Indexing theory during an emerging health crisis: how U.S. TV news indexed elite perspectives and amplified COVID-19 misinformation

ABSTRACT

This paper applies indexing theory to test whether U.S. television news about COVID-19 covered misinformed elite viewpoints and whether indexing patterns were consistent across networks. We extend theory by investigating an emerging crisis where information was in flux. We conducted a content analysis of U.S. broadcast and cable news coverage of two COVID-19 issues: masks and disinfectants/UV light. Coverage responded to changes in health institution guidance related to mask wearing, but: (a) mentioning partisan elites was related to misleading content, (b) at times mentioning health institutions was related to decreased inclusion of correct information, and (c) at times patterns of indexing differed across networks. Findings suggest that indexing practices may encourage misinformation spread during emerging crises, especially on partisan news.

KEYWORDS:

• Misinformation
• COVID-19
• television
• content analysis
• indexing theory

View correction statement:

Correction

Disclosure statement

No potential conflict of interest was reported by the author(s).

Correction Statement

This article was originally published with errors, which have now been corrected in the online version. Please see Correction (https://doi.org/10.1080/23808985.2022.2185386)

Notes

1 See a study published in 2017 by Hamlin titled Ultraviolet Irradiation of Blood: ‘The Cure that Time Forgot’? available at https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6122858/.

2 We applied spaCy using the English multi-task CNN pretrained model trained on OntoNotes.

3 We filtered out irrelevant entity types (i.e. language, date, time, percent, money, quantity, ordinal, and cardinal) since such entities would not correspond to the people and organizations of interest in this project.

4 Specifically, coders were asked to identify whether (a) any information relevant to the topic was correct, (b) any information relevant to the topic was misleading, and (c) any information relevant to the topic was incorrect. If coders identified misleading or incorrect information in a segment, they took one more step to indicate whether that misleading or incorrect information was corrected [0 = not corrected at all, 1 = implied correction (e.g. the head of the CDC now says that guidance could be changing), or 2 = clear correction of information (e.g. The President’s remarks were widely condemned by medical experts)]. Coders struggled to make meaningful distinctions among the correct/implied correction/clear correction categories and the misleading/incorrect categories; so, after coding, the ‘correct,’ ‘implied correction,’ and ‘clear correction’ categories were combined to indicate a final ‘any correct’ claims code and the ‘misleading’ and ‘incorrect’ categories were combined to indicate misleading/incorrect claims.

5 We also ran a multinomial logistic regression; the results were consistent.

6 (1) Whether the before/after CDC announcement variable interacted with the networks, (2) whether a segment mentioned a Republican/conservative interacted with the networks, (3) whether a segment mentioned a Democratic/liberal interacted with the networks, and (4) whether a segment mentioned a health elite.

7 (1) Whether the before/after CDC announcement variable interacted with mention of a Republican/conservative, (2) whether the before/after CDC announcement variable interacted with mention of a Democratic/liberal, (3) whether the before/after CDC announcement variable interacted with mention of a health elite.

8 The emmeans package estimates probabilities by averaging over the levels of the other variables in the model.

9 Only four of the segments relevant to the disinfectant/UV light topic aired prior to April 23, 2020, when Trump made his statement about this topic, so we cannot compare correct/misleading content before and after his statement.

10 The low number of elite mentions may be due to our conservative decision to include full names (e.g. Donald Trump) and not include only last names (e.g. Trump) in the list of elites when there are numerous public-facing people with the same last name. When we do include ‘Trump’ in the list of Republican elites, the pattern of our findings is largely the same as what we cover in this paper.

Additional information

Funding

This work was supported by Mark Gibson, CEO, Capital Markets, JLL Americas.