Terrified or Enraged? Emotional Microfoundations of Public Counterterror Attitudes

Abstract Despite the widespread assumption of terrorism's “terrifying” effect, there has been little systematic testing of the specific emotional microfoundations underlying public opinion about terrorism. While fear is one well-recognized emotional response to terror threats, in societies where terrorism is rare, anger may play a more pivotal role, with distinct consequences for citizens’ downstream political attitudes. To test the impact of these emotional mechanisms on public opinion in the wake of terrorism, I employ a multi-arm mechanism experiment (n = 5,499) in the United States that manipulates both exposure to news about different types of terror attacks and the encouraged emotional response. I supplement this experimental study with observational analyses of the emotional content of social media posts in the wake of sixteen real-world terror attacks in the United States. I find that not only is anger the dominant emotional response to terrorism across both studies, but also that punitive motivations and support for retaliation are both directly shaped by experimentally induced anger after exposure to news about terrorism. These findings illuminate strategic incentives shaping militants’ use of terror tactics, electoral constraints leaders face in formulating counterterror policy, and the emotional mechanisms fueling cycles of political violence.

threat following terrorist attacks, only a minority report high levels of fear or anxiety. 4he minority that do experience high fear are often less supportive of aggressive counterterror policies, 5 even though publics in general tend to be more militant following terrorism. 6The public also reacts more strongly to out-group perpetrators than in-group ones, even when they are similarly violent. 7Together, these findings suggest that the dominant driver of public opinion about terrorism is often not fear and a desire to reduce personal risk, but anger and a desire for vengeance against perpetrators whom citizens deem worthy of punishment.
I investigate these distinct emotional mechanisms using a parallel-encouragement experiment in the United States (n = 5,499).This design is explicitly structured to test causal mechanisms by manipulating both exposure to news about different types of terrorism and the encouraged emotional response. 8I find that, while both anger and fear increase support for retaliation, anger is both a more dominant emotional response and the only emotion that, when induced, increases punitive motives for retaliation.Inducing anger also significantly shrinks partisan gaps in support for retaliation following attacks by different perpetrators, providing further evidence of anger's causal role in shaping public attitudes.
Because public opinion can influence state security policy choices and shape militant incentives, 9 the degree of public anger following terrorist violence has problematic implications for retaliatory spirals of conflict. 10If terror attacks primarily engender outrage and the desire to punish perpetrators, they could incentivize electorally minded leaders to exact vengeance, even if this engenders a "backlash" that increases future terror risk. 11These results thus shed light on a core puzzle in the terrorism literature: why groups that use terrorism rarely achieve concessions yet often goad democratic governments to overreact with military action, making terrorism more effective as a strategy of provocation rather than attrition. 12This research also highlights the increasingly partisan emotional pathways through which terrorism shapes attitudes toward different perpetrators, pointing to potentially changing in-group/out-group perceptions in an era of heightened partisan polarization and posing a thorny challenge for leaders seeking to effectively counter both threats. 13ear and Risk: The "Terror" in Terrorism Terrorism has a strong effect on the political attitudes of the mass public.After terror attacks, citizens prefer costlier counterterror policies, at times seeking out concessions, but more often supporting more militant policies and parties. 14The predominant explanation for these preferences, in both scholarly and public accounts, centers on perceptions of risk and the emotion of fear. 15Whether due to rational updating about personal risk following attacks on civilians or inflated risk estimates stemming from cognitive biases, citizens are thought to experience increased fear after terrorism that, in turn, shapes their political attitudes. 16he identity of terrorism's victims is central to this conception of fear's role in influencing public opinion.Citizens concerned with their personal safety witness attacks on civilians and update (rationally or not) their assessments of their own future risk of being targeted, particularly when the victims are perceived as "similar to them." 17 These elevated risk perceptions, even if vastly overestimated, 18 make citizens more fearful after terror attacks than they would have been had the violence not targeted civilians.
However, few have directly examined the centrality of this posited emotional pathway or the relative importance of civilian victimization in shaping public opinion.The importance of fear is simply assumed, perhaps because the public does report increased levels of threat in the wake of terrorist violence and routinely lists terrorism among their top security concerns.Yet, the public's concern about terrorism is a measure of threat perceptions, not the emotional response of fear.Though "threat" and "fear" are often used interchangeably in discussing public concerns about terrorism, they are not the same thing.Fear is an emotional response to threat, and it is only one of many possible emotional responses. 19Anger is another, and has distinct implications for citizens' political motivations and attitudes.

