1

The Politics of Threat: How Terrorism News Shapes Foreign Policy Attitudes

Forthcoming, Journal of Politics

Shana Kushner Gadarian Robert Wood Johnson Scholar in Health Policy University of California-Berkeley sgadarian@berkeley.edu

This project was supported by a grant from Princeton University’s Policy Research Institute for the

Region. Thank you to Bethany Albertson, Matthew Baum, Larry Bartels, Ted Brader, Marty Gilens,

Tali Mendelberg and seminar participants at Princeton for comments on earlier drafts.

2

Abstract: In this paper, I argue that the features of the media environment after 9/11, particularly

the media’s emphasis on threatening information and evocative imagery, increased the public’s

probability of supporting the policies advocated by political leaders, principally the president. Using

the National Election Studies 2000-2004 panel and a controlled, randomized experiment, I

demonstrate that citizens form significantly different foreign policy views when the information

environment is emotionally powerful than when it is free of emotion, even when the factual

information is exactly the same. Citizens concerned about terrorism are more likely to adopt the

hawkish foreign policy views communicated in threatening news stories when that policy is matched

with fear-inducing cues than when it is not. These findings suggest that the role of the media is

broader than simply providing a conduit for elites to speak to the public; the media influences the

public through their own means as well.

Keywords: terrorism, public opinion, political communication, emotion

3

Terrorism shattered America’s sense of invulnerability and unparalleled might on a sunny

September morning. Almost overnight, the American landscape went from one of prosperity, safety,

and power to one of threat, fear, and uncertainty. Threat and fear are not simply psychological

phenomena - they are politically consequential for how elites and the mass media communicate to

the public and, ultimately, for opinion formation. In times of crisis, citizens turn to political leaders

and the media to make sense of new and frightening events. The contours of the information

environment in turn influence how people prefer the government to react to threat. In this paper, I

argue that the features of the media after 9/11, particularly the media’s emphasis on threatening

information and evocative imagery, increased the public’s probability of supporting the hawkish

policies advocated by political leaders, principally the president. To the extent that the media

reinforces feelings of threat and political elites advocate a dominant set of policies, the public is apt

to support those policies.

While terrorism is inherently dramatic and threatening (Gans 1979), in a competitive media

environment, journalists and editors have incentives to use emotionally powerful visuals and

storylines to gain and maintain ever-shrinking news audiences, and these elements of news coverage

can strengthen the public’s sense of threat. Using an analysis of the National Election Studies (NES)

2000-2002-2004 panel and a controlled, randomized media experiment, I demonstrate that citizens

form significantly different foreign policy views when the information environment emphasizes

emotion than when it is free of emotion, even when factual information is identical. Citizens

concerned about terrorism are more likely to adopt the hawkish foreign policy views communicated

in threatening news stories when they are matched with fear-inducing cues than when they are not.

When people concerned about terrorism hear that a terrorist attack is likely and will bring fire and

destruction worse than 9/11, they adopt hawkish policies. When these people actually see the fire,

they react even more strongly.

4

Unlike other models of public opinion (Baum 2003; Baum and Groeling 2010; Berinsky

2009; Zaller 1992), my Threat Model contends that the mass media influence foreign policy opinion

through not only providing information, that is, communicating a set of policy options from

political leaders, but also through covering threatening issues such as terrorism in an evocative way.

In other models, the media function mostly as a conduit for elite cues; the distribution of elites

supporting each position tells citizens how to align their own views to policies (Berinsky 2009; Zaller

1992) or an interaction of the cues and the credibility of the cue shape opinion (Baum and Groeling

2010). My model, however, argues that although the media influence citizens through providing

foreign policy cues, the media also affect attitudes in their own right through evoking emotion. This

theory takes information seriously, but it also explicitly considers the ways in which the mass media

can influence opinions through the tone and presentation of information.1

Theories of the media’s influence that focus on policy content but ignore the often

threatening nature of news coverage are likely to underestimate the media’s effect on the public.

Since the media tend to cover foreign policy in threatening ways (Nacos, Bloch-Elkon, and Shapiro

2007), it is particularly important to account for how the emotional aspects of media coverage may

increase persuasion and shape the public’s policy preferences. When the news media emphasize

threat with emotional cues, those cues influence American foreign policy attitudes.

Threat and the anxiety that accompanies feelings of risk increase support for strong political

leaders as well as support for punitive policies that tend to escalate conflict (Gordon and Arian 2001;

Landau et al 2004). Fear leads individuals to search for information, and leads to less reliance on

long-standing political predispositions such as partisanship, increasing the potential for persuasion

(Brader 2006; Marcus et al 2000). If political leaders evoke fear in the public through the media

without being challenged by political opponents, citizens’ attitudes are likely to be influenced by the

foreign policy choices presented by the most prominent political leader, most likely, the president.

5

In considering how psychological mechanisms as well as the information environment

influence foreign policy attitudes, my Threat Model differs from previous work on foreign policy

attitudes and provides a unique contribution to the study of public opinion. Huddy and colleagues’

(2002; 2003; 2005) work on support for counterterrorism policies argues effectively that perceptions

of national threat increase support for hawkish policy. However, in my experiment described below,

the design pushes these findings further by demonstrating that emotion rather than information

leads public opinion in a more hawkish direction in a way that Huddy et al cannot since the authors

did not manipulate threatening news coverage experimentally. Merolla and Zechmeister (2009)

demonstrate that increased support for militant policies in times of threat is not limited to the

American political context. They use media stories to experimentally induce threat in both Mexican

and American samples, but they do not manipulate the emotional content independently of the

threatening information as my experiment does. As such, I am able to show that emotional cues

rather than threatening information alone influence attitudes.

The rest of this section outlines the components of my Threat Model: 1. the psychology of

threat, 2. how the political environment influences attitudes, and 3. the information environment’s

role in spreading and reflecting threat.

Fear itself – The psychological underpinnings of foreign policy attitudes

In my Threat Model, individuals receive information about foreign policy primarily through

the mass media, which disproportionately focus on threat and cover terrorism in sensationalistic

ways (Iyengar 1991; Nacos et al 2007). Additionally, in threatening times, the political spectrum

tends to conform to a “one-message” environment where opposition to the president’s foreign

policy lessens and, in turn, the media reflects one particular message about the appropriate foreign

policy approach (Zaller 1992). The War on Terror news frame that emerged after 9/11 argued, in

6

line with the president’s positions, that a military solution was necessary to counteract terrorism and

this frame received little serious opposition from Democratic leaders or the press (Entman 2003).

Individual perceptions of threat lead citizens to support policies that they believe will

neutralize the source of threat and protect them (Gordon and Arian 2001). The presence of threat is

associated with in-group solidarity (Turner et al 1984), ethnocentrism (Feldman and Stenner 1997),

an increased reliance on enemy images (Hermann 1986), and decreased support for civil liberties

(Davis and Silver 2004). In foreign policy, the threat of terrorism is associated with increased

support for retaliatory action by the government (Huddy et al 2002), greater support for overseas

involvement (Merolla and Zechmeister 2009), and support for George W. Bush (Huddy et al 2003;

Landau et al 2004). In sum, threat is related to a range of shifts toward more punitive preferences

and behavior. Therefore, I expect that individuals who perceive more threat from terrorism will support more

hawkish foreign policy. Conversely, heightened threat will decrease support for dovish foreign policy.

