A Vocal Minority: Assessing the Representativeness of Tweeters and Tweets

• In Event: Advances in Analysis of Twitter Data

Abstract

Social media data can provide new windows into important political phenomena, but it’s often unclear how to assess the representativeness of both users and their posts. We develop three samples of Twitter users: two blended samples recruited from representative panel surveys linked with profile information and one based on administrative data (a commercial voter file) matched to profile information. We discuss the demographic and behavioral composition of each sample, including measures of online political engagement. We show that each method results in a sample that is demographically similar to RDD-based estimates of the U.S. adult Twitter population, but that posting behavior differs across the samples. However, one consistent finding is that only a small share of users generate most actual tweets, suggesting that studies focused on posts (rather than users) capture a narrow set of voices online. In addition, we show that surveys are constrained when it comes to rare kinds of social media behavior, while panels based on administrative data are less limited. We conclude by developing general recommendations for researchers looking to study social media by linking accounts with external benchmark data sources.

Authors

• Adam G. Hughes, Pew Research Center

• David Lazer, Northeastern University

• Stefan McCabe, Northeastern University

• William Hobbs, Cornell University

• Emma Remy, Pew Research Center

• Sono Shah, Pew Research Center