Anger and Morality: The Politics of Retribution
Anger is a powerful driver of political attitudes and behavior and is particularly important in violent political conflict. 20Anger's deep-seated impact is likely because anger is a moral emotion, elicited not simply when a negative, threatening event occurs but when that event or action is perceived as both purposeful and unjust. 21Anger is thus rooted in value-based appraisals of the legitimacy or fairness 14.Godefroidt 2022.15.Albertson and Gadarian 2015; Mueller and Stewart 2012; Nacos, Bloch-Elkon, and Shapiro 2011.16.Getmansky and Zeitzoff 2014; Silverman, Kent, and Gelpi 2022.17.Avdan and Webb 2019.18. Mueller and Stewart 2012.19.Huddy et al. 2021.20.Petersen and Zukerman 2010.21.Haidt 2003; Roseman 1996; Scherer 2000.  of an action and the blameworthiness of the perpetrator and may moralize and polarize political attitudes. 22ognitive-appraisal theories of emotion posit that anger and fear, though both negatively valenced, stem from distinct perceptual foundations and thus lead to different behavioral tendencies. 23Fear stems from an appraisal of heightened personal risk or weakness, activating an avoidance tendency.Anger is triggered by appraisals of comparative strength, combined with attributions of blame toward someone perceived to have transgressed norms, engendering an approach tendency.The emotions are not mutually exclusive-self-reports of fear and anger are often highly correlated-but, to the extent that individuals feel both emotions, one is usually experienced with higher intensity than the other and so will be a more prominent driver of political attitudes. 24rom a theory-of-ethics perspective, fear is associated with a consequentialist logic, while anger engenders deontological thinking: actions must follow moral rules rather than simply achieving the best outcomes. 25This has implications for the motives underlying public support for retaliation and conciliation following terrorist violence.For fearful individuals, preferences for retaliation are likely motivated by defensive impulses to prevent future violence, while, for angry individuals, support for retaliation may be motivated by offensive aims, 26 punishing the perpetrators simply because they deserve it.These different retaliatory logics parallel the motives underlying public opinion on criminal justice: rehabilitation, incapacitation, deterrence, and punishment. 27The first three motives are consequentialist, designed to reduce the future risk of violence by the perpetrators themselves or by other, likeminded perpetrators.Punitive motives, on the other hand, are deontological, focused on avenging historic wrongdoing. 28Of the four, punitive motives have repeatedly been found to be the most dominant predictor of support for harsh criminal justice policies, including the death penalty. 29Political science research has documented a similar connection between punitive motives and support for the use of military force. 30However, the emotional antecedents of these retributive preferences in foreign policy have been less well studied. 31 theorize that the reason exposure to terrorism news so often increases public support for retaliation and decreases support for concessions is because these 22. Clifford 2019; Lickel et al. 2006; Petersen 2010.23.Roseman 1996; Scherer 2000.24.Rhodes-Purdy, Navarre, and Utych 2021.25.García-Ponce, Young, and Zeitzoff 2023.26.Coan et al. 2021.27.McKee and Feather 2008.28.Crockett, Özdemir, and Fehr 2014.From an evolutionary standpoint, the prevalence of punitive motives in humans may be because it is an adaptive strategy that deters future bad behavior by others (McDermott, Lopez, and Hatemi 2017), but at the individual level the motive is still past oriented.
29. Carlsmith, Darley, and Robinson 2002.30.Liberman 2006; Stein 2015.31.Pagano and Huo 2007.Terrified or Enraged?827 attacks make people angry and this, in turn, makes punishment a primary motive when deciding which policies to support (Figure 1).While fear would generate a flight or fight response that could increase support for retaliation in some cases, 32 it would be motivated by an impulse to prevent future violence. 33In contrast, angry citizens' support for retaliation should be motivated by a desire not simply to prevent future attacks but also to punish past ones.Moreover, while fear can increase risk aversion and support for conciliatory policies in some cases, 34 angry publics should uniformly oppose concessions for attackers they deem worthy of punishment.This means that, relative to fear, public anger in the wake of terrorism could increase the range of circumstances under which citizens would support retaliation.

