What Predicts the Game Frame? Media Ownership, Electoral Context, and Campaign News

Abstract

While scholars have often bemoaned journalists’ heavy use of game-framed and “horse-race” coverage of elections, the contexts most likely to produce game-framed news have not yet been well identified. Our data collection across three election cycles (2004, 2006, 2008) and various levels of elective office (candidates for governor and U.S Senate), and across multiple media markets and types of news organizations allows us to examine the extent to which all three classes of contextual variables—the internal news-making context, the media economic market context, and the electoral political context—influence the provision of game-frame election coverage. We find that news organizations’ choices to rely heavily on game-frame election stories are dependent on both news-making and political contexts. These findings contribute to the ongoing debate on the relationship of media ownership to news quality, tempered by firm evidence that news outlets of all kinds tend to focus on the “game” of politics when electoral contests are close.

Keywords:

• framing
• game frame
• horse race
• elections
• news

Notes

1. However, evidence does suggest that the game-frame dominance over issue coverage does not necessarily extend to gubernatorial and Senate races and that the amount at these levels is contingent on a host of other race-specific factors (Kahn, 1991, 1995)—some of which we explore in this study.

2. Although there is in theory endless space for news (of all kinds) on the Internet, see Prior (2007) for a discussion of why this seemingly endless space for serious news does not improve democratic learning on the whole. And as we discuss further, the limitations of traditional news media still constrain the news content that most citizens encounter.

3. Work on economic theories of news has identified two general mechanisms by which market features shape political news content: market competition and market level audience demographics. We focus on the former in this study, primarily because we do not have a strong theoretical expectation about how demographic characteristics such as those identified by Hamilton (2004) would predict game-frame coverage as a particular type of election news. Diagnostics of block demographic variables showed no significant influence of demographics and indicated a better fit when not included. Thus, the empirical evidence seems to support the lack of compelling theoretical expectation.

4. Also see Fowler and Ridout (2010), who note that the majority of campaign activity, as measured by political advertising, doesn’t begin until after September 1.

5. NewsBank provides the most comprehensive archive of U.S. local newspapers and back issues are difficult to obtain due to cost and availability. In some cases in which NewsBank’s holdings were incomplete, additional stories were retrieved via the Nexis database or hard copies were retrieved via interlibrary loan. The initial sample of 278 newspapers was whittled to 259 because of missing observations, due to the fact that the few stories from these newspapers did not fit the key categories of story type discussed. The list of newspapers included in our sample is available from the authors upon request. On average, the initial 278 newspapers in the sample represent 44% of all daily newspapers published in each state.

6. Although our data set contains more than 12,000 coded news articles, some data are missing because information could not be located (note that the N for stories in our model is 10,784). Because our data and analysis are multi-level, if we are missing an indicator at the paper level, it results in the loss of all the news stories for that paper. Our sample contains a handful of papers for which we could not verify information on either corporate ownership or circulation, resulting in the loss of several hundred news stories for the models presented. We also lost stories that for whatever reason did not contain a date of publication in our electronic archives.

7. This variable was coded by asking coders to answer the question, “In your opinion, what is the primary focus of the story?” Intercoder reliability on this variable as on our other variables averaged .61 or greater, which Landis and Koch (1977, p. 165) categorize as “substantial” agreement. Please see supplemental Appendix B for more information on coding and ICR.

8. Cable penetration data are taken from TVB.org, the not-for-profit trade association for the broadcast industry. The measure combines wired cable services and alternative delivery systems (i.e., satellite), thus because the measure double counts households with both, the maximum value is slightly over 100%.

9. We should note that the circulation rate is for the individual newspapers and not the circulation numbers for the newspaper chain in cases where papers are owned by chain owners.

10. We estimate a random effect model, as opposed to a fixed effect model, for several reasons: (1) Given the hierarchical structure of the data, the fixed effects model will produce far more parameters, as dummy variables would need to be entered for news organizations and races. (2) The random effects model better accounts for different cluster sizes—for instance, news organizations that release varying numbers of stories. For smaller news organizations, the intercept is more likely to pool around the common intercept. However, it is important to note that in practice, there are typically negligible differences between the fixed and random effects models.

11. Here we also tested interactions between the ownership structures and market penetration. None of the interaction terms yielded significant results.

Additional information

Notes on contributors

Johanna Dunaway

Johanna Dunaway is an Associate Professor of Political Science and Mass Communication, Department of Political Science, Louisiana State University.

Regina G. Lawrence

Regina G. Lawrence is Jesse H. Jones Centennial Chair in Communication, School of Journalism, The University of Texas at Austin.