What Do We Learn About Social Networks When We Only Sample Individuals? Not Much

Abstract

Much of the empirical analysis of social networks is based on a sample of individuals, rather than a sample of matches between pairs of individuals. This paper asks whether that approach is useful when one wants to understand the determinants of variables that are inherently dyadic, such as relationships. After reviewing the shortcomings of the data used in the literature, we use Monte Carlo simulation to show that the answer is positive only when relationships are themselves randomly formed, a very special and uninteresting case. Additional work that supports strategies to collect dyadic data as part of surveys usually used by economists seems to be needed.