Innocents, Out-Groups, and Partisanship
Why is terrorism so likely to evoke anger and the desire for retributive retaliation?First, terrorist attacks often target civilians.Attacking these blameless "innocents" transgresses modern norms regarding the use of force and may be seen as particularly extreme or unjust by the public, 35 a core appraisal tied to anger. 36Thus, public anger and support for retaliation may be higher after attacks on civilian, as opposed to military, targets.On the other hand, terrorist attacks may trigger anger connected to the identity of the perpetrators: an ethnic or religious out-group.Individuals have been shown to express more anger and support greater punishment 37 toward negative 33.Coan et al. 2021; Silverman, Kent, and Gelpi 2022.34.Gould and Klor 2010; Lerner et al. 2003; Thomas 2014.35.Abrahms 2006; Fortna 2015.36.Haidt 2003.37. Piazza 2015; Schiller, Baumgartner, and Knoch 2014.actions by out-group offenders, 38 particularly when the out-group is perceived as "entitative." 39Anger can also be exacerbated by the ultimate attribution error, an intergroup cognitive bias whereby negative actions by out-group members are more readily attributed to character flaws. 40Indeed, though scholarly definitions of terrorism focus on the victims as the defining feature of terrorist (as opposed to guerrilla) violence, 41 the mass public readily labels attacks on military targets as terrorism, 42 but is less likely to label white attackers as terrorists, even when they target civilians. 43Collectively, this evidence suggests that public anger following a terror attack is driven more by the perceived blameworthiness of out-group, as opposed to in-group, perpetrators.The public's emotional responses to terrorism may also be shaped by citizens' preexisting political attitudes.While past studies have demonstrated that jihadist terrorism increases hawkish preferences across the political spectrum, 44 responses to terrorism have become increasingly politicized.In the United States, Democrats have grown wary of the "war on terror" against jihadist groups and more concerned about other terror threats, including a resurgent white nationalist movement. 45The growing association between the Republican Party and the "alt-right" movement under the Trump administration, likewise, may have dampened support from Republicans (and increased support from Democrats) for policies to counter domestic white nationalist groups. 46This could be due to a recategorization of salient in-group/ out-group dynamics in an era of heightened partisan polarization. 47Democrats may now view white nationalist perpetrators as part of a partisan out-group, rather than as ethnic and religious in-group members. 48Moreover, Democrats may be more likely to perceive the victims of white nationalist violence as in-group members, due to their assumed partisan, ethnic, or religious affiliation.Since anger toward offenders is stronger when the victim is seen as an in-group member and the perpetrators as an out-group, 49 this recategorization may differentially shape the emotional and political responses of Republicans and Democrats to attacks by different perpetrators.

Research Design
The present research tests this theoretical argument using a preregistered survey experiment with a parallel-encouragement design that manipulates both exposure 38.Van Prooijen 2006.39.Lickel et al. 2006.40.Coleman 2013; Pettigrew 1979.41.Ganor 2002.42.Huff and Kertzer 2018.43.D'Orazio and Salehyan 2018.44.Godefroidt 2022.45.Snook et al. 2022.46.Caton and Mullinix 2022.47. Mason 2018.48.Kam and Kinder 2007.49.Gordijn, Wigboldus, and Yzerbyt 2001; Van Prooijen 2006.Terrified or Enraged?829 to terrorism news and the encouraged emotional response before assessing political attitudes (Figure 2). 50The survey was conducted using Lucid Marketplace in March 2021 among 5,499 US adults, with rough census representation on age, region, gender, ethnicity, and partisanship. 51nipulating both the treatment and mediator is an important innovation to directly test causal mechanisms in experimental studies, as it addresses the risk of post-treatment bias that could otherwise be induced by conditioning on posttreatment variables (i.e., experienced emotions) in a traditional design where only the treatment level is manipulated. 52

