The Symbiosis of News Coverage and Aggregate Online Search Behavior: Obama, Rumors, and Presidential Politics

Abstract

Using a relatively new approach, this study examines the agenda-setting effects of television and newspaper coverage of a prominent rumor from the 2008 presidential election: the rumor that Barack Obama was secretly Muslim. In doing so, we look at the relationship between online information-seeking behavior and mass media news coverage, expecting online behavior, such as search, to be a function of exposure to conventional news coverage rather than vice versa. Using Google search trends as a novel search behavior measure, we demonstrate that volume of news coverage positively predicts spikes in aggregate search.

A previous version of this article was presented at the 2009 AEJMC conference in Boston, MA.

Notes

¹Our final search term was BODY (obama w/50 muslim and rumor or secret or secretly or secretly arab or he's an arab or he is arab or muslim name or false rumor).

²For more information about Google Trends, see http://www.google.com/trends.

³Only searches conducted in the United States were included in the results.

⁴For example, the search term Is Obama Muslim had only one-fifth of the searches, as did Obama Muslim.

*p < .05. **p ≤ .001.

Note. R ² = .579; Liung-Box Q = 23.788, df = 17, p = .125.

⁵On the suggestion of an anonymous reviewer, we also tested both of our hypotheses regarding newspaper coverage using the entire U.S. daily newspaper database in LexisNexis. The results were very consistent with our tests using the selected newspapers. Using the same procedures discussed in the Method section, our expanded search yielded 839 articles. The same-day bivariate relationship between newspaper coverage and Google searches was r = .182, p = .022. We then ran an OLS regression and newspapers again failed to reach significance (p = .086, Durbin-Watson =.56). Finally, we ran a 1, 0, 0 ARIMA model using television and newspaper stories as predictors. Television stories remained a significant predictor of searches (p = .038), whereas newspaper stories were again not significant (p = .386; Model Statistics: R ² = .580; Liung-Box Q = 23.673, df = 17, p = .129). The expanded newspaper search also continued to support H2: Bivariate correlations were strongest on the same day and gradually became weaker on subsequent days (Same-day: r = .182, p = .022; Day 2: r = .108, p = .180; Day 3: r = .072, p = .371; Day 4: r = .061, p = .451; 1 week later: r = .018, p = .829). Finally, as in the selected search, a day's searches also failed to predict newspaper coverage 1 week later in the expanded search (r = −.069, p = .402).

Additional information

Notes on contributors

Brian Weeks

Brian Weeks (M.A., University of Minnesota, 2010) was a graduate student at the University of Minnesota–Twin Cities at the time of this study. He is presently a Ph.D. student in the School of Communication at The Ohio State University investigating political communication, rumors, and media effects.

Brian Southwell

Brian Southwell (Ph.D., University of Pennsylvania, 2002) is an Associate Professor in the School of Journalism and Mass Communication, University of Minnesota–Twin Cities. His research interests include campaign evaluation methods, including time-based designs, and the intersection of mass communication and interpersonal interaction.