Specifying network boundaries is fundamental in the study of social structures of elite networks. However, traditional methods do not offer clear criteria on either size or composition of the elite, and rely on numerous ad hoc decisions. A methodological framework that is inductive, reproducible and suitable for comparative research is proposed. First, a comprehensive dataset of the 5079 affiliation networks of all potentially powerful sectors in Denmark was assembled. Second, these heterogeneous affiliation networks were weighted to account for potential level of social integration. Third, a weighted modification of k-cores is used to identify a power elite of 423 individuals.