Experimental Treatments
The first treatment is exposure to a fictitious news report about a terrorist attack in the US that "killed at least two US citizens and wounded dozens more," with two varying features: the victim of the attack (civilians at a mall or troops at a base) and the perpetrators of the attack (jihadists or white nationalists).Until debriefing, subjects FIGURE 2. Survey Flow 50.Imai, Tingley, and Yamamoto 2013.51.For additional sample descriptives, see section A in the online supplement.52.Acharya, Blackwell, and Sen 2018; Imai, Tingley, and Yamamoto 2013.In observational surveys or standard experiments where only exposure to terrorism news is randomized, mediation analysis is used to assess effects of emotions on attitudes (e.g., Fisk, Merolla, and Ramos 2019; Huddy et al. 2005; Lambert  et al. 2010; Skitka et al. 2006).Though these studies find associations between anger and militancy, this analysis relies on a strong sequential ignorability assumption.believe this is a real article about a terror attack that happened the day before. 53mportantly, exposure to terrorism news may have different effects than physical proximity to a terror attack. 54However, in the US and many other Western countries, the vast majority of the public's real-world exposure to terrorism comes from media reports rather than personal experience. 55Using news about terrorism as a treatment to assess the effects of terrorism on public opinion is thus a common approach. 56ubjects in the natural mediator arm do not receive any emotion prime and are asked to write their thoughts on the article before proceeding to answer questions about their counterterror attitudes.Subjects in the manipulated mediator arms are exposed to a second treatment to induce different emotions: anger, fear, or reassurance.Because negative emotions-particularly anger-are theorized to be naturally high after exposure to terrorism news, the reassurance prime sought to dampen this emotional response to assess the effect of exposure to terrorism news, independent of the anger that follows.These arms use a well-established emotion induction called an autobiographical emotional memory task (AEMT). 57Respondents are prompted to recall and write about aspects of terrorism that have led them to experience the assigned emotion strongly.This addresses two key threats to inference that otherwise impact the study of emotions in politics: ex ante characteristics that lead individuals to have specific emotional responses to terrorism, and the role new information about a threat plays in triggering the emotion. 58

Ethical Considerations
This study involves the use of deception.Research participants believe they are reading real articles, but these stories were developed by the researcher.Using fake or altered news stories attributed to real media is a frequent approach in experimental political science. 59However, this deception can pose risks.For example, prior to debriefing, participants may experience emotional distress from reading news about political violence.There is also a concern that participants might become suspicious of news media and/or academic researchers after being made aware of the deception.It is thus important to consider whether deception is necessary for the research. 60.An extended discussion of experimental ethics is given later and in section C of the online supplement.See section D of the online supplement for information about treatment believability and manipulation checks.
54. Terrified or Enraged?831 In this study, deception is critical because the experiment examines the role that visceral emotional reactions to encountering news about terrorism play in shaping citizens' political attitudes.While hypothetical scenarios or abstract vignettes may be suitable for the study of some attitudes, 61 these tools are less well suited for inducing the strong emotional responses citizens feel after exposure to terrorism news.Indeed, abstract scenarios are particularly problematic when addressing issues concerning morality and justice, 62 as this study does, because research participants' answers to "decontextualized hypothetical scenarios may not accurately reflect moral decisions in everyday life." 63ince experimental realism is so important to the design's validity, one might instead consider using real articles about past attacks.However, this too is problematic.Because real-world attacks naturally vary on so many dimensions, disentangling the effects of the attributes of interest (victim and perpetrator identity) on emotional and political responses to terrorism would be extremely confounded. 64This problem is further compounded by the distinct ways the media covers attacks by different perpetrators. 65Even if it were possible to identify two real attacks similar to each other except on the dimensions of interest, given the widespread real-world coverage, respondents are likely to have already developed various preconceived beliefs about the attacks and what should be (or already has been) done in response.These beliefs could be shaped by a host of longer-term processes that take place following terrorism, including elite framing, 66 making it nearly impossible to assess the independent causal effect of these immediate, bottom-up emotional reactions to terrorism news on the public's political attitudes.
Ultimately, this research investigates core questions regarding how the public responds emotionally to terrorism, shaping the incentives of political leaders formulating counterterror policy.Given these considerations, I conclude that deception is necessary to effectively assess the role of public outrage in potentially exacerbating vengeful cycles of terrorist and counterterrorist violence, a question of immense political importance with stark real-world consequences for the populations impacted by this violence.
In addition to the need for deception, the Principles and Guidance for Human Subjects Research of the American Political Science Association (APSA) outline four other considerations for researchers using deception: whether the study involves minimum risk of harm for participants; whether consent would have been withdrawn if participants had full knowledge of the deception; potential power imbalances between the researcher and participants; and whether debriefing of participants is possible. 67I discuss each of these considerations, the various other APSA principles for 61.Brutger et al. 2022.62. Schein 2020.63.FeldmanHall et al. 2012, 432.64.Huff and Kertzer 2018.65.Dreier et al. 2022.66. Nacos, Bloch-Elkon, and Shapiro 2011; Norris, Kern, and Just 2004.67.American Political Science Association 2020.human subject research, and the potential broader impacts of the use of deception for trust in media and researchers in section C of the online supplement.