Media and foreign policy attitudes

The news media play a key role in shaping public opinion about the fundamental issues

facing the nation. Particularly in foreign affairs, where complicated issues as well as low knowledge

challenge the public’s ability to form attitudes, the media can affect opinion (Baum 2003; Berinsky

2009). Yet, until recently, the media’s role in forming foreign policy preferences was mostly ignored

in the literature on how people form attitudes about war and conflict (Hurwitz and Peffley 1987;

Mueller 1973). Terrorism can influence the public by having a venue via the mainstream media

(Norris, Kern, and Just 2003), and also through the way that terrorism stories are presented to the

public. Terrorism is newsworthy given its dramatic nature and receives inordinate coverage on the

national news. Even before the recent focus on terrorism, terrorism news predominated over other

types of potentially threatening news. Iyengar (1991) found that ABC, NBC, and CBS broadcast

more stories on hijackings in the 1980s than on poverty, unemployment, race, and crime combined.

7

Terrorism receives a considerable amount of coverage, and the coverage itself tends toward

sensationalism (Nacos et al 2007).

George W. Bush’s framing of the 9/11 attacks as the commencement of a global war against

terrorism was not significantly challenged by Democratic leaders or the press for some time after

9/11 (Entman 2003). For example, in New York Times coverage from 2001 to 2005, more than 70

percent of stories that utilized a “terrorism” frame took a pro-military engagement/hawkish tone

(Boydstun and Glazier 2008). Although hawkish frames generally dominated news coverage of the

War on Terror, the coverage did not have a complete lack of opposition. Guardino and Hayes

(forthcoming) found that the national network news featured sources opposed to the Iraq war, but

those opposition voices came mostly from overseas whereas the pro-military arguments came

mostly from “official” sources. So while dovish viewpoints were available to the public, they came

from less credible sources, suggesting a legitimization of hawkish views. Additionally, the imagery in

terrorism stories supported the foreign policy recommendations favored by the president (Entman

2003). Griffin (2004) found that photographs in Time, Newsweek, and US News after 9/11 depicted

the events in ways congruent with the president’s War on Terrorism frame.

The role of visual imagery in how the media affects attitudes

In the Threat Theory, the media act not only to provide the public with cues about elites’

foreign policy positions but also to impart emotional emphasis to foreign policy and terrorism

stories, primarily through the images accompanying news stories. Visual imagery may powerfully

arouse emotions and influence attitudes independently of news story messages (Brader 2006). Visual

information, particularly negative imagery, captures viewers’ attention and interest, is easier to

understand than verbal information, and makes retrieving information easier (Neuman, Just, and

Crigler 1992; Newhagen and Reeves 1992). Negative visuals like those in terrorism stories are likely

to induce emotion since they evoke a fear of death (Landau et al 2004) and remind viewers of their

8

traumatic emotional experience on 9/11. Surveys of Americans immediately after 9/11 showed a

strong connection between media exposure to the attacks and post-traumatic stress disorder

(Schuster et al 2001). The more television coverage people watched on 9/11 and afterward, the

more severe their stress responses, especially among people who saw vivid images of the attacks.

How threat and media interact

The interaction of threat and media on foreign policy opinion is rarely explored in the media

and public opinion literatures. Notable exceptions are studies by Huddy et al (2003), Hermann

(1986), and Merolla and Zechmeister (2009). Collectively, these studies find that threatening

information and images increase support for hawkish foreign policy while reassuring information

causes liberalization on defense issues. Yet none of this work asks whether, at a given level of threat,

different media messages with varying levels of emotion influence attitudes differently.

The perception of threat after 9/11 was compounded by media coverage that portrayed

“vivid and unceasing depictions of death and destruction” (Landau et al 2004), meaning that the

effect of threat may be conditional on media consumption. Druckman and McDermott (2008)

found that distress increases framing effects, suggesting that individuals concerned about terrorism

may be more open to persuasion and the messages offered about foreign policy through the mass

media. Thus, the influence of threat on opinion may depend on exposure to coverage of terrorist

acts or threats. In other words, threat should have a larger influence on attitudes for those people

exposed to fear-inducing media stories. If this is true, then we should expect the interaction of

perceived threat and exposure to emotionally threatening media content to predict hawkish opinion.

That is, increased media consumption in combination with heightened threat will increase support for hawkish foreign

policy. However, this relationship exists primarily when media content provides emotional coverage of threat.

9

Testing the Threat Theory 2002

From my Threat Theory, I expect that individuals who perceive high threat from terrorism

will support the most hawkish policies. In addition, I expect that individuals exposed to frightening

media coverage of foreign policy will be the most likely to support hawkish policy.

The Threat Theory implies that one of the major ways that the media can influence attitudes

is by enhancing feelings of fear and vulnerability. I use television versus newspaper consumption as

distinct measures of more emotional and less emotional media exposure to test the implication that

emotion, not just information, moves foreign policy attitudes. I utilize the 2000-2002-2004 NES

panel to model the effect of threat, television news, and newspaper consumption on an additive

index of foreign policy attitudes from 2002. The panel allows me to look at the same people before

and after 9/11 and to demonstrate that the type of media exposure significantly affected policy

attitudes. The policy index contains six items weighted equally: approval of going to war in Iraq,

approval of Bush’s handling of terrorism, spending on foreign aid (reverse-coded), defense, border

security, and homeland security (Cronbach’s α = .62).2 The index represents the range of policies

that the United States could use to fight terrorism – from diplomacy and aid (dovish) to military

action (hawkish). The spending questions asked respondents whether the federal budget should be

increased, decreased, or remain the same. The index ranges from -1 to 1 and is constructed so that

higher values signify more hawkish attitudes. While some policies, such as homeland security

spending, were relatively uncontroversial, it should be noted that in 2002, support for policies such

as war in Iraq represented the far end of the hawkish spectrum.

I model the effect of threat perception, television news consumption, newspaper reading,

and the interactions of threat and television and threat and newspaper, controlling for ideology and

partisanship, in two ways – using OLS and two-staged least squares (2SLS). Since my major interest

is modeling the influence of threat and media exposure on attitudes, it is important to take into

10

account the potentially cyclical relationship between threat and media exposure. Media exposure is

potentially both a cause and a result of threat perception. People who feel threatened by terrorism

might watch more TV news to cope with their anxiety about terrorism, but that information might

increase threat perception, leading to more TV news watching, in a mutually reinforcing cycle.

