Framing Senate Candidates in US Elections Coverage: A Topic-Network Approach

• In Event: Structure and Change in Strategic Communication Networks

Abstract

The growth in computational tools big data analysis allows researchers to revisit looming questions in prominent communication theories, such as framing (Matthes, 2009). In political communication, scholars have focused on two crucial frames in media coverage of elections, strategy and issue frames (Cappella & Jamieson, 1997) Strategy frames focus on winning and losing, language of war, performance and style, and horse-races. Issue frames focus on candidates’ approach to policy and decision making. The emphasis of strategy or issue framing can have an impact on cynicism towards the political process news items and political knowledge(Zoizner, 2018). However, framing research was often limited by methodological constrains, including lengthy and costly manual content analyses. In this study, we combine topic modeling and network analysis to automatically identify issue and strategy frames in a large corpus, and examine changes in strategy and issue framing emphasis in the news coverage of 330 Senate candidates in five election cycles (2008-2016).

While studies on the effects of issue and strategy framing were largely consistent in their findings, questions regarding the factors that shape the use of frames and changes in use over time offer mixed findings. Some researchers argued for a rise in strategy framing over time (Patterson, 1994), while others found inconsistent use patterns(Benoit, Stein, & Hansen, 2005). Some have pointed to moderators such as media type (Aalberg, Strömbäck, & De Vreese, 2011), yet more factors remain unexplored.

One possible reason for the mixed findings might be methodological constrains. Framing studies often required the creation of a codebook, training coders and a lengthy process of hand coding. This limited the possible breadth of analyses, making it hard to analyze contexts such as the one used in this study – the comparative study of multiple Senate candidates over multiple election cycles. The method we suggest combines topic modeling and network analysis in a multi-stage process. As our initial findings show, this method is efficient in estimating issue and strategy frame ratios in corpora and individual texts.

We first conduct topic modeling - a semi-automated, unsupervised method for text analysis (Maiya & Rolfe, 2014). The algorithm extracts a set of topics, frequency distributions of words that tend to co-occur, from which the given corpus could have been created. Second, we create a topic-network by calculating the pairwise similarity between topics based on word distributions. Each topic serves as a node in the network, and words similarity as edges. Finally, we use community detection (Walktrap, see Pons and Latapy, 2005) to identify coherent clusters of topics, conceptualized as frames. All coverage of 330 candidates in the 2008-2016 U.S. Senate elections was downloaded from LexisNexis (periods of six months before elections date for each case). To validate whether our classification extracts media frames, we randomly sample 100 articles from our data and hand-code them for their use of strategy and issue framing based on Aalberg et al. (2011). After validation we further analyze the use of frames over campaign timelines.

This study offers an inductive, unsupervised, and cost-efficient method for framing analysis, that also advances our understanding of framing dynamics in the news coverage of elections. Future studies could employ this method to comprehensively study framing dynamics in other political contexts.


References
Aalberg, T., Strömbäck, J., & De Vreese, C. H. (2011). The framing of politics as strategy and game: A review of concepts, operationalizations and key findings. Journalism, 13(2), 162–178.
Benoit, W. L., Stein, K. A., & Hansen, G. J. (2005). New York Times Coverage of Presidential Campaigns. Journalism & Mass Communication Quarterly, 82(2), 356–376.
Cappella, J. N., & Jamieson, K. H. (1997). Spiral of cynicism: the press and the public good. New York: Oxford University Press.
Maiya, A. S., & Rolfe, R. M. (2014). Topic Similarity Networks: Visual Analytics for Large Document Sets. ArXiv:1409.7591 [Cs, Stat]. Retrieved from
Matthes, J. (2009). What’s in a frame? A content analysis of media framing studies in the world’s leading communication journals, 1990-2005. Journalism & Mass Communication Quarterly, 86(2), 349–367.
Patterson, T. E. (1994). Out of order: An incisive and boldly original critique of the news media’s domination of America’s political process. New York, NY: Vintage.
Pons, P., & Latapy, M. (2005). Computing Communities in Large Networks Using Random Walks. In pInar Yolum, T. Güngör, F. Gürgen, & C. Özturan (Eds.), Computer and Information Sciences - ISCIS 2005 (pp. 284–293). Springer Berlin Heidelberg.
Zoizner, A. (2018). The Consequences of Strategic News Coverage for Democracy: A Meta-Analysis. Communication Research, 1–23.

Authors

• Dror Walter, Georgia State University

• Yotam Ophir, University of Pennsylvania