INTRODUCTION
Do individuals shy away from or, instead, actively consume information about public threats? Most extant scholarship suggests that people privilege negative information, and some (but not all) argue that threats motivate attention, yet most research focuses on single-country contexts and information that is particularly relevant. Consider terrorism: while it has increased in global scope and lethality, experiences and concerns vary across countries. Does this variation lead to different responses to news about terrorist threat (vs. positive news), or does human nature spur heightened attention regardless of context? We lack consensus on this question. Where news of terrorism is more salient, some argue the public is prone to pay more attention to the threat (e.g., Merolla et al., Reference Merolla, Montalvo, Zechmeister and Zizumbo-Colunga2011); however, others note that consistently high terror warnings may leave the public complacent with respect to news of terrorist threat (e.g., Hoffman et al., Reference Hoffman, Kowal and Kaire de Francisco2013).Footnote 1 Where terrorism is less prominent, individuals may startle easily or, conversely, they could disregard what might seem to be out-of-left-field threats. Via a multi-country research design, we assess the degree to which people across distinct contexts react the same, or differently, to news of a collective threat. We focus on the case of international terrorist threat, yet also test the robustness of our conclusions to domestic terrorist threat, economic threat, and crime.
The general proposition we test is that individuals exposed to news about a pressing threat will be more inclined to “tune in” to this information, compared to exposure to more positive news. Consistent responses across distinct contexts, and across threats, would affirm the robustness of negativity bias—the tendency to privilege negative information—in the context of media stories about threat. We investigate this topic with data from nine original experiments conducted in eight countries in 2012. Five studies were conducted with near-nationally representative adult samples via internet panels in France, Spain, Turkey, the United Kingdom, and the United States. Another four were conducted in face-to-face interviews with representative adult samples in major urban areas in Albania, Ecuador, Peru, and Turkey. Each treatment condition contained one of a set of news stories about either a threat or “good times.” Subsequent to this exposure to a news story, subjects responded to a survey that began with questions designed to measure information acquisition.
Across all nine studies, those presented with news of international terrorist threat demonstrate higher mean levels of information acquisition compared to those presented with positive news. Moreover, individuals in the threat conditions were more motivated to return to the article to acquire or confirm information. Furthermore, the pattern of results is robust to tests of reactions to news about domestic terrorist threat, potential economic recession, and a crime threat. Finally, they also hold in a follow-up study in which the information acquisition question is identical across experimental conditions. In short, this project demonstrates that the effect of news about pressing collective threats on the public's orientation toward information is exceptionally general. The effect is not conditional on whether the threat has saturated the environment, and is found consistently across different study modes, questions, and types of threat.
TERRORIST THREAT AND INFORMATION ACQUISITION
A number of theoretical perspectives suggest that negative information that accompanies threat can stimulate attention and information seeking. We ask the following question: Is this tendency present in individuals irrespective of the status quo environment with respect to threat levels and experiences?
A long line of scholarship finds a tendency to privilege negative information (see, e.g., Baumeister et al., Reference Baumeister, Bratslavsky and Finkenauer2001). For example, individuals are more sensitive to losses than gains (Kahneman and Tversky, Reference Kahneman and Tversky1984; Slovic, Reference Slovic1969). Negative considerations of candidates are stronger predictors of feeling thermometer ratings than positive considerations (Lau, Reference Lau1982). Many find that negative campaign advertisements have greater influence, especially with respect to conveying information, than positive advertisements (Freedman et al., Reference Freedman, Franz and Goldstein2004; Fridkin and Kenney, Reference Fridkin and Kenney2008; Geer, Reference Geer2006; Lau, Reference Lau1982; for a meta-analysis, see Lau et al., Reference Lau, Sigelman and Rovner2007). Negative ads also evoke a greater physiological response and individuals exposed to negative political ads are more likely to recognize information from those ads (though they also tend to over-report recognition of information; Bradley et al., Reference Bradley, Angelini and Lee2007). In general, electrocortical activity in the brain responds differently when evaluating negative stimuli compared to positive or neutral stimuli (Ito et al., Reference Ito, Larsen, Kyle Smith and Cacioppo1998). Some find that individuals in a negative emotional state, in particular an anxious one, are more likely to learn the issue positions of candidates (Brader, Reference Brader2005, Reference Brader2006; Marcus et al., Reference Marcus, Russell Neuman and MacKuen2000), pay attention to and learn about national security threats (Huddy et al., Reference Huddy, Feldman, Cassese, Russell Neuman, Marcus, Crigler and MacKuen2007), and seek out and retain threat-relevant information (Albertson and Gadarian, Reference Albertson and Gadarian2015; Gadarian and Albertson, Reference Gadarian and Albertson2014).