Measurement
The core dependent variables in this study are respondents' support for conciliatory or retaliatory policies against terrorists and, crucially, their rationales for supporting retaliation: punitive ("punishing the perpetrators for their wrongs") or preventative (e.g., "incapacitating the perpetrators so they cannot commit future attacks").Subjects are also asked to self-report the degree to which they feel various emotions about terrorism.Emotions are measured with a composite index, where anger is the average of anger, outrage, and disgust (α = 0.84) and fear is the average of fear, anxiety, and worry (α = 0.86).Previous research has found these measures load onto two distinct factors. 68

Hypotheses
The survey experiment tests eight preregistered hypotheses, with comparisons across different treatment arms providing causal leverage on different questions. 69The first five hypotheses are tested using the natural mediator arms of the experiment, where subjects read news about terrorism but do not receive an AEMT task: • H1: Exposure to news of an attack will increase support for retaliation and decrease support for conciliation with terrorists compared to no attack.• H2: In particular, exposure to news of an attack will increase punitive rationales for retaliation against terrorists compared to no attack.• H3: Exposure to news of attacks on civilian targets will not result in different levels of support for retaliation or conciliation relative to attacks on military targets.• H4: Exposure to news of attacks by jihadists will increase support for retaliation and reduce support for conciliation relative to exposure to news of attacks by white nationalists.• H5: Exposure to news of attacks by white nationalists will interact with partisanship such that Republicans will be less likely than Democrats to support retaliation and more likely to support conciliation.
The other three hypotheses leverage the double randomization of the parallelencouragement design, examining subjects across the manipulated mediator arms of the study: 68.Valentino et al. 2011.See section B of the online supplement for a full list of variables and question wording.Respondent partisanship data are from the Lucid survey panel, avoiding concerns about potential post-treatment bias in reporting partisan identity.
69.For the test of one additional preregistered hypothesis regarding the direct effect of manipulating only emotions (no terrorism news) on attitudes, see section F.2 in the online supplement.
Terrified or Enraged?833 • H6: Exposure to news of any type of attack in conjunction with an anger prime will increase support for retaliation and decrease support for conciliation relative to news exposure in conjunction with a reassurance prime.• H7: Exposure to news of any type of attack in conjunction with an anger prime will increase punitive rationales for retaliation against terrorists relative to both a reassurance and a fear prime.• H8: Differences in political attitudes between respondents exposed versus not exposed to terrorism news will be smaller in the manipulated anger arms than they are in the natural mediator arms.

Emotions and Exposure to Terrorism News
The first set of analyses compares the five treatment arms in the natural mediator arms.
In four of these conditions, subjects are exposed to news stories about different types of terror attacks.These are compared to the control: subjects who read a nonpolitical story.
The dominant emotional response of subjects to terrorism news is anger rather than fear: 62 percent report higher anger than fear following exposure to terrorism news; only 18 percent report higher fear.A paired sample t-test of anger and fear within each news treatment arm indicates that self-reported anger is significantly higher than fear regardless of attack type (Figure 3a).These patterns are consistent across ethnicity, gender, religion, residence, education, and income.Moreover, after exposure to news of an attack, anger, but not fear, significantly increases, regardless of perpetrator Notes: Figure 3a displays mean levels of anger (orange, lighter) and fear (navy, darker) in each condition on a scale from 1 ("not at all") to 5 ("extremely").Black bars represent standard errors of the mean.See also Table H.1 in the online supplement.Figure 3b displays the coefficient estimate of the effect of exposure to terrorism news on anger (orange) and fear (navy) compared to no terrorism news, with 90% and 95% confidence intervals.See also Table H.2 in the online supplement.Natural mediator arms only.

