Dichotomizing Network Data Can Change The Meaning of Actor Centrality

This paper explores the consequences of calculating actor centrality-how connected actors are to the other members of a larger group-with dichotomized network data. I argue that the meaning of actor centrality may change when researchers convert valued network data into a binary network. Data from eight archival data sets suggests that centrality describes overall connectedness when the network is dichotomized at a low cut point-e.g., when even acquaintances count as ties-but transforms into a measure of cliquishness/relationship exclusivity when the network is dichotomized at a high cut point-e.g., when friends, but not acquaintances, count as ties. Researchers who dichotomize network data may unknowingly use centrality estimates that do not map onto their theoretical construct of interest.