Ignoring endogeneity may lead to biased coefficients and high variance estimates in linear

regression, leading to incorrect inferences (Achen 1986). Given these potential drawbacks, I model

the relationship between threat and foreign policy attitudes using 2SLS as well as OLS. In the OLS

models, ideology and partisanship are measured in 2000 while television watching, newspaper

reading, and threat perception are measured in 2002.3 All independent variables are scaled to range

between 0 and 1 with higher values indicating more newspaper reading, television watching, and

threat perception. Higher values on the ideology and partisanship measures indicate more

conservative and Republican views. In the 2SLS models, I instrument television watching and

newspaper reading and their interactions with measures from the 2000 wave and include the 2000

ideology, partisanship, and 2002 threat measures.4

If the media influence attitudes through providing threatening information, then I expect

newspaper reading to affect attitudes; however, if the media influence attitudes through heightening

the effect of emotion rather than simply providing information, then television news exposure

should have a larger effect than newspaper reading. News consumption is measured by two variables

that capture media exposure: the average number of days in the past week the respondent watched

national and local television news and the number of days in the last week respondents read a daily

newspaper. Newspaper reading remained steady between the 2000 and 2002 panel waves while

television watching increased significantly. On average, in 2000, respondents read a newspaper 3.44

days a week, with 25 percent of respondents not reading a newspaper at all. In 2002, 21 percent of

respondents read no newspaper in the past week and the average respondent read a newspaper 3.67

11

days per week. In comparison, television news consumption increased from 3.12 days in 2000 to

3.75 days in 2002 (p<.01).5

Almost a year after the terrorist attacks, American citizens were still quite concerned about

future terrorism. I operationalized threat perception as respondents’ belief about the likelihood of

another terrorist attack, on a 4-point scale ranging from “very unlikely” to “very likely.” Forty-eight

percent of the 1,100 respondents in 2002 thought that another major terrorist attack was likely in the

next year, while another 19 percent believed that a terrorist attack was very likely, leaving one-third of

respondents less concerned about terrorism. Only 8 percent of respondents answered that terrorism

was very unlikely in the next year. This level of concern over foreign affairs was high in comparison

to respondents’ level of worry in 2000. There were no questions about terrorism in 2000, but the

NES did ask respondents in 2000 and 2002 how worried they were over the prospect of both

conventional and nuclear war. Concern over both types of war increased substantially over time.6

One of the main findings of these models is that television news significantly increased

support for hawkish policies whereas newspaper reading did not, suggesting that emotional coverage

mattered more than information on its own. However, television news watching did not significantly

affect attitudes in the absence of threat perception. The left side of Table 1 displays the results of the

OLS models. Respondents concerned about future terrorism reacted to threatening news by

increasing hawkishness while the one-third of respondents unconcerned by terrorism were unmoved

by television news. The effect of television exposure ranges from -.03, when a respondent believed

that terrorism was very unlikely in the next year, to .18 at the highest level of threat. Figure 1

presents the expected hawkishness scores for a respondent at varying levels of threat and television

watching, holding partisanship, ideology, and newspaper reading at their means. The figure

demonstrates that: 1. as respondents believed that terrorism was more likely, they preferred more

hawkish types of foreign policy and 2. television watching increased hawkishness among those

12

concerned about terrorism. The lower line displays the effect of television watching on attitudes for

respondents at the lowest level of threat; the flatness of the line demonstrates that citizens

unconcerned about a terrorist attack were unaffected by television news exposure. In contrast,

respondents at the highest level of threat were more responsive to media messages, as demonstrated

by the steeper top line of Figure 1. As they were exposed to more television news, high threat

respondents increased their hawkishness from .55 to .75, moving toward the top of the hawkishness

scale. This increase in support for hawkish policies had a substantively important effect on future

political preferences. For a moderate, Independent respondent, increasing hawkishness from .55 to

.75 in 2002 translated into an increased probability of supporting George W. Bush in the 2004

election by 14 percentage points.7

Unlike television exposure, newspaper reading had no significant effect on foreign policy

attitudes either on its own or when controlling for television exposure. Newspaper readers are more

able to tailor the information that they are exposed to – individuals not worried about terrorism may

simply have chosen to avoid terrorism stories or avoid the most threatening aspects of stories and

therefore the threatening information did not influence their attitudes. Alternatively, low threat and

high threat respondents may simply watch different news altogether, which is not possible to

ascertain using this measure of media use. Yet the differences between newspaper reading and

television exposure may also be due to the more emotional nature of television news – television

stories allow individuals to visualize terrorism.

Table 1 also displays results from 2SLS models regressing foreign policy attitudes on

respondents threat perception and exogenous measures of media consumption in 2002. The right

side of Figure 1 shows the predicted values of foreign policy holding the newspaper reading,

ideology, and partisanship at their means. The dependent variable is the same index as above. These

2SLS models reveal the same basic patterns as the OLS models with one significant difference.

13

Consistent with the findings from the previous models, respondents high in threat prefer

significantly more hawkish policy than those lower in threat. Unlike the OLS models, the 2SLS

models show that newspaper reading may increase support for hawkish policies among high threat

citizens, although the effect of newspaper reading is significantly attenuated when accounting for

television watching, meaning that newspaper consumption does not have a direct effect on attitudes

– the direct effect is carried only by the more emotive television content. This suggests that the

threatening information captured in newspaper stories can influence foreign policy attitudes but that

emotive television coverage mattered more.

These models suggest that television news, even very threatening news that catches viewers’

attention, may influence viewers in a variety of ways – consumers’ threat perceptions influence how

they understand what they watch. High and low threat respondents take away different types of

information and emotion from television coverage of foreign policy. Overall, both the OLS and the

2SLS models provide support for the hypothesis that an increased sense of vulnerability about

terrorism led to more support for a variety of policies linked to the War on Terror – more defense

and border spending, higher support for the Iraq war, and less support for dovish policies such as

foreign aid. Additionally, the findings show that television watching rather than newspaper reading

moved foreign policy attitudes, suggesting that emotionally potent coverage of terrorism influenced

attitudes more significantly than information itself.

Respondents threatened by terrorism were open to persuasion, and in the mostly one-sided

political environment, these threatened individuals adopted the president’s position on foreign

policy. Support for hawkish policy was especially high among people watching the nightly television

news, news that ran hundreds of terrorism segments in the year after 9/11. Because the national

news reflected the president’s War on Terror framing (Entman 2003), respondents already

concerned about terrorism received messages to help connect their sense of threat to hawkish policy

14

options. Threat matters for political attitudes because the president and the media imbued the threat

of terrorism with political meaning that citizens used to shape their attitudes. Television news, with

its more emotional coverage of terrorism, mattered more for attitudes than did less evocative

newspaper coverage, a dynamic explored in the next section.

The Threat Experiment

Utilizing the 2006 Threat Experiment, this section shows that when news stories pair

threatening information with fear cues, threatened respondents are significantly more likely to

support militaristic foreign policy than respondents who only receive threatening information.

