Media Coverage and Public Approval of the U.S. Supreme Court

Abstract

When citizens believe the U.S. Supreme Court makes decisions in an insincere or politicized manner, their specific support for the institution can decline. The Court’s relative aversion to publicity means the media are the primary source of information about its decisions. We design a survey experiment that varies the type of coverage—game frame or principled. Game-frame coverage reduces agreement with and acceptance of the decision discussed. We then classify and analyze more than 1,000 transcripts of broadcast coverage of salient Court decisions from 1990–2010. Not only has game-frame coverage of the Court increased, but this coverage partially explains recent declines in specific support for the institution.

Keywords:

• Supreme Court
• public opinion
• television news media
• game frame

Supplemental Material

Supplemental data for this article can be accessed at https://doi.org/10.1080/10584609.2018.1467517.

Notes

1. Indeed, television reporting of the Court’s decisions tends to be infrequent, brief, and at times factually incorrect (Slotnick & Segal, 1998).

2. This corpus of transcripts represents the universe of coverage of politically salient decisions 1990–2010 on the three major networks and Fox News, CNN, and MSNBC. Politically salient decisions are those reported on the front page of the next day’s New York Times (Epstein & Segal, 2000). Our measurement strategy is conservative. Even some salient cases fail to receive television coverage (Slotnick & Segal, 1998). Searching for coverage of only salient cases as opposed to all decisions, if anything, biases us against finding an association between coverage and public opinion.

3. Johnston and Bartels (2010) and Zilis (2015) also present evidence supporting this general conclusion. These studies motivate our inquiry, which builds on this literature by considering the role of game frames as a driver of politicized perceptions and by studying the effect of aggregate coverage on public opinion over time.

4. We selected our two cases using several criteria. First, decisions must be issued between 2000 and 2015 to ensure a large number of observations for MSNBC, which began broadcasting in 1996, and Fox News Network, which began an incremental rollout in the late 1990s. Second, a decision must be covered on the front page of the New York Times (Epstein et al., 2012), a common criterion for salience in studies of the Court (Epstein & Segal, 2000).

5. This duration of time ensures a larger number of observations while also capturing a range of coverage from initial reporting on the decision, to subsequent interpretation of arguments and discussion of implications.

6. According to an LNA Content report generated September 21, 2015 (upon data collection), LNA coverage for ABC starts January 1, 1980; for CBS February 1, 1990; for CNN February 10, 1990; for NBC July 1, 1993; for MSNBC November 8, 1999; and for Fox News December 28, 1997. These dates were collected via the sourcebook on LNA, and confirmed with the assistance of librarian Stephanie Braunstein and technical specialists at LNA. See supplemental Appendix I for an inclusive list of all shows in the sample and show coverage in LNA. Of course, these variations in data availability by network suggest that our measure potentially undercounts coverage in certain years. Fortunately, our primary statistical conclusion, that increases in the volume of game-framed coverage are associated with decreases in specific support for the Court, is robust to restricting our latter analysis to the periods 1993–2010, 1997–2010, and 2005–2010. These latter analyses do not include control variables due to a lack of degrees of freedom to estimate the full error correction model. As such, we are confident that our inferences are not conditional on variation in data availability by network from LNA. Future scholars could and should consider expanding the time horizon of this analysis as data become available.

7. The nature of Supreme Court cases means that the words “winner” and “loser” could conceivably be utilized by broadcasters without any additional game framing in describing case outcomes. “win,” “lose,” and derivatives thereof comprise about 20% of the volume of game-frame words in our sample. We created an alternative measure of game framing by removing these words from the dictionary. The alternative measure correlates at 0.98 with the measure we use here, and all statistical results are robust to this alternative. Therefore, the fact that coverage of a Court decision may contain the words “winner” and “loser” as a function of an adversarial judicial system does not appear to impact this analysis.

8. For robustness, we utilized supervised ensemble classification methods as an alternative strategy. Results were robust to this method. Contact the authors for details.

9. The total volume of stories is positively and statistically significantly correlated with the average amount of game framing, and both indices generally increase over time. As such, disentangling the effect of coverage increases from the effect of the average amount of game framing in any single given story is challenging give our data. Separate Wilcoxon rank-sum tests indicate that there is a statistically significant difference in the distribution of both game framing in individual stories as well as total stories in the latter (more recent) half of our data set; both these variables appear to increase over time. Fortunately, our results do not depend on the dictionary classification exercise alone, mitigating against this concern.

10. One option might be to present unframed coverage of a decision as a control. This approach is problematic because, as our text analysis demonstrates, most coverage features some proportion of game-frame language, making “neutral” coverage unlikely. Moreover, people often perceive biases in neutral coverage (Vallone et al., 1985), so such coverage is unlikely to even be without any effect.

11. It is possible to scale these survey responses as an index (percentage favorable-unfavorable). Results reported below are robust to this alternative.

12. The specific support series has already been first-differenced to render it stationary, due to the results of unit root tests. The results reported in Table 2 are robust to using the undifferenced, unit root, series as the basis for the dependent variable. This alternative strategy would, however, require that all the variables in the model were not of the same order of integration, contra the advice of both Grant and Lebo (2016) and Keele and colleagues (2016). We therefore do not present these alternative results here.

Additional information

Funding

This work was supported by the Media Effects Lab and the Carville Professorship (held by Searles), both in the Manship School of Mass Communication, and Louisiana State University.

Notes on contributors

Matthew P. Hitt

Matthew P. Hitt is an Assistant Professor in the Department of Political Science at Colorado State University, USA. Kathleen Searles is an Assistant Professor in the Manship School of Mass Communication and the Department of Political Science at Louisiana State University, USA.

Kathleen Searles

Matthew P. Hitt is an Assistant Professor in the Department of Political Science at Colorado State University, USA. Kathleen Searles is an Assistant Professor in the Manship School of Mass Communication and the Department of Political Science at Louisiana State University, USA.