Negative information is more likely to draw one's attention, which in turn makes this information more effective in the formation of impressions of people or things, decision-making, and, the domain we are focused on, information acquisition. Why does negative information tend toward such strong effects? There are several different causal explanations for negativity bias. According to expectancy-contrast theories, given that most experiences are positive, negative information stands out as extreme, and thus it is more effective in shaping attention and decision-making (Helson, Reference Helson1964; Lau, Reference Lau1982, Reference Lau1985; Sherif and Sherif, Reference Sherif, Sherif, Sherif and Sherif1967; see also Fiske, Reference Fiske1980). A kindred explanation for negativity bias is found in frequency-weight theories, which posit that negative cues are more effective in that they are more informative or novel (Lau, Reference Lau1982, Reference Lau1985). Others argue that a mechanism lies in the negative emotions, particularly anxiety, that arise in reaction to threat (MacLeod and Mathews, Reference MacLeod and Mathews1988; Marcus et al., Reference Marcus, Russell Neuman and MacKuen2000). Negative emotions can activate a person's surveillance (threat-detecting) system, which “stimulates peoples’ attention (Marcus and MacKuen, Reference Marcus and MacKuen1993, 678).” Underlying most, if not all, of these perspectives is the notion that a tendency to privilege negative information is a basic human trait, which evolved over time, as it has been beneficial to human survival (Baumeister et al., Reference Baumeister, Bratslavsky and Finkenauer2001). Our goal is not to adjudicate across the mechanisms highlighted in these different theories; they all suggest that the negative information that accompanies public threats should increase information acquisition.
Nonetheless, some scholarship runs counter to this expectation. For example, Nadeau et al. (Reference Nadeau, Niemi and Amato1995) find that anxiety (absent hope) does not have the expected effects on learning, in the case of Quebec and language policy. Variation in the emotions evoked by a threat could cause variation in information orientations. Terrorism tends to activate both anger and fear (see Carver, Reference Carver2004; Harmon-Jones et al., Reference Harmon-Jones, Harmon-Jones, Abramson and Peterson2009; Lerner and Keltner, Reference Lerner and Keltner2000; Merolla and Zechmeister, Reference Merolla and Zechmeister2009). While fear is associated with more thoughtful processing, individuals made angry may favor more direct action over information seeking and processing (MacKuen et al., Reference MacKuen, Wolak, Keele and Marcus2010; Tiedens and Linton, Reference Tiedens and Linton2001; see also Valentino et al., Reference Valentino, Hutchings, Banks and Davis2008). The particular context into which a threat enters may also matter. With respect to terrorism, countries vary in their experiences with plots and the degree to which the environment is already replete (or not) with threats. Against a norm of constant threat, positive information may be more novel and grab one's attention. In short, it is an open question just how robust is the notion that threat stimulates information acquisition across distinct countries and threat contexts.
DATA AND METHODS
To test the degree to which a collective threat stimulates information acquisition, and to assess the generalizability of such a connection, we examine data from nine original experiments conducted in eight countries in the summer of 2012 (Merolla and Zechmeister, Reference Merolla and Zechmeister2018). The countries were selected to provide a range of contexts along two key dimensions: nature of democracy and experience with terrorist attacks. In France, Spain, Turkey, the United Kingdom, and the United States, we implemented the study via the internet to near-nationally representative adult samples. In Turkey (again), Albania, Ecuador, and Peru, we embedded the study within face-to-face interviews of representative adult samples in the major metropolitan areas.Footnote 2 At the time, Spain, Turkey, the United Kingdom, and the United States had experienced an attack by Al Qaeda. France was a likely target for a future attack, and in the intervening years the country has experienced a number of atrocious attacks, including the Paris bombing attacks in the fall of 2015. At the time of the study and since, an international terrorist attack seemed quite remote in Albania, Ecuador, and Peru; however, the latter two countries had indirect and direct experience with domestic terrorism, something that we took into account in the study design.