FIGURE 3. Relative Levels of Anger and Fear in Response to Terrorism News
or victim identity (Figure 3b). 70In short, subjects are significantly more outraged after exposure to news of an attack, but are not significantly more terrified.This belies the popular narrative of terrorism's invariably fear-inducing effect on civilian populations.

Effects of Exposure to Terrorism News on Political Attitudes
Given the prominence of anger following terror attacks, exposure to any form of terrorism news was hypothesized to increase support for retaliation and decrease support for conciliation (H1), while also increasing punitive rationales for the use of force (H2).I find, however, that exposure to terrorism news only marginally increases support for retaliation (Table 1, model 1) and does not significantly affect support 70.Respondents who live geographically close to the attack, in the attacked state or a state directly bordering it (n = 469), also exhibit higher anger (but not fear) after exposure to an attack.
Terrified or Enraged?835 for conciliation or punitive motivations. 71These muted effects are surprising because past research has demonstrated a small but consistent rightward shift following terrorism, particularly in the US. 72 closer examination, it is clear that the small main effects are driven by heterogeneity in partisans' responses to different forms of terrorism, as hypothesized in H5: Republicans are significantly less angry, punitive, and retaliatory than Democrats following an attack by a white nationalist, while the pattern is reversed for jihadists (Figure 4). 73Because Democrats and Republicans have opposing emotional and political responses to news about different attacks, the overall effect of exposure to terrorism news on support for retaliation is small.Moreover, because Democrats are less supportive of retaliation at baseline, attacks on the military are actually marginally more likely to increase support for retaliation than attacks on civilians, and attacks by white nationalists are significantly more likely than Notes: Figure 4 displays the estimates from OLS models examining the interaction of party ID and perpetrator identity on self-reported anger, punitive motives, and support for retaliation.Jihadists are the reference category, represented by the dotted vertical line.Natural mediator arms only.See also attacks by jihadists to do so, in contrast to hypotheses H3 and H4, respectively (Table 1, models 2 and 3).
Together, these results indicate that the level of anger after terrorist violence varies by both party and perpetrator identity, which may shape differential partisan support for retributive retaliation.

The Causal Impact of Anger
But does anger play a causal role in motivating support for retaliation among Democrats and Republicans, or is it simply a downstream consequence of ex ante political attitudes?To directly examine this question, we next turn to the manipulated mediator arms, analyzing attitudes across arms where the type of attack differed but the emotional prime was the same.
Two hypotheses were specified regarding the indirect effect of emotions induced in the wake of terrorism on political attitudes: both induced anger and fear will increase support for retaliation and decrease support for conciliation (H6); and only anger will  Terrified or Enraged?837 increase the importance of punitive rationales in justifying this response (H7).Figure 5 illustrates preferences for retaliation and conciliation among those in the manipulated fear arms and anger arms as compared to those who received no prime. 74While support for conciliation is unmoved, fear and anger both increase support for retaliation-however, note that the effect of anger (p = 0.079) is not as precisely estimated as that of fear.However, the rationales for these retaliatory preferences among fearful and angry respondents vary, as predicted.Inducing respondents to feel angry-but not fearful-significantly increases punitive rationales for the use of force (Figure 6).74.The reassurance prime, whose aim was to reduce the negative emotions of fear and anger respondents might naturally feel after terrorism, did not significantly reduce either emotion as compared to "no prime" (see section D.2 in the online supplement).Thus, when analyzing the effect of anger and fear on these attitudes, I compare respondents in these arms to respondents who received no emotional prime.Results are, however, substantively similar when comparing respondents in the anger and fear arms to the reassurance arms; see section F.1 in the online supplement.

