Common large-scale approaches to inferring social structure make use of digital traces such as online social networks or mobile communication data. However, these networks are often agnostic of node and edge representativity and type. This talk investigates the structure of a social network sourced from administrative registers for an entire population based on family, household, work, school, and next-door neighbor relations, alongside rich demographic node attributes. We revisit three of the most common concepts in social network analysis: degree, closure and distance. We find that observed degrees are the result of a combination of degree distributions in various layers, disqualifying common explanatory mechanisms such as preferential attachment. Low node-to-node distances are realized through particular edge types that shortcut paths in already clustered areas. Measuring closure across layers shows how we can realistically capture the extent to which people have closed or open network opportunity structures. Finally, we highlight how people's network structure varies greatly along demographic axes such as age, income and level of education. This shows that understanding of both the type of edge and the part of the population that is considered is of great importance. Therefore, leveraging register data to capture the social structure of a complete population is one of the most fruitful ways forward to obtain actionable insights and ultimately evidence-based policies.