As shown in the previous section, threatened individuals exposed to the most television

news adopted the most hawkish attitudes. However, the survey data cannot disentangle exactly why

the television news exposure so effectively influenced attitudes - whether the threatening

information or the emotionally evocative presentation of the information increased support for

hawkish policy. Using an experiment, this section further tests the hypothesis that television news

stories increase support for hawkish policy through emotionally threatening cues.

To test the effect of threatening media content on policy attitudes, I designed and ran the

2006 Threat Experiment, a media experiment with a nationally representative sample of 1,229 adults

recruited through YouGov/Polimetrix. Because the experimental design required subjects to watch

a news segment, the experiment took place on-line. Experiments provide a way to determine the

causal impact of media exposure on attitudes in a way that secondary survey data cannot, making an

experiment a better methodological choice. Using survey data alone, it is not possible to determine

whether media cause feelings of threat, or whether pre-existing feelings of threat determine attention

to news and threatening information in particular. Additionally, media exposure self reports such as

those used in the NES analysis are prone to over-reporting because citizens have a difficult time

15

estimating their news exposure (Prior 2009). Lastly, even with a perfect measure of media exposure,

secondary data is unable to distinguish the effect of information from the effect of emotion.

Given these issues, an experiment provides the advantage of control over both exposure and

content. Respondents are randomly assigned to experimental conditions and exposed to the same

information within each experimental condition, meaning that the selection mechanism for exposure

is random rather than determined by subjects’ characteristics. The randomization enhances the

ability to make inferences about how the media affect attitudes.

Experimental treatments

The experimental treatments manipulated both the presence of threat (threatening v. non-

threatening verbal information) and the emotional cue (threatening visuals v. non-threatening

visuals) in real television stories about recent terrorist attacks. The experiment took place online

during fall 2006 with a representative sample of Americans. The sample is composed of 52 percent

female respondents, 80 percent white respondents, 23 percent with bachelors’ degrees or higher, and

an average age of 46 years old. Participants answered a variety of attitude questions and were

randomly assigned to watch one of three videos. In the control condition, participants watched an

irrelevant news story on India’s economy that provided neither verbal information on threat nor a

visual emotional cue to threat. Subjects in the two treatment conditions watched a story entitled

“Wave of Terror” that outlined recent terrorist attacks and suggested that a terrorist attack was

imminent. The experimental story suggested that, given the pattern of terrorist attacks in the past

year, the public should expect further attacks and featured terrorism experts warning the public. All

subjects in the treatment conditions received exactly the same verbal information about terrorism --

government authorities were troubled by the increase in smaller-scale terrorist attacks like the ones

in London and Madrid and the attacks might signal a new wave of terrorism.

16

Only the visual imagery differed between the two treatment conditions: the “scary visuals”

group viewed the terrorism story with evocative imagery of terrorism while the “neutral visuals”

group viewed the same story with neutral imagery. Both terrorism stories highlighted the threat from

terrorism, but the neutral visuals condition used a non-emotional tone and no visual imagery of

terror. In the scary visuals treatment, the terrorism news story was edited to enhance the threatening

nature of the visual imagery. This editing added video such as the burning World Trade Center and

bloodied victims of the 2005 London transit bombings. In the neutral visuals condition, subjects

watched the same story on terrorism with the visuals replaced by less violent imagery. For example,

in place of the London victims, the neutral condition included maps of the London subway system.

Table 2 summarizes the design of the experiment and the online appendix provides the text of the

experimental treatments and compares the types of images seen in each treatment condition.

The experimental design thus clearly teases apart the unique aspect of media coverage that

my model predicts will affect attitudes – threatening visual cues that convey emotional content –

from the unemotional information about the likelihood of a terrorist threat. To the extent that

experimental respondents receive a threatening message about terrorism paired with emotional

imagery, these respondents should be the most likely to support hawkish foreign policy. Across the

conditions, the message about the increased likelihood of threat is the same while the emotional

presentation differs, so any differences in opinion that occur can be attributed to differences in the

imagery/emotion rather than to differences in the threatening information or other dimensions such

as framing. If there are differences in opinion only between the control group and the treatment

groups but not between the neutral visuals condition and the scary visuals condition, then

threatening story content influences opinions but how the content is presented does not. If there are

differences between the neutral visuals condition and the scary visuals condition, then presentational

factors like the visual content affect opinion.

17

The Threat Experiment provides a hard test for pinning down the mechanism by which the

media matter. The “Wave of Terror” story that respondents in the treatment conditions watched

contained threatening information about potential terrorist attacks but did not contain overt policy

content. The story itself did not advocate hawkish policies or quote sources arguing for more

homeland security funding or sending more troops abroad. In addition, the experimental treatment

consisted of only one, 2 ½ minute story about terrorism abroad in the midst of more than five years

and literally thousands of other news stories about terrorism. Even though respondents were asked

their reactions and attitudes directly after the story, the treatment was relatively mild compared to

the vast amount of relevant news available to Americans.

To confirm that the scary visuals condition evoked respondents’ emotions as anticipated,

this study used two manipulation checks to measure emotional reactions to the news stories. Prior to

the Threat Experiment, all images were pre-tested by a separate sample of 79 students to confirm

that the terrorism images provoked anxiety. Respondents viewed a series of both neutral and

evocative photos in random order and answered how afraid, hopeful, and angry each photo made

them, on a scale from 1 “not at all” to 10 “very”. On average, the neutral images evoked significantly

less fear than the terrorism images. For the seven neutral images the mean fear score was 1.60 while

the average fear score for the terrorism pictures was 4.84 (p<.01). Respondents who saw a neutral

image of a double-decker London bus (Mbus = 1.70) were significantly less fearful than those who

saw the iconic image of the London bus attacked on July 7, 2005 (Mexplosion= 4.91, p<.01).

As another measure, after watching the news story, each subject received a set of questions

designed to tap into particular negative emotions; respondents rated how fearful (worried, fearful,

anxious), sad (sad, depressed, grief stricken), and angry (angry, mad, and furious), they felt.

Respondents rated their emotional reactions on a 9 point scale from “did not feel the emotion” at

the low end to “felt the emotion very strongly” at the high end of the scale, which I rescaled to vary

18

between 0 and 1. All of these negative emotions loaded strongly onto a single dimension in a factor

analysis (Eigenvalue =2.00, Cronbach’s α = .87). Respondents in the scary visuals condition felt

significantly more negative emotions (Msv = .43) than subjects in the control condition (Mc = .27,

p<.01). Additionally, respondents concerned about terrorism prior to watching the terrorism story

reported feeling significantly more negative emotions in the scary visuals condition than either of the

other two conditions (Mc= .28, Mnv= .48. Msv = .54, p<.02 both differences), suggesting that the

visuals worked as intended.