Anger
In the anger treatment, the coefficient on support for punitive motives (β = 0.137) is 2.5 times that in the fear treatment (β = 0.056). 75This increased support for retributive retaliation among angry individuals is notable, given that individuals are likely to under-report the extent to which retributive motives guide their preferences. 76In contrast, encouraging respondents to feel fearful only significantly increases the motive to incapacitate would-be attackers, as predicted by an appraisal framework that theorizes fear as an avoidant emotion primarily aimed at reducing future harm.There is thus a significant indirect effect of exposure to terrorism news that flows through the emotional experience of anger to shape punitive motives and support for retaliation.
The changes in underlying motives engendered by anger and fear help explain the different political attitudes partisans were shown to hold in the wake of attacks by different perpetrators.Inducing Republicans and Democrats to experience the same emotional reaction to jihadist and white nationalist violence, then, should reduce the partisan gap in support for retaliation, making the interaction of perpetrator identity and partisanship nonsignificant (H8). 77his is indeed what we find.The partisan gap in political attitudes in conditions in which attacker ID varies, but the emotional prime (e.g., anger) is the same, are smaller than the partisan gap in the natural mediator arm of the study, where no emotional prime is given.Without priming emotions, the effect size of the partisanship-perpetrator interaction on support for retaliation is β = −0.72.When priming anger, it is much smaller: β = −0.33.This means anger experienced in the wake of exposure to terrorism news explains nearly half of partisanship's differential effect on support for retaliation following white nationalist versus jihadist attacks.This is primarily driven by changes in responses to jihadist violence, where partisan attitudes are more dissimilar at baseline (Figure 7).After receiving an anger prime, Democrats, who are otherwise significantly less retaliatory and punitive toward jihadists than Republicans are, adopt much more retributive preferences.For Republicans, on the other hand, induced anger leads them to adopt a similarly retributive stance toward white nationalists as they do toward jihadists.
to give each tweet a score representing the percentage of words in the tweet present in the LIWC anxiety and anger dictionaries.For example, the following tweet about the San Bernardino attack received a high anger score: "Angered by the news of the San Bernardino mass shooting today; not just saddened, angered."In contrast, this tweet following the El Paso Walmart shooting scored high on anxiety, but low on anger: "Im like so freaking terrified, I live in El Paso and the shooting is just ridiculous like im scared their gonna come to my area and its not safe right now." For nine out of sixteen terror attacks, paired sample t-tests indicate that the average anger score of tweets is significantly higher than the average anxiety score, while anxiety is significantly higher after only four attacks (Figure 8). 80Thus, people are often more likely to express anger than fear after real-world terrorist violence, even in the immediate aftermath of these events, when fear would likely be at its height.Specifically, among these sixteen attacks, those perpetrated by white nationalists appear most likely to engender anger on Twitter.Given the partisan skewedness of the platform, with Democrats over-represented among all users and among the most prolific ones, 81 this trend is consistent with the experimental results: that partisanship and perpetrator identity interact to influence emotional and political  Terrified or Enraged?841 responses to terrorism.While social media analysis has important limitations, these data strongly suggest that anger has played a predominant role in shaping public responses to many real-world terror attacks over the past decade.

