News, Politics, and Negativity

AbstractWork in political communication has discussed the ongoing predominance of negative news, but has offered few convincing accounts for this focus. A growing body of literature shows that humans regularly pay more attention to negative information than to positive information, however. This article argues that we should view the nature of news content in part as a consequence of this asymmetry bias observed in human behavior. A psychophysiological experiment capturing viewers’ reactions to actual news content shows that negative news elicits stronger and more sustained reactions than does positive news. Results are discussed as they pertain to political behavior and communication, and to politics and political institutions more generally.Keywords: negativity biasmass mediapolitical communicationpsychophysiology Notes1. We do not provide citations to this work here, but the literature on each of these topics is discussed in detail later.2. There is some work that links the negativity bias to evolution, in short: it may be evolutionarily advantageous to prioritize negative over position information; humans may thus be hard-wired to do this; and media content may reflect this tendency. We do not discuss this possibility in detail here; see, for example, Shoemaker, Citation1996; Fuller, Citation2010; Soroka, Citation2014.3. For early work see Feldman, Citation1966; Hodges, Citation1974; Hamilton and Huffman, Citation1971. For more recent work see Fiske, Citation1980; Ronis and Lipinski, Citation1985; Singh and Teoh, Citation2000; Van der Pligt and Eiser, Citation1980; Vonk, Citation1993, Citation1996.4. The implication is that there will be an especially large degree of information processing when someone in a bad mood receives bad news. See Forgas, Citation1992. See also a related body of work on mood-congruence and mood-state-dependent memory (e.g., Bower, Citation1981; Ucros, Citation1989). That said, the emphasis in this work is not on the relative importance of negativity, but rather the relationship between one’s ability to remember positive or negative information based on his or her current emotional state.5. Consider also, for instance, work on “person memory” (e.g., Ybarra & Stephan, 1996), work on performance evaluations of employees and students (e.g., Ganzach, Citation1995; Rowe, Citation1989), work on the effects of positive versus negative events on psychological distress (e.g., Hobfoll, Citation1988, Wells, Hobfoll, & Lavin Citation1999), and on daily “mood” (e.g., David, Green, Martin, & Suls, Citation1997).6. The literature is vast, but see, for example, Tversky, Slovic, and Kahneman, Citation1990; Kahneman and Thaler, Citation1991; Shoemaker and Kunreuther, Citation1979; Arkes and Blumer, Citation1985; Diamond, Citation1988. For a partial review, see Edwards, Citation1996.7. Though note that these “negative voting” results have been contested by other authors, suggesting alternative hypotheses that account for the regularity with which presidents’ parties lose seats in midterm elections (e.g., Hinckley, Citation1981; Cover, Citation1986; Born, Citation1990). Recent work suggests a story more in line with Kernell (Citation1977), but based on a prospect theory account that emphasizes the relationship between disappointment with the current presidential administration and electoral turnout (Patty, Citation2006). Aragones’ (Citation1997) work suggests a related negative-reaction account for declining popularity the longer a candidate stays in office.8. These findings reflected observations in several earlier studies, including Campbell, Converse, Miller, and Stokes’ (1960) work on electoral behavior, and Mueller’s (1973) study of U.S. foreign policy.9. Though note that prospect theory, loss aversion, and asymmetry more broadly construed have played an important role in a number of political science subfields as well. There exist several recent reviews of the political science literature informed by prospect theory; see, for example, Levy, 2003; McDermott, Citation2004; Mercer, Citation2005.10. For a thorough review of psychophysiological approach in communication studies, see Ravaja, Citation2004.11. The experiment was run at 2 different times—42 respondents participating in early 2010, and the remaining 21 in fall 2012. There are no significant differences in results across the two groups, so we lump them together in analyses here.12. There is no evidence that these expert coders had different physiological reactions to news stories; dropping them from the analyses has no significant impact on results, so we include them here.13. See Potter and Bolls (Citation2012) for a particularly useful discussion of heart rate (and other physiological measures as well, including skin conductance) in media studies. These authors also note that the relationship between heart rate and attentiveness is still being explored; and some work suggests that heart rate variability may be a better indicator of information processing (e.g., Ravaja, Citation2004). Our sense of the literature is in line with Potter and Bolls; however, there is a considerable body of evidence (including Lang’s seminal work, cited earlier) suggesting that decreased heart rate indicates increased information processing (i.e., increased attentiveness).14. This and all subsequent analyses were conducted both with a simple dummy variable to capture tone, and by using the interval-level measure produced by the coders. Both work similarly in every case. For the sake of simplicity, we use the simple dummy variable here.15. Basic descriptive statistics for the dependent variable, SCRs, are as follows: mean, 1.602; standard deviation, 1.326; min, 0; max, 7.16. Indeed, respondent IDs account for roughly 45% of the total variance (248.208/551.60).17. Basic descriptive statistics for the dependent variable, SCRs, are as follows: mean, 15.096; standard deviation, .344; min, 13.616; max, 19.174.18. Basic descriptive statistics for the dependent variable, SCRs, are as follows: mean, 76.197; standard deviation, 13.330; min, 30.782; max, 167.612.19. Results are available upon request.20. Predicted values and associated margins of error are based on leaving all other variables in the data set at their actual values, and shifting just the direct and interacted values for tone.21. See note 2.22. Preliminary tests did not reveal stronger negativity biases in these more conservative voters; but this is no surprise given how few conservative voters there are to work with.Additional informationNotes on contributorsStuart SorokaStuart Soroka is Michael W. Traugott Collegiate Professor of Communication Studies and Political Science, and Faculty Associate in the Center for Political Studies at the Institute for Social Research, University of Michigan. Stephen McAdams is Professor and Canada Research Chair in the Department of Music Research, Schulich School of Music, McGill University.Stephen McAdamsStuart Soroka is Michael W. Traugott Collegiate Professor of Communication Studies and Political Science, and Faculty Associate in the Center for Political Studies at the Institute for Social Research, University of Michigan. Stephen McAdams is Professor and Canada Research Chair in the Department of Music Research, Schulich School of Music, McGill University.