March 30, 2017

Tracking The Tones of Media Coverage

olu-eletu-38649We already know that the tone of media coverage influences people’s attitudes and opinions. But is that influence conditional? Amber Boydstun, an associate professor of political science at the University of California, Davis, addressed this question at last Thursday’s Text as Data seminar titled The Conditional Effects of Media Tone on Public Opinion: The Case of Immigration.

The seminar, based on her Policy Frames Project, which is in collaboration with Dallas Card (Carnegie Mellon University) and Noah Smith (University of Washington), is centered around two hypotheses. First, the effects of media coverage on public attitudes towards immigration is stronger the more coverage there is. And, second, the effects of media coverage on public attitudes towards immigration is weaker the more varied that coverage is.

To test the two hypotheses, the team initially used keyword searches to identify all relevant articles in LexisNexis for a given issue. For instance, immigration has multiple frames, or dimensions, such as the political or constitutional frames. Within each frame are associated keywords like federal and congress or illegal and citizenship. For this particular policy issue, 38,283 archived articles in the database from 1980 to 2012 were collected from 12 major US news organizations such as the New York Times, The Wall Street Journal and The Washington Post.

The following step was manual coding, or trained-human coding, to identify tone and frames of the articles. Approximately 10% of the 38,283 articles were manually coded, meaning over 4,000 articles were individually read and categorized by human coders. The articles are identified as either implicit, like a hard news piece exploring immigration effects on low-wage job prospects or explicit, like an op-ed that points the finger at immigrants for taking Americans’ jobs, and pro-immigration or anti-immigration, followed by identifying the article’s frame. Finally, the team used NLP techniques to mimic the manual coding for the remaining articles.

Their findings show that media tone has a significant effect on people’s attitudes, at least in the case of immigration. That effect is stronger when there’s more overall attention but weaker when information is unconcentrated. The group suspects that their hypotheses also apply to other policy issues, and plan to test their collected data on other topics like same-sex marriage, gun control, tobacco/smoking, abortion, and climate change.

More generally, identifying the “empirical regularities” of how media cues operate is academically and politically beneficial. Further, the Policy Frames Codebook, developed for the project, provides a structure for coding the many frames a policy issue takes, and acts a guiding tool for other researchers seeking to track an issue of their interest and its multiple frames over time.

by Nayla Al-Mamlouk