A0952
Title: Inferring social connectedness with endogenously selected groups
Authors: Eric Auerbach - Northwestern University (United States) [presenting]
Abstract: Researchers often study social connectedness by first clustering individuals into groups based on the observed network ties. However, standard inference procedures typically ignore the fact that the group definitions are data-driven, leading to biased estimators and invalid inferences. A new method is proposed for inferring social connectedness that remains valid even when the groups are endogenously selected. Theoretical guarantees are first established. Then the approach is evaluated via simulations and in an empirical application.
