of large-scale digital data has transformed the study of social structure with online traces and complete national statistical registries offering unprecedented opportunities to connect individual behavior to collective outcomes. Yet abundant data is not always available. Some fields remain scarce in data: these require model-centric reasoning as opposed to direct measurement-driven inference.
In this thesis, I combine both regimes. Using population-complete family registries from Denmark, I show that the large-scale structure of family networks is governed primarily by partner-change dynamics, which act as shortcuts linking otherwise slowly expanding kin groups, while partner-choice homophily plays a secondary role. In the domain of opinion dynamics, where observations of interpersonal influence are rare, I analyze a new weighted-median model of opinion shifts. I demonstrate how network topology controls whether
consensus emerges or disagreements persist, and develop a meanfield approximation that enables short-run forecasting in large networks. Finally, using 4.3 million marriages linked to registry data, I move beyond one-dimensional measures of homophily to study partner choice as a high-dimensional phenomenon. I show that intersections of age, education, occupation, industry, and income jointly segregate the partner market, narrowing local opportunities and sharpening global boundaries in ways invisible to single-trait analyses.
Together, these studies advance a mechanistic understanding of how tie formation, dissolution, and influence generating recurring patterns of social structure. They highlight how explanatory strategies must adapt to data abundance or scarcity, and how accounting for the semantics of ties is essential in explaining the patterned social fabric we all inhabit.