The basic study protocols were standardized. Participants first consented to the study, and then responded to a pre-treatment survey that asked about demographic and socioeconomic characteristics and political predispositions. The subjects were randomly assigned to a “good times” news condition or a “threat” news condition.Footnote 3 Those in the treated conditions were asked to read a news story. They were then asked two close-ended questions that asked them to recall two facts from that article, with the option to consult the news story as needed. They were then asked questions about their emotional state and others not focused on here, and then were debriefed.
Our focus is on individuals assigned either to the good times condition or a threat condition. Table 1 presents the number of observations within each of these cells (and mode), for each country. The core feature of the experimental design is the set of short (~400 to 500 words) news stories that were randomly assigned to treated subjects and followed a similar template across all countries (see the Supplementary Appendix for experimental treatments). For all treatments, the information presented was drawn from actual sources but edited together by the authors and modeled after instruments used in similar types of research (e.g., Gadarian, Reference Gadarian2010; Merolla and Zechmeister, Reference Merolla and Zechmeister2009).
Table 1 Countries, Mode, and Observations in Study Treatment Conditions
The intention of the good times news story was to present positive indicators from around the world and within the country. The treatment begins with a statement that the country is “headed toward a time of increased well-being.” It refers to positive trends in areas such as education, the environment, and health in the country and, as well, the world. The first paragraph ends with a note that, according to a recent survey, a “majority” in that country report “moderate to high levels of life satisfaction.” The next four paragraphs focus on positive information about education, the environment, science (e.g., energy use), and health and welfare. Each paragraph situates information about the country in the context of broader, global positive news.
We included two international terrorism threat conditions, which varied only with respect to the last paragraph: for reasons unrelated to this study, one condition (labeled Int'l terror in Table 1) did not end with a reminder of democratic values, while the other did (labeled Int'l terror reminder in Table 1). In four cases, we also included a domestic terrorist threat news story (Ecuador, Peru, Spain, and Turkey face-to-face); in six cases, we included an economic threat story (Albania, France, Spain, Turkey online, the United Kingdom, and the United States); and in one country we included a story about crime as a threat (Ecuador). This allows us to check the robustness of our conclusions to news of other threats.
For the international terrorist threat news story, the first paragraph referenced warnings that the country is “on the brink of experiencing a major terrorist attack,” placed this in the context of increased global vulnerability, and noted that a majority of individuals in the country are somewhat to very worried about a future attack. The next paragraph referenced the increased danger posed by terrorism and referenced the 2008 Mumbai, India attack by Al Qaeda. The third paragraph referenced Al Qaeda's intentions to continue to mount coordinated, lethal attacks on citizens in various public areas. The fourth paragraph referenced the risk of biological and chemical weapons. The fifth paragraph referenced a statement by a public official about the lethal intentions of terrorists. The international terrorist threat with a reminder of democratic values news story differed only in that it ended with an additional, final sentence that referenced a statement by all leaders in the political system urging people “to protect democracy” by adhering to “core democratic values, such as liberty and tolerance, and respect for fundamental democratic practices, such as free and fair elections and an independent judiciary.”
The domestic terrorist news, economic recession, and crime stories also were developed with a common structure. The Supplementary Appendix contains the outline (in English) of the treatment followed in each country and the full (language-specific) treatment used in each respective case.
As a manipulation check, we included questions to assess the extent to which the threat conditions increase negative emotions and decrease positive emotions relative to the good times conditions. Respondents were presented with 10 emotions, as recommended by Marcus et al. (Reference Marcus, MacKuen, Wolak, Keele and Redlawsk2006) and asked, for each one, to “indicate to what extent you are feeling this way right now” on a 1–5 scale. The emotions are: Afraid, Anxious, Worried, Enthusiastic, Hopeful, Proud, Hatred, Contempt, Bitterness, and Resentful.