Threat and emotion’s effects on foreign policy attitudes

To establish that emotionally evocative media coverage increases support for hawkish

policies, I measure respondents’ foreign policy attitudes by experimental condition using a set of five

dependent variables. All dependent variables were measured after exposure to the experimental

treatments. The militarism measure asked respondents whether they thought that the best way to

ensure peace was through military strength or diplomacy. The dichotomous measure is scored from

0 to 1, with 1 representing the “military” position. The spending index is a four item index; each

questions within the index asked respondents, “Should the federal government should increase,

decrease, or keep funding levels the same?” on foreign aid, defense, border security, and homeland

security. Each question weighed equally in the index (Cronbach’s α= .70), which was scaled from -1

(spend less) to 1 (spend more), and foreign aid was reverse coded. These measures are identical in

wording to the measures from the NES utilized in the previous section.

To test whether threat has broad consequences on foreign policy, respondents also answered

several questions on their attitudes about specific military engagements in Sudan and Iraq. In 2002,

Americans who perceived a high threat of terrorism were supportive of using the United States

military to destroy terrorist groups in Sudan (Pew 2002). Respondent’s reaction to the Darfur region

of Sudan provides a way to measure whether, given limited resources, citizens want the United

19

States to spend funds for military or dovish means. The Sudan measure is a combination of two

questions that asked respondents about hypothetical actions that the U.S. government could take–

military action to destroy terrorist groups and sending humanitarian foreign aid to Darfur.

Respondents gave their preferences toward military action and humanitarian relief separately using a

5-point scale that ranged from strongly oppose to strongly favor, with “neither favor nor oppose” as

the middle position. I created the Sudan measure by categorizing respondents who wanted to send

aid but not troops as doves with a score of 0 and respondents who wanted to send troops but not

aid as hawks with a score of 1.8 The Iraq measure asked respondents to evaluate retrospectively

whether they believed that the war in Iraq was worth the cost, and is scored on a 5 point scale from

“strongly disagree” at 0 to 1 “strongly agree”. Lastly, the Government approval measure is a three item

index (Cronbach’s α = .70) that asked respondents to evaluate how well they believed the

government was doing in reducing the threat of terrorism, how much they approved of the

president’s handling of terrorism, and whether they believed that the ability of terrorists to carry out

terrorist attacks was greater, less, or the same than before 9/11. Higher scores indicated higher

approval of the Bush administration and are categorized as being more hawkish.

To test the hypothesis that high threat respondents are especially likely to support hawkish

policy when exposed to threatening news paired with fear cues, Figure 2 shows the differences in

means between treatments and the 90 percent confidence interval around the differences.9 This

figure compares the treatment effects on all foreign policy measures for all respondents. The black

circles represent the differences in attitudes between respondents in the neutral visuals condition

and the control condition, and the white open diamonds represent the differences in opinion

between the scary visuals condition and the neutral visuals condition. Movements in a hawkish

direction are shown to the right of the vertical dotted line while movements toward the dovish end

appear to the left of the line. Respondents are divided by their pre-test level of threat, measured as

20

how likely the respondent believed a terrorist attack was in the next year. The 553 low threat

respondents believed that terrorism was “(very) unlikely” while the 674 high threat respondents

answered that terrorism was “(very) likely”.

Consistent with the findings from the NES, low threat subjects are unresponsive to both the

message and the presentation of television news, with the exception of the decision to send troops

versus aid to Sudan. While subjects unconcerned about terrorism are no more likely to support

hawkish policy in the treatment conditions than in the control condition across 4 of the 5 dependent

variables, low threat subjects in the scary visuals condition actually adopt a more dovish position

than respondents in the neutral visuals condition. While 26 percent of low threat respondents in the

neutral visuals condition preferred sending troops to fight terrorism in Sudan rather than send

humanitarian aid, only 13 percent of those in the scary visuals condition did. This move toward

dovishness suggests that the imagery associated with terrorism may not automatically be associated

with hawkish policy outcomes. Terrorism in Sudan and the Darfur situation are not salient issues

closely associated with the War on Terror, meaning that the threatening imagery may easily sway

respondent to support a variety of positions, including dovish ones. Among those individuals

unconcerned about terrorism, neither the threatening information contained in the neutral visuals

story nor the evocative imagery in the scary visuals story affected foreign policy attitudes for the

majority of measures.

Citizens often choose between not watching a terrorism story at all and seeing an evocative

story highlighting death and destruction. The experiment allows a test of the effect of that choice on

foreign policy attitudes by comparing the foreign policy views of high threat respondents in the

scary visuals condition to those in the control condition. Across all five foreign policy measures,

high threat respondents in the scary visuals condition are about 6 percent more supportive of

hawkish policy than high threat subjects in the control condition.10 To compare the magnitude of

21

these effects, the largest effects of the scary visuals conditions are in the case of overall militarism

(10 percentage points) and in willingness to bomb Sudan (9 percentage points). These effects are

slightly less than the effect of being a Republican (14 percentage points) and about a third of size of

voting for Bush in 2004 (34 percentage points) on respondents’ hawkishness in the 2004 NES. The

experiment also allows for a tougher test of the hypothesis that the emotional nature of terrorism

news increases support for militaristic policies.

If threatening information about terrorism alone affected attitudes, then support for hawkish

policies should be equal in the neutral visuals and scary visuals conditions. Figure 2 demonstrates

though, that the scary visuals condition increased support for hawkish policies over the neutral

visuals condition among high threat subjects. In contrast to low threat respondents unmoved by

either information or emotion, high threat respondents react to the threatening information and

emotional imagery in the scary visual condition by supporting increasingly more militant foreign

policy. Compared to respondents who watched the Wave of Terrorism story without the emotional

imagery, respondents in the scary visuals condition are more supportive of military solutions to

international problems by .07 on the 0 to1 scale (p<.08), prefer more spending on areas such as

defense and border security by .09 (p<.02), rate the government’s handling of terrorism more

positively by .06 (p<.07), and a larger percentage of respondents prefer to send troops to Sudan

rather than foreign aid by .11 (p<.06). High threat subjects are more supportive of the Iraq war than

low threat subjects, but the experimental treatments failed to increase support for the war among

those high in threat. Although not all of the differences between the neutral and scary visuals

conditions are significant at the conventional p<.05 level, the overall pattern of findings

demonstrates that threatening media content can increase hawkishness but that the most powerful

effects from television news comes from stories that utilize emotional presentations. These findings

demonstrate that watching one news story about potential terrorist attacks that included emotion-

22

inducing imagery increased support for a variety of militaristic policies while the less emotional

neutral visuals condition did not

According to the hypotheses, I expect that terrorism stories significantly affect attitudes

because they evoke negative emotions like anxiety that make citizens more open to political

persuasion. To more directly test whether the scary visuals condition affects attitudes through

evoking respondents’ emotions, Figure 3 compares the mean levels of foreign policy attitudes

among respondents who did not react emotionally to the scary visuals condition (low emotion) and

those with stronger emotional reactions (high emotion) and 90 percent confidence intervals around

the means.11 As Figure 3 demonstrates, the scary visuals condition affected foreign policy attitudes

through evoking emotion. Subjects who reacted emotionally preferred more militaristic policy

overall (Mdifference = .41, t=8.78, p<.01), supported more defense and border security spending

(Mdifference = .36, t=7.02, p<.01), evaluated the Bush administration’s handling of terrorism more

positively (Mdifference = .18, t=4.51, p<.01), and were more supportive of the use of the military

overseas in Sudan (Mdifference = .22, t=6.31, p<.01), than respondents unemotional after watching the

scary visuals condition. The size of the emotion effect is roughly equivalent to the difference

between Democrats’ and Independents’ foreign policy views in this sample. The effect of emotion is

to substantively increase support for a range of more militaristic policies that respondents do not

prefer in the absence of emotion.

