The Origins of Large-Scale Structure in Family Networks

Abstract: Family relations form the backbone of human social structure, yet the behavioral origins of large-scale family networks remain poorly understood. Leveraging a complete Danish longitudinal registry—including ~6 million individuals and ~8 million parent–child relations—we analyze how individual partner behaviors aggregate to shape network structure. While partner-choice homophily (e.g., similarity by education, geography, age) has been hypothesized to drive network formation, our findings show it plays a minor role. Instead, partner-change behavior—individuals leaving one partner for another—acts like ‘shortcuts’ in network theory, significantly reducing network path lengths and fostering meso-scale connectivity. Moreover, partner-change is self-exciting: individuals with more prior partners are more likely to change again, producing highly connected hubs. Incorporating these mechanisms into generative models captures key empirical network metrics accurately, whereas homophily-based models fail. This work provides the first quantification of the behavioral micro-foundations of family network structure.