We performed a principal components factor analysis on the 10 questions for the pooled dataset, and found 3 factors with eigenvalues over 1.0. The first (eigenvalue 3.94) is characterized by negative emotions related to anger, with high rotated factor loadings for hatred, contempt, bitterness, and resentful (≥0.75). The second (eigenvalue 2.07) is characterized by negative emotions related to fear, with high rotated factor loadings for afraid, anxious and worried (0.81 and higher). The final factor (eigenvalue is 1.08) is characterized by positive emotions, with enthusiastic, hopeful, and proud loading highly (≥0.78). We regressed dummy variables for the international terrorism threat conditions (the only threat conditions included in all studies) on each emotions factor. Individuals in the international terrorism conditions are significantly more angry and less positive than individuals in the good times condition (p < 0.001). We do not find a significant difference on the anxiety/fear factor. Thus, the terrorism treatments were effective in increasing some negative emotions and decreasing positive emotions relative to those in good times.Footnote 4
Our core dependent variable is based on the two close-ended questions that were asked following presentation of the news stories (and prior to the emotions battery). During this time, the participants were offered the chance to return to the article. The first question was very similar across all treatment conditions. It referenced survey results that were found in the first paragraph of each article,Footnote 5 and asked whether the finding applied to “more than half” or “less than half” of those interviewed. For the good times condition, the question asked about the proportion reported to be moderately to highly satisfied with their lives; in the case of the terrorist threat condition, the question asked about the proportion worried about the threat of terrorism. In both cases, the correct answer (per the news story) is more than half. The second question was also dichotomous and varied across treatments. For good times, the question was whether the news story reported that global air quality has improved or deteriorated in the past decade (correct answer is improved); for the international terrorist threat conditions it asked whether more or less than 100 people were killed in Al Qaeda's attack in Mumbai (the correct answer is more than 100).Footnote 6
RESULTS
Are individuals presented with news about terrorist threat motivated to acquire and recall more information than those who are presented with news about “good times?” The data allow us several ways to triangulate over this question. First, we created an Information Acquisition tally (0, 1, or 2) of the number of correct responses individuals gave to the two post-treatment questions. We assess our expectation first by comparing differences in values on this measure between the good times condition and each terrorism condition.Footnote 7 As a robustness check, we also assess correct responses to only the first question, which was more similar across conditions.Footnote 8
Figure 1 presents mean values on Information Acquisition for the good times and the international terrorist conditions; the whiskers represent 95% confidence intervals. Across all nine studies, without exception, the mean is higher in the threat conditions compared to good times. As the variable is trichotomous, we assess differences using chi-squared tests. The difference between the terror threat and the good times conditions is significant at p < 0.01, in 16 out of 18 tests.Footnote 9 To probe further, we examined just the proportion who gave a correct response to the first question, which was more standard across conditions. In each case, a higher proportion responded correctly in the terrorism conditions compared to good times. In 12 out of 18 pairwise comparisons (via difference of proportion tests), the difference is statistically significant at p < 0.1, two-tailed (see Appendix Figure 1). In sum, we find strong support for our expectation: those in terror threat conditions demonstrate having attained more correct information than those in the contrasting “good times” condition.Footnote 10
Figure 1 Mean Information Acquisition Levels across Treated Conditions, by Study
How robust is this finding to other threat conditions? We also examined information acquisition for those in the four domestic terrorist threat conditions (included in the studies in Ecuador, Peru, Spain, and Turkey face-to-face). In each case, mean Information Acquisition is significantly higher in the domestic terrorist threat conditions than in the good times condition, and these differences are significant according to chi-squared tests (p < 0.01, in three cases; p < 0.02 in the case of Turkey face-to-face study; see Appendix Figure 2 for values). We then extended our analysis to the other threat conditions included in the study: economic threat (Albania, France, Spain, Turkey Online, United Kingdom, and United States) and crime threat (Ecuador). In every case, Information Acquisition is higher in the threat condition, and each chi-squared test is significant at p < 0.01 (see Appendix Figure 2). In short, the results are robust across threat types and country contexts.
The study further contains a behavioral component, which allows us to ask whether individuals were more likely simply to recall the information, were motivated to become sufficiently engaged to answer the questions correctly, or both? In short, we can assess whether people merely acquired more information from the threat news stories on the first read, or whether some were induced to put more effort into seeking accurate information. In the online studies, the subjects were provided the opportunity to return to the article prior to answering the question, and that process was recorded—that is, we have a variable that indicates whether the individual answered the question on the first attempt or selected to return to the news story and then come back to answer the information question. As a measure of mere recall, individuals are counted as correctly answering each information question if they got it correct on the first attempt, with those who got it wrong or those who returned to the article coded as 0. We add together the two to create a “mere recall” measure, scored 0, 1, or 2. We further look at those who chose to return to the article rather than answer on the first try. This reflects engagement in the sense of being motivated to pursue correct answers. We created a dummy variable for whether the subject returned to the article for each question, and then add these together to create a “motivated to return” measure (0, 1, or 2). We display results for Mere Recall in Figure 2 and for Motivated to Return in Figure 3.
