Combining Survey and Commercial Data to Study Political Behavior on Social Media

• In Event: Communicating Politics on Twitter

Abstract

Social media platforms allow individuals to directly contact their elected officials, but the prevalence of this kind of political behavior - and the attributes of those who engage in it - are largely unknown. We use two Twitter data sources covering President Trump’s impeachment trial to measure how U.S. adults engage with elected officials on Twitter. First, we contracted with a commercial data vendor to collect a census of public tweets that mention members of Congress and use the term “impeachment.” Second, we conduct a representative survey of U.S. adult Twitter users and link responses with their social media data. The first data source includes hundreds of thousands of impeachment-related tweets directed at elected officials. However, a relatively small number of accounts generate a large share of these tweets, and a large number of them are duplicative. Using representative survey data we find that relatively few (and more politically engaged) Twitter users tweet at lawmakers, while a larger share follow members of Congress on the platform and retweet their posts. In sum, these results show that commercial social media data sources may overstate political behavior online, relative to survey-based benchmarks. We discuss the implications of this disconnect for political communication and behavior research and describe methods for assessing the representativeness of tweets.

Authors

• Adam G. Hughes, Pew Research Center

• Galen Stocking, Pew Research Center

• Sono Shah, Pew Research Center

• Michael Barthel, Pew Research Center