The experiment demonstrates that more than 5 years after the 9/11 attacks, exposure to a

single terrorism story led high threat respondents to prefer more hawkish types of foreign policy

when the story presented both a threatening message and emotional presentation. What makes this

finding especially striking is that, in the years since 9/11, American citizens saw frequent news

stories about terrorism and had stored knowledge and beliefs about counter-terrorism policy.

Equally remarkable is the fact that, even though the public is both interested and knowledgeable

23

about terrorism, watching one more frightening terrorism story significantly influenced attitudes of

respondents concerned about terrorism more than 5 years after 9/11.

Conclusion

In times of crisis, citizens turn to the government and mass media for answers, comfort, and

protection. In the hours after the 9/11 terrorist attacks, millions of Americans turned on televisions

in their homes, schools, and workplaces to make sense of events and to understand how they should

react. But television heightened threat instead of providing comfort because it focused on

threatening, emotionally laden images of terror. Television watchers saw the Twin Towers fall in

New York City, saw crowds of office workers running away from the disaster as fire fighters and

police ran toward the collapsed buildings, and saw the desperate faces of bystanders. Bearing witness

in this way led many Americans to feel profound emotions – anger at the hijackers, sadness over the

loss of life, and fear of future attacks. The emotional cues in the media combined with the

perception that another attack was likely shaped the types of policies that the public demanded to

address terrorism in the aftermath of the attacks.

In this article, I argue that evocative, emotionally powerful images of terrorism cued the

public into supporting hawkish foreign policy in the years after 9/11. Using an experiment that

tightly controlled the type of terrorism stories respondents saw, this article demonstrates that

emotionally charged news influences attitudes significantly more than information on its own.

However, not all citizens reacted to the information environment in the same way. Those individuals

already concerned about future terrorism were especially supportive of hawkish policy when

exposed to news coverage, particularly news coverage that was emotional in nature. Americans

convinced that terrorism was likely in the future were responsive to the message and emotional

content of terrorism news stories. The mass media covered the War on Terror heavily from the

9/11 attacks onward, and the findings from this paper illustrate that the amount and type of

24

information that citizens received from the media profoundly affected policy attitudes, conditional

on individuals’ level of concern over terrorism.

While the focus of this particular study is terrorism, the results hold implications for other

policy domains. In policy areas where there are resonant images reflected in mass media coverage

that can induce negative emotions, these findings suggest that elites have an opportunity to persuade

citizens to support preferred policies, particularly punitive policies. In particular, past research

demonstrates that perceptions of threat shape attitudes toward crime, immigration, and public health

(Brader, Valentino, and Suhay 2008; Iyengar 1991; Witte and Allen 2000). What the Threat Theory

can add to this research is an understanding of how mass media coverage of these issues may

increase support for punitive/hawkish policies - that is, through evoking emotion.

Issues of war and peace are fundamental to democracy, and this paper reveals conditions

under which the mass public will support the use of force. This paper fills a hole in the literature on

public attitudes in times of crisis by demonstrating that public support for hawkish policy is not the

only possible outcome after terrorist attacks, but under the conditions where the public feels

threatened and the mass media covers terrorism in a sensationalistic manner, support for hawkish

policy will increase. These findings also suggest that the role of the media is broader than simply

providing a conduit for elites to speak to the public; the media influences the public through its own

means as well. This article outlines how the media play a role in convincing some members of the

public to support different policies than they otherwise might. While respondents high in threat

adopted significantly more hawkish attitudes after watching the scary visuals condition than the

control condition, this article should also provided some relief to those concerned about the

unchecked power of fear to adulterate the public’s attitudes on foreign policy and war. Emotion did

alter some respondents’ attitudes, but the largest effect was on individuals open to the threatening

messages in the treatment conditions, suggesting that people can rely on certain mechanisms, such

25

as diminishing the threat of terrorism, to cope with threatening news. Yet, if the media can activate

citizens’ sense of impending danger, my findings show that when political leaders trumpet the right

threatening message and the mass media use the right images, the public may consent to war, even

though a citizenry with less emotion may have chosen otherwise.

26

Table 1: The effect of threat and media consumption on foreign policy attitudes 2002

Foreign policy index 2002

OLS 2SLS

Television

model Newspaper

model All

media Television

model Newspaper

model All

media

PID 2000 0.22 0.20 0.22 0.18 0.22 0.18 (0.030) (0.030) (0.030) (0.035) (0.032) (0.038) Ideology 2000 0.14 0.15 0.14 0.16 0.14 0.16 (0.027) (0.027) (0.027) (0.028) (0.028) (0.029) Threat of Terrorism 2002 -0.01 0.14 0.01 0.10 0.12 0.10 (0.053) (0.046) (0.061) (0.030) (0.030) (0.030) TV watching 2002 -0.03 -0.03 -0.50 -0.53 (0.059) (0.060) (0.184) (0.210) TV * Threat 0.22 0.22 0.83 0.89 (0.083) (0.084) (0.239) (0.292) Newspaper reading 2002 0.03 0.01 -0.36 -0.06 (0.048) (0.048) (0.16) (0.20) Newspaper * Threat -0.02 -0.03 0.60 -0.13 (0.070) (0.069) (0.25) (0.330) Constant 0.37 0.33 0.36 0.34 0.34 0.35 (0.038) (0.035) (0.044) (0.040) (0.030) (0.041)

Observations 1158 1163 1157 1047 1058 1046 R2 0.15 0.13 0.15 0.16 0.15 0.16

Source: NES 2000-2004 panel. Model Specification: OLS and 2SLS. Coefficients in bold are significant at p < .05. Standard errors are in parentheses. Foreign policy index is a 6-item additive index – spending on foreign aid, defense, border security, and homeland security/ war on terror, support for the war in Iraq, approval of the president’s handling of terrorism. The indices range from -1 to 1, higher values are more hawkish. Partisanship and ideology are measured in 2000 and range from 0 to 1, with higher values indicating Republican and conservative. Television and newspaper consumption are measured in 2002 as the number of days a week respondents watched TV news/read a daily newspaper, recoded to range between 0 and 1 in the OLS models. Television and newspaper consumption in the 2SLS models are instrumented from TV watching and newspaper reading in 2000 and their interactions are instrumented as well. (See footnote 4 for details) Threat perception is respondent’s belief about the likelihood of another terrorist attack ranging from “terrorism not very likely” (0) to “terrorism very like” (1).