Figure 2 Mean on Mere Recall, by Online Study
Figure 3 Mean on Motivated to Return, by Online Study
As per Figure 2, the mean number correct on Mere Recall (that is, correct answers among those who did not return to the article) is fairly high. More importantly, in all cases except the United States study, those in the international terror threat conditions were more likely to answer correctly on the first attempt (leaving the United States case aside, a chi-squared test of the difference between the respective threat condition and the good times condition is significant at p < 0.1 for each of the eight comparisons).Footnote 11 This provides evidence that, on average, a terror threat news environment stimulates immediate attention and strong recall.
As per Figure 3, we find that tendencies to return to the news story prior to answering the questions are higher in each of the international terrorist threat conditions compared to the good times condition. As the measure is trichotomous, we performed chi-squared tests of the difference between good times and the international threat conditions, and find statistically significant results in four of the ten comparisons at p < 0.1 (and in four additional cases if a higher threshold is considered, p < 0.2). Though not shown here, the same pattern is found for the domestic terror threat condition included in the Spain study and the economic threat conditions included in these five studies. In short, threat motivates individuals to pay close attention (making them more likely to get it right on the first attempt, mere recall) and, among some, it increases motivations to seek out correct answers when provided the opportunity.
As a final test, we examine a different, directly comparable information question from a United States online study that was conducted in 2016 for a separate purpose. Across all conditions, the study contained an identical question, which asked subjects to recall how many paragraphs were in the article they read (correct response = 1, 0 otherwise). The study included a similar terrorism treatment to the one in 2012, and also had three other terrorism conditions that referenced either Hillary Clinton's stance on the issue, Donald Trump's stance, or both. The terrorism only condition had four paragraphs, while the ones with Clinton's stance, Trump's stance, and both stances, all had six paragraphs. The good times story was different: it focused on only one topic, was more engaging, and was nonpolitical (it was about a dog who gained fame on Facebook). It also had six paragraphs. Per difference in proportions tests to compare across good times and each terrorism condition, we find significantly higher rates of recall for the conditions with terrorism news only (p < 0.01, two-tailed), terrorism with Clinton's stance (p = 0.02, two-tailed), and terrorism with both candidates’ stances (p = 0.01, two-tailed). The proportion correct also is higher in the terrorism with Trump's stance condition, though outside a conventional significance level (p = 0.18, two-tailed).Footnote 12 In short, the notion that collective threat provokes information acquisition proves quite robust.
CONCLUSION
Scholars of public opinion and political behavior have identified a number of differences in the ways in which citizens evaluate and engage in politics under threatening versus better times. We take a step back from political evaluations and behaviors, per se, and examine differences in the extent to which individuals acquire and recall information offered by two distinct news environments: one characterized by threat and the other characterized by indicators of well-being and progress. Given that absorbing relevant information is foundational to the types of attitudes citizens express and the extent to which, and how, they engage in the world around them, it is important to understand how information acquisition and recall differs across bad and good times.
Across nine studies in eight countries, we find strong evidence that individuals are more likely to acquire and recall information when presented with news about threat than with news about better times. In addition, we find similar results in a follow-up study with a question that is identical across conditions and yet peripheral to the actual information content.
The study environment is intentionally artificial so as to increase internal validity, and this comes with some costs to external validity. We bolster against this by using layered news reports with information drawn from actual sources and by drawing from near representative samples of adults in major metropolitan areas and countries. As Shadish et al. (Reference Shadish, Cook and Campbell2002) instruct, external validity is about finding similar results across modes, places, subjects, and instruments. We find similar patterns of results across different modes (online versus face-to-face), across distinct countries, and across stories that varied in small (across the international terrorist threat conditions) to larger (across all threat conditions) ways. Thus, the study provides important perspective on general human inclinations toward threatening (versus positive) news. Simply put, it is human nature to pay more attention to bad news about one's environment as opposed to positive news.
SUPPLEMENTARY MATERIAL
To view supplementary material for this article, please visit https://doi.org/10.1017/XPS.2018.4
