A Bottom-Up Theory of Public Opinion about

Foreign Policy

Joshua D. Kertzer⇤ and Thomas Zeitzo↵†

Last revised: November 7, 2016

Abstract: If public opinion about foreign policy is such an elite-driven process, whydoes the public often disagree with what elites have to say? We argue here that elite-cue-taking models in IR are both overly pessimistic and unnecessarily restrictive. The publicmay lack information about the world around them, but it does not lack principles,and information need not only cascade from the top down. We present the resultsfrom five survey experiments where we show that cues from social peers are at least asstrong as those from political elites. Our theory and results build on a growing numberof findings that individuals are embedded in a social context that combines with theirgeneral orientations towards foreign policy in shaping responses towards the world aroundthem. Thus, we suggest the public is perhaps better equipped for espousing judgmentsin foreign a↵airs than many of our top-down models claim.

9463 words (including footnotes and bibliography)

⇤Assistant Professor of Government, Harvard University. 1737 Cambridge St, Cambridge MA 02138. Email:jkertzer@gov.harvard.edu. Web: http://people.fas.harvard.edu/˜jkertzer/.

†Assistant Professor, School of Public A↵airs, American University. 4400 Massachusetts Avenue NW, WashingtonDC 20016. Email: zeitzo↵@american.edu. Web: http://www.zeitzo↵.com/.

1 Introduction

In July 2014, another wave of violence erupted in the Middle East, as Israel responded to a barrage

of rockets from Gaza by launching airstrikes, and eventually, a ground incursion intent on degrading

Hamas’ military capabilities and destroying a web of underground tunnels being used to launch

covert attacks. In Washington, both Democrats and Republicans firmly sided with Israel: the Senate

passed an unanimous resolution blaming Hamas for the conflict, and both prominent Democrats and

Republicans gave staunch defenses of Israel’s right to defend itself. In an interview on ABC on July

20, Secretary of State John Kerry summed up the White House’s position — and with it, the

Republicans’ position as well — that “when three young Israeli kids are taken and murdered and

Hamas applauds it. . . and then starts rocketing Israel when they’re looking for the people who did

it, you know, that’s out of balance by any standard” (ABC News, 2014).

Although both Democrats and Republicans in Washington were united in their support of Israel,

a series of polls found that Democrats and Republicans in the public were divided. In a Pew poll

from July 24-27, 60% of Republicans blamed Hamas for the violence, while Democrats were split,

with 29% blaming Hamas, and 26% blaming Israel (Pew Research Center, 2014). A Gallup poll from

July 22-23 detected a similar pattern: 65% of Republicans thought Israel’s actions were justified, but

Democrats were divided, as 31% backed the Israeli response, and 47% called it unjustified (Jones,

2014). This pattern — where political elites are united but the public is divided — is particularly

interesting for political scientists because it violates the assumptions of a commonly held theory

about public opinion, in which the public knows relatively little about foreign a↵airs and thus

structures its beliefs by taking cues from trusted, partisan elites — a top-down process in which

members of the public adeptly swallow whatever their preferred elite cue-givers feed them. Yet if

the mass public knows so little and can only regurgitate carefully pureed talking points, why does

it often disagree with what elites have to say?

We argue here that partisan elite cue taking models are both overly pessimistic, and unnecessarily

restrictive: the public may often lack information, but it doesn’t lack principles, and information

need not cascade from the top down. We present the results from five survey experiments where we

explore the limits of elite partisan cues in foreign a↵airs. Across all five experiments, fielded in three

2

studies across two years, we show that cues from social peers are as least as strong as those from

political elites, and in some cases, stronger. Additionally, even in the absence of cues, individuals

have general predispositions towards foreign policy they can rely on when forming attitudes towards

specific policy issues. Together, these findings suggest that the role of elite cues should be understood

in a broader context about the information environment in which citizens are embedded, and the

role of political orientations beyond partisanship. We make this argument in three parts. First, we

review the literature on public opinion about foreign a↵airs, showing how scholars in the past half

century have oscillated from pessimism to optimism and back again. Second, we point to a number

of both theoretical and empirical reasons that should encourage us to relax some of the assumptions

undergirding top-down models of public opinion. We then present our barrage of experimental

results, and conclude by discussing some of the implications of our findings.

2 Three images of the public in foreign a↵airs

The public opinion about foreign policy literature is rich and multifarious, but like Caesar onto Gaul,

we can crudely divide it into three parts.

In the aftermath of the Second World War arose what came to be known as the “Almond-

Lippmann consensus” (Almond, 1950; Lippmann, 1955): a pessimistic view that held that public

opinion on foreign policy issues was ill-informed and ill-structured (Holsti, 2004). Kennan (1951,

59) compared democratic publics to “one of those prehistoric monsters with a body as long as

this room and a brain the size of a pin”, while Almond (1950, 232) suggested that the American

public’s reaction to international events “has no depth and no structure.” Perhaps unsurprisingly,

many of the advocates of this cynical view tended to be foreign policy realists, eager to insulate the

intricacies of foreign policy-making from what they saw as an unsophisticated and emotional public

(Morgenthau, 1948).

In reaction to the postwar cynics (and more methodologically sophisticated counterparts, like

Converse 1964) have come a series of optimistic rejoinders showing that foreign policy attitudes in-

deed have structure (Hurwitz and Pe✏ey, 1987; Holsti, 1992), and that the public reacts predictably

and prudently to world events (Page and Shapiro, 1992; Jentleson, 1992; Kertzer, 2013), most no-

3

tably casualties (Mueller, 1971; Gartner, 2008). The public has principles when it came to foreign

policy: it likes victory (Eichenberg, 2005) and success (Gelpi, Feaver, and Reifler, 2009), dislikes

inconsistency (Tomz, 2007), likes multilateralism (Chaudoin, Milner, and Tingley, 2010), and has

stable and well-structured foreign policy orientations (Holsti, 1979; Wittkopf, 1990; Herrmann, Tet-

lock, and Visser, 1999; Rathbun, 2007), rooted in core values (Rathbun et al., 2016; Goren et al.,

2016) and encoded into our genes (McDermott et al., 2009). Although these approaches are remark-

ably varied, what they share is a sense that public opinion about foreign policy is characterized by

order rather than chaos, and that the sources of this order can be derived from within the public

itself.

In response to these optimists is a third school that also finds predictability in public opinion

about foreign a↵airs, but credits it not to the public, but to the elites they listen to. Responding

in particular to event-driven theories of public opinion, this latter camp points out that the mass

public is “rationally ignorant” about politics in general, but especially foreign policy issues, which

are, by definition, foreign, and relatively far removed from most people’s daily lives (Rosenau,

1965), resulting in an important information asymmetry between elites and the public they govern

(Colaresi, 2007; Baum and Groeling, 2010). In the heat of the crisis in Ukraine in early March 2014,

for example, only one in six Americans could correctly locate Ukraine on a map (Dropp, Kertzer,

and Zeitzo↵, 2014). To “learn what they need to know” (Lupia and McCubbins, 2000) and make

political judgments, members of the public thus turn to trusted cue-givers, typically prominent

members of their preferred political party.3 As a result, the balance of public opinion on foreign

policy issues is largely driven in a top-down fashion by the balance of elite opinion (Brody, 1991;

Zaller, 1992; Berinsky, 2007, 2009). Actual events matter on the ground less than what prominent

Democrats and Republicans have to say about them, and when these elites are divided — and the

media environment reports these divisions (Groeling and Baum, 2008; Baum and Groeling