A network-based approach to estimating partisanship and analyzing change in polarization during the 2016 general election

Abstract

Communication and media research lacks an accessible and systematic approach to measuring political partisanship in decentralized media environments. In this dissertation, a network-based measurement of partisanship is proposed and then used to analyze social media users during a highly contentious general election. Study I (Chapter 2) introduces rtweet, a newly developed open-source software package designed to collect Twitter data. Study II (Chapter 3) then uses rtweet to gather publicly available Twitter data and demonstrate a network-based approach to estimating partisanship. Finally, Study 3 (Chapter 4) extends this network-based approach to analyze change over time in network polarization among partisan and non-partisan users during the 2016 general election. Results showcase the range and validity of network-based estimates of partisanship and provide clear evidence of partisan selective exposure and network polarization on Twitter as proximity to the election increases.