Finding Patterns in Media Framing of Policy Issues over Time

Amber Boydstun, Associate Professor of Political Science

When a policy issue like immigration or gun control is covered in the media, the tone of the coverage (e.g., pro-immigration vs. anti-immigration) can affect citizens’ perceptions of the issue, even if the facts of the policy issue itself have not changed. Studies also suggest that the tone of coverage is inherently linked with the frame used to portray the issue in the news—that is, which aspect of the issue (e.g., economic, legal, moral) is emphasized. For example, a news story about the hardships faced by immigrants seeking a better life in America is likely to send a pro-immigration signal, whereas a story about crimes committed by undocumented immigrants is likely to send an anti-immigration signal. Thus, in order to understand how public attitudes are influenced by media coverage, we need to understand the common patterns of how tone and framing operate both across policy issues and across time: What are the empirical regularities of how frames and tone cues “behave” in the news? For instance, in news coverage of a given issue, does a given frame consistently get used with a particular tone, or can it shift over time? Do frames evolve over time, morphing into one another? Does the presence of multiple frames previously ignored in media coverage tend to forecast a shift in tone and, thus, a shift in public opinion? This project aims to answer these and other questions using new data and new methods. By combining manual annotation with computational modeling, we have tracked both the tone (pro/anti/neutral) and the emphasis frames (e.g., morality, economic) used in U.S. news stories on each of four policy issues—immigration, gun control, same-sex marriage, and smoking/tobacco—over a 30+ year period (1980–2012). The result is a corpus of nearly 150,000 news articles on these four important policy issues. Our next step is to strategize about how to use this data to answer our questions of interest.