Discussion and Implications
This study shows that anger and the desire for retribution can shape citizens' attitudes in the wake of terrorism.It thus provides a deeper understanding of the distinct emotional microfoundations driving public opinion about terrorism, with key implications for how we understand cycles of terrorist violence.
First, the dominant emotional response of citizens to exposure to terrorism news is often anger rather than fear.This may influence the strategic calculations of militant organizations.Terrorist tactics may help a weaker group provoke retaliation from a state, bringing the group followers and recruits in the long run. 82However, the attack is unlikely to lead to concessions, 83 unless the group can sustain a large enough campaign to truly engender a feeling of vulnerability and fear in a target population.This may explain why studies that show terrorism "works" in securing concessions are often done in countries fighting civil wars. 84This study thus contributes to a growing body of work assessing how affect shapes strategy in international conflict. 85econd, anger experienced in the wake of terrorism increases citizens' focus on punishing perpetrators, not just preventing future terrorism.This has problematic implications for cycles of political violence, as angry publics will likely pressure their leaders to retaliate under a broader range of circumstances than fearful ones, including when retaliation risks a terrorist backlash. 86Interrogating these distinct emotional motivations undergirding public support for retaliation is thus important for understanding when political violence is likely to escalate into retaliatory spirals. 87he increasingly partisan response of Americans to terrorist violence is another striking finding of this study.This is a stark departure from historic patterns, in which terror attacks often unified public opinion, such as following the Oklahoma City bombing and the September 11 attacks.These changes are likely driven, in part, by the increased politicization of terrorism in the US since the 2016 election of Donald Trump.However, polarization over counterterror policy was growing before the Trump administration and is also prevalent in other countries. 88These partisan divides pose a difficult political dilemma for elected officials seeking to marshal public support for policies to counter both types of threats. 89An important future 82.Kydd and Walter 2006.83.Abrahms 2006; Fortna 2015.84.Thomas 2014.85.Little and Zeitzoff 2017; Renshon, Lee, and Tingley 2017; Schnakenberg and Wayne 2022.86.Piazza and Choi 2018.87.McDermott, Lopez, and Hatemi 2017.88.Carothers and O'Donohue 2019.89.Schultz 2017.direction for this work, then, is to examine the conditions under which attacks trigger more uniform (as opposed to polarized) public outrage.Based on the findings of this study, it is likely that changing perceptions of the in-group/out-group status of terrorism's victims and perpetrators may play a prominent role. 90inally, these findings have important policy implications for how leaders interact with their citizenry in the wake of terror attacks.For one, political elites may be overestimating the extent to which citizens are terrified of terrorist violence.If the public is angry rather than scared following terrorism, attempts by leaders to reduce public fear may backfire if citizens interpret this messaging as allowing blameworthy perpetrators to avoid punishment.This may be particularly challenging for dovish leaders. 91Of course, the public's anger in response to terrorism also means hawkish leaders can seize on public outrage for electoral benefit and to pursue more aggressive counterterror policies. 92In particular, moralized counterterror frames will likely garner more ready support from an outraged public than frames focused on risk reduction. 93This may explain the frequent use of this rhetorical strategy by leaders, such as George W. Bush's famed "axis of evil" speech, 94 Israeli prime minister Benjamin Netanyahu's statement that "evil has to be resisted," 95 and proposals for counterterror policies that emphasize punishment, such as Donald Trump's 2016 call to "take out" the families of suspected terrorists. 96Thus, in the wake of terrorism, leaders are constrained by public anger and its increasingly partisan manifestations, but they also have significant power to channel this outrage into a range of potential counterterror policies at home and abroad.

FIGURE 1 .
FIGURE 1. Theorized Causal Path of the Effect of Terrorism News on Attitudes Figure 5 displays the estimates from OLS models regressing the randomized emotion prime (anger or fear) on support for retaliation and conciliation.No emotion prime is the reference category, represented by the vertical dotted line.Excludes arms with no terrorism news.See also TableH.6 in the online supplement.

FIGURE 5 .
FIGURE 5. Effect of Distinct Emotions After Exposure to Terrorism News on Political Attitudes Figure 8 displays the mean percentage of anger (orange, lighter) and anxiety (navy, darker) words in all tweets identified as relevant during the seven days following each attack.Error bars represent 95% confidence intervals.See also TableG.2 in the online supplement.

FIGURE 8 .
FIGURE 8. Sentiment Analysis of Tweets Surrounding Real-World Terror Attacks

TABLE 1 .
Support for Retaliation by Terrorism Type and Partisanship Notes: Table1displays the results of five OLS models regressing attack type on support for retaliation.Model 1: any attack compared to no attack.Models 2 and 5: military to civilian attacks.Model 3 and 4: white nationalist to jihadist attacks.Models 4 and 5 include a three-level partisanship variable interacted with attack type, with Democrats as the reference category.Natural mediator arms only.^p<.10; *p<.05; **p<.01;***p<.001.
FIGURE 4. Interaction of Partisanship and Perpetrator ID on Responses to Terrorism News 71.See Tables H.3-H.4 (model 1) in the online supplement.72.Godefroidt 2022.73.Republicans are also marginally more retaliatory than Democrats following an attack on the military (Table 1, model 5).
Figure 6 displays the estimates from OLS models regressing the randomized emotion prime (anger or fear) on motives for retaliation: rehabilitation, incapacitation, deterrence, and punishment.See also Table H.7 in the online supplement.FIGURE 6.Effect of Distinct Emotions After Exposure to Terrorism News on Punitive Rationales Notes: