Hybrid representative sampling of social media

Abstract

Social media communities are understudied due to the difficulty of obtaining data with any amount of generalizability. This research proposes a method of hybrid probability sampling of social media ecosystems (collections of communities) using network theory, stratified sampling methods, theoretical saturation, and widely available social media activity statistics. Because of the unique venue of research, these methods create an opt-in sample, where every member of a community is given an equal chance to participate. The proposed methods were then used to obtain a sample representative of a digital political ecosystem with a population size of 8 million. The results illuminate stark polarization resulting from algorithmic opinion aggregation, with implications in online extremism, media literacy, demographic representation in public discourse, and more.