27

Table 2: The Threat Experiment Design

Story watched Story content Visual content

Control condition

Indian economy

Neutral “This is India booming.”

Neutral

Neutral visuals condition

Wave of Terror

Threatening “Is this the beginning of a wave of attacks?”

Neutral

Scary visuals condition

Wave of Terror

Threatening “Is this the beginning of a wave of attacks?”

Threatening

28

Figure 1: Effect of threat and media use on foreign policy attitudes 2002

High threat

Low threat

.2.3

.4.5

.6.7

.8

Haw

kis

hness

No TV Mean TV Every day

TV watching per week

OLS model

High threat

Low threat

.2.3

.4.5

.6.7

.8

Haw

kis

hness

No TV Mean TV Every day

TV watching per week

IV model

Source: 2000-2004 NES

Predicted values and 95% CI

Foreign Policy Attitudes 2002

29

Figure 2: Scary visuals condition increase hawkishness among high threat respondents

Hawkishness

-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4

Iraq Worth Cost

Approval

of Handling

of Terrorism

Sudan

Foreign Policy

Spending

Militarism

Diff(Neutral-Control)

Diff(Scary-Neutral)

Foreign Policy Attitudes by Condition

(Low threat subjects)

Hawkishness

-0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5

Iraq Worth Cost

Approval

of Handling

of Terrorism

Sudan

Foreign Policy

Spending

Militarism

Diff(Neutral-Control)Diff(Scary-Neutral)

Foreign Policy Attitudes by Condition

(High threat subjects)

30

Figure 3: Emotion mediates the effect of the scary visuals condition

31

References

Achen, Christopher. 1986. The Statistical Analysis of Quasi-experiments. Berkley,CA: University of

California Press.

Bartels, Larry. 1991. Instrument and “Quasi-Instrumental” Variables. American Journal of Political

Science. 35(3): 777-800.

Baum, Matthew. 2003. Soft News Goes to War: Public Opinion and American Foreign Policy in the New Media

Age. Princeton, NJ: Princeton University Press.

Baum, Matthew and Timothy Groeling. 2010. War Stories: The Causes and Consequences of Public Views of

War. Princeton: Princeton University Press

Berinsky, Adam 2009. America at War: Public Opinion During Wartime, From WWII to Iraq. Chicago:

University of Chicago Press.

Boydstun, Amber and Rebecca Glazier. 2008. From Spreading Freedom to WMDs: Framing

Dynamics and the War on Terror. Presented at the Midwest Political Science Association.

Chicago.

Brader, Ted. 2006. Campaigning for Hearts and Minds: How Emotional Appeals in Political Ads Work.

Chicago: The University of Chicago Press.

Brader, Ted., Nicholas Valentino, and Elizabeth Suhay. 2008. What Triggers Public Opposition to

Immigration? Anxiety, Group Cues, and the Immigration Threat. American Journal of Political

Science. 54(4).

Davis, Darren and Brian Silver. 2004. Civil Liberties v. Security: Public Opinion in the Context of

the Terrorist Attacks on America. American Journal of Political Science. 48(1): 28-47.

Druckman, James and Rose McDermott. 2008. Emotion and the Framing of Risky Choice. Political

Behavior. http://www.springerlink.com/content/104963/

32

Entman, Robert 2003. Cascading Activation: Contesting the White House’s Frame After 9/11.

Political Communication. 20: 415-432.

Feldman, Stanley and Karen Stenner. 1997. Perceived Threat and Authoritarianism. Political

Psychology. 18: 741-770.

Gadarian, Shana Kushner. 2008. The Politics of Threat: Terrorism, Media, and Foreign Policy

Opinion. Doctoral Dissertation. Princeton University.

Gans, Herbert. 1979. Deciding What’s News: A Study of CBS Evening News, NBC Nightly News,

Newsweek, and Time. New York: Pantheon.

Gordon, Carol and Asher Arian. 2001. Threat and Decision Making. Journal of Conflict Resolution.

45(2): 196-215.

Griffin, Michael. 2004. Picturing America’s “War on Terrorism” in Afghanistan and Iraq.

Journalism. 5(4): 381-402.

Guardino, Matt and Danny Hayes. forthcoming. Whose Views Made the News? Media Coverage

and the March to War in Iraq. Political Communication.

Hermann, Richard 1986. The Power of Perceptions in Foreign Policy Decision Making: Do Views

of the Soviet Union Determine the Policy Choices of American Leaders? American Journal

of Political Science 30(4): 841-875.

Huddy, Leonie, Stanely Feldman, Tereza Capelos, and Colin Provist. 2002. The Consequences of

Terrorism: Disentangling the Effects of Personal and National Threat. Political Psychology.

23(3): 485-509.

Huddy, Leonie, Stanley Feldman, Gallya Lahav, and Charles Taber. 2003. “Fear and Terrorism:

Psychological Reactions to 9/11,” in Framing Terrorism: The News Media, the Government, and the

Public. P. Norris, M. Kern, M. Just, eds. New York: Routeledge.

Huddy, Leonie, Stanley Feldman, Charles Taber, and Gallya Lahav. 2005. The Politics of Threat:

33

Cognitive and Affective Reactions to 9/11. American Journal of Political Science. 49(3): 610-625.

Hurwitz, Jon and Mark Peffley. 1987. How Are Foreign Policy Attitudes Structured? A Hierarchical

Model. American Political Science Review. 81(4): 1099-1120.

Iyengar, Shanto. 1991. Is Anyone Responsible? How Television Frames Political Issues. Chicago:

University of Chicago Press.

Landau, Mark, Sheldon Solomon, Jeff Greenberg, Florette Cohen, Tom Pyszczynski, Jamie Arndt,

Claude Miller, Danielle Ogilvie, Alison Cook. 2004. Deliver Us from Evil: The Effects of

Mortality Salience and Reminders of 9/11 on Support for President George W. Bush.

Personality and Social Psychology Bulletin. 30(9): 1136-1150.

Marcus, George, W. Russell Neuman, Michael MacKuen. 2000. Affective Intelligence and Political

Judgment. Chicago: University of Chicago Press.

Merolla, Jennifer and Elizabeth. Zechmeister. 2009. Democracy at Risk: How Terrorist Threats Affect the

Public. Chicago: University of Chicago Press.

Mueller, Jon. 1973. War, Presidents, and Public Opinion. New York: Wiley.

Nacos, Brigette, Yaeli Bloch-Elkon, and Robert Shapiro. 2007. Post-9/11 Terrorism Threats, News

Coverage, and Public Perceptions in the United States. International Journal of Conflict and

Violence. 1(2): 105-126.

The National Election Studies (www.electionstudies.org). THE NATIONAL ELECTION

STUDY 2000-2002-2004 FULL PANEL FILE [dataset]. Ann Arbor, MI: University of

Michigan, Center for Political Studies [producer and distributor].

Neuman, W. Russell, Marion Just, and Ann Crigler. 1992. Common Knowledge. Chicago: University of

Chicago Press.

