Traditional approaches and digital trace data: Applying framing analysis to social media

This paper combines a well-established approach in the social sciences, framing theory, and digital trace data. Framing refers to the process of making certain aspects of a topic more salient and in that way influencing its evaluation by an audience. Political actors and advocacy groups aim to advance their political agenda by emphasizing a certain point of view over competing conceptions regarding the problems underlying a topic and possible solutions. The paper outlines new research directions that arise thanks to digital trace data as compared to traditional framing analysis that typically concentrates on news discourses. As an empirical application, the paper analyzes all tweets of elites from both U.S. political parties (members of Congress, party accounts, presidential candidates, governors, the President) during the year 2015. There are two diverging expectations: On the one hand, partisan framing might vary over topics, as these represent varying degrees of partisan conflict. On the other hand, politicians and parties might expect political gains from aggressive or even divisive framing strategies across all policy fields. Such an excessive framing would hint at strong partisan predispositions. The severe polarization of the American political system might overshadow all political debates on Twitter, in line with notions of a ”culture war” that spills over to social phenomena. Applying computational text analysis, the paper finds that across topics, partisan framing is observable, but especially in contentious topics like Obamacare, climate change and the policies towards Israel. A qualitative analysis identifies common frames across topics. For example, Republicans frame the topics climate change and Obamacare as an illegitimate government overreach in their tweets, whereas Democrats frame regulation on both issues as necessary and beneficial to the current and future generations. Finally, the paper discusses the potentials offered by machine learning techniques and topic models to classify frames.