Statistical inference and misperceptions in social networks

Abstract

We investigate how individuals form beliefs about population behavior using statistical inference based on observations of their social relations, which are biased by the friendship paradox. We analyze equilibrium behavior under strategic complementarity and show that population behavior depends on the share of individuals that either do or do not correct their estimate using information about neighbors' representativeness, which we capture by network degree. We illustrate the role of sample size for estimation precision and outline how our framework is useful for further analysis of behavior by agents embedded in networks.