Newhagen, John and Byron Reeves. 1992. The Evening’s Bad News: Effects of Compelling

Negative Television Images on Memory. Journal of Communication, 42(2): 25-41.

34

Norris, Pippa., Montague Kern, and Marion Just, eds. 2003. Framing Terrorism: The News Media, the

Government, and the Public. New York: Routledge.

Pew Research Center for the People and the Press. 2002. January 2002. Americans Favor Force

in Iraq, Somalia, and Sudan: News Interest Index. January 22, 2004.

Prior, Markus. 2009. The Immensely Inflated News Audience: Assessing Bias in Self-Reported News

Exposure. Public Opinion Quarterly. 73 (1): 130-143.

Schuster, Mark, Marc Elliott, Annie Zhou, David Kanouse, Janina Morrison, Sandra Berry. 2001. A

National Survey of Stress Reactions after the September 11, 2001, Terrorist Attacks. New

England Journal of Medicine. 345(20): 1507-1512.

Turner, John, Michael Hogg, P. Turner, and P. Smith. 1984. Failure and Defeat as Determinants of

Group Cohesiveness. British Journal of Social Psychology. 23: 97-111.

Witte, Kim and Mike Allen. 2000. A Meta-analysis of Fear Appeals: Implications for Effective

Public Health Campaigns. Health Education and Behavior. 27(5): 608-632.

Zaller, John. 1992. The Nature and Origin of Public Opinion. New York: Cambridge University Press.

1 The Threat Theory’s emphasis on threat and emotion does not imply that I believe threat is the

predominant determinant of foreign policy attitudes or that predispositions such as partisanship and

ideology are irrelevant to opinion formation. While partisanship and ideology may be less helpful

cues in foreign policy than domestic policy, foreign policy opinion in the post-9/11 era is

significantly affected by partisanship, elite positions on foreign policy, and beliefs about information

credibility. The Threat Theory argues that terrorism stories moved the public toward hawkish policy

because the presidential administration effectively framed hawkish policy as the remedy for

terrorism and that that link between terrorism and policy is strongest when connected to

emotionally evocative images. Even as views on the War on Terror became more polarized by party,

35

emotionally charged news could influence attitudes by increasing support for the policies advocated

by one’s favored elite or for the dominant hawkish policy if no strong alternatives were offered by

elites outside the presidential administration. Emotion may be most powerful in a “one-message”

political environment, yet it does not preclude the possibility that other political actors can harness

emotions to argue for alternative policies or the possibility that the emotion may be so powerful as

to persuade citizens to permanently shift their attitudes. The moderating effects of partisanship on

foreign policy attitudes are explored further in Gadarian (2008).

2 The online appendix http://rwj.berkeley.edu/scholars/gadarian.php includes question wordings

and codings. This foreign policy index captures a broad foreign policy dimension of how

respondents want the government to handle international problems: through diplomacy or the

military. Using factor analysis with varimax rotation, the 6 items loaded strongly on one dimension

(Eigenvalue = 1.37). The 2002 index correlates with a 7-point diplomacy-militarism scale from the

2004 NES wave (r=.36, p<.01) and the same index measured in 2004 correlates with the 2004

diplomacy-militarism scale (r =57, p<.01).

3 Conservative and Republican identification both increased in the NES panel between the 2000 and

2002 survey waves as did support for hawkish foreign policy. To control for the possibility that

respondents updated their partisanship and ideology to remain consistent with their foreign policy

preferences, the models use respondents’ partisanship and ideology measured in the 2000 wave

rather than the 2002 measure.

4 The panel structure is useful for finding instruments for the media exposure variable in 2002

because respondents answered the same media questions in 2000. The television questions asked

respondents how many days in the last week they watched local and national television. In order to

get a “purged” measure of TV watching in 2002, I regressed the average of respondents’ national

36

and local news watching in 2002 on respondent’s average television watching in 2000, race, age,

education, political sophistication, partisanship, ideology, marital status, and whether they lived in

the West. I then use the predicted measure of TV watching in the second stage equations. Finding a

useful instrument for threat perception is much more difficult since the 2000 NES did not ask about

the threat of terrorist attacks. Threat perception may be closely related to individual personality

dispositions such as trait-anxiety, for which there are few good measures on the NES. The lack of

appropriate prior measures inhibits my ability to find appropriate instruments. While ignoring

endogeneity may wreak havoc on statistical estimations, using weak instruments may also seriously

bias statistical inferences (Bartels 1991). Because I doubt the appropriateness of any instrument for

threat perception, I include the original 2002 threat measure in all of the models in Table 1.

5 Media exposure self-reports may be problematic because individuals have trouble estimating their

own exposure (Prior 2009). Yet, these models are concerned with whether media consumption

increases or decreases rather than with the actual number of days respondents watch/read.

Additionally, the experiment later in the argument controls exposure rather than relying on self-

reports.

6 In 2000, 50 percent of respondents answered that they were somewhat or very concerned about

the prospect of conventional war, which increased to 84 percent by 2002 (χ2 = 31.8, p<.01).

Similarly, anxiety over nuclear war increased over time, in 2000, 43 percent of respondents were

either somewhat (35 percent) or very worried (8 percent), while in 2002, 75 percent of the same

respondents were somewhat (50 percent) or very worried (25 percent) (χ2 = 92.8, p<.01).

7 These calculations are based on a model that using the NES 2000-2004 panel to predict the

probability that a respondent would vote for George W. Bush in the 2004 election based on

respondent’s partisanship, ideology, gender, race (all measured in the 2000 wave) and 2002

37

hawkishness (measured with the foreign policy index). I then calculated a white, male respondent’s

probability of voting for Bush when ideology and partisanship were held at their 2000 mean

(moderate Independent). When hawkishness increases from .55 to .75, the respondent goes from

being indifferent between Kerry and Bush (49 percent probability of voting for Bush) to being a

strong Bush supporter (63 percent probability of voting for Bush).

8 Respondents who preferred a combination of dovish and hawkish action were excluded from this

measure.

9 Because the Threat Theory expects that evocative imagery should only move respondents toward

the hawkish end of the scale, Figure 2 utilizes one-tailed significance tests and shows 90 percent

confidence intervals.

10 Comparing the mean hawkishness score among high threat respondents in the control condition

to high threat respondents in the scary visuals condition, the difference in means across the 5

dependent variables ranges in size from .03 to .10 on scales that range from 0 to 1, with the

exception of the spending scale that ranges from -1 to 1.

11 The Emotion variable is constructed from respondents’ emotional reactions to the scary visuals

condition that were described in the manipulation check section. Fear and anger scaled together on

one dimension in a factor analysis, so I combined these two emotional responses into one emotion

index and divided respondents by the median. Respondents were categorized as low emotion

(n=514) if they were below the median on either the fear index or the anger index. Respondents

were categorized as high emotion (n=715) if their scores on either the fear or anger indices was

above the median or if they were above the median on both indices.