ABSTRACT
In a competitive digital environment, news outlets employ attention-gaining strategies, including the portrayal of conflict. Yet such coverage may backfire. Using an experiment comparing exposure to disagreeable or uncivil conflict news, we examine how articles with different types of conflict relate to intended news engagement through the mediating variables of perceived incivility, news source credibility, and/or surveillance emotions and test the moderating effects of conflict avoidance. We find that in textual, digital settings, uncivil conflict news is high risk and low reward: perceived incivility can increase engagement through surveillance emotions but also can decrease engagement directly and through decreased news credibility.
Acknowledgments
The authors would like to thank the Center for Media Engagement (CME), as well as CME’s funders The Democracy Fund, Hewlett-Foundation, and Rita Allen Foundation for funding this project.
Disclosure Statement
No potential conflict of interest was reported by the author(s).
Notes
1 A factor analysis indicated that participants’ responses to the emotion items in our study loaded on the same factor and the items were highly correlated (Pearson’s r = .73, p < .001).
2 To ensure that we adequately manipulated disagreeable and uncivil conflict, we tested the stimuli via MTurk.com. We measured participants’ reactions to the articles by asking them to describe the presence of various uncivil behaviors (e.g., polite/rude, calm/agitated, unexaggerated/exaggerated; averaged measure range from 1 = extremely civil to 5 = extremely uncivil). Using one-way t-tests, we verified that all uncivil conflict articles were perceived as significantly more uncivil than the midpoint of 3 (Range M = 3.35, SD = 0.76 to M = 3.65, SD = 0.72) and all of the disagreeable conflict articles were perceived as neutral or significantly civil (Range M = 2.89, SD = 0.69 to M = 2.90, SD = 0.61).
3 We also included an item about whether the content was emotional/unemotional, following Mutz and Reeves' (Citation2005) perceived incivility measure. We dropped the emotional/unemotional item from the analysis. The results were nearly identical when this item was included. The only difference was that the moderated mediation was significant with a 90% CI rather than a 95% CI (see Appendix C).
4 Non-hypothesized indirect pathways were not included in the models but are included in Appendix D. The only difference in results is that the total indirect effect is no longer significant.
5 The interactions among the article conflict, headline variables, and issue were not significant. We report statistical tests related to the headline and issue conditions in Appendix E.
6 Hayes (Citation2018) suggests comparing partially standardized indirect effect coefficients when independent variables are dichotomous. These coefficients are nearly identical to those presented in the main text (Appendix G).
7 We asked participants to complete three items related to their expectations of the article content, with response options ranging from 1 to 5. The mean responses were significantly less than the midpoint of 3 (and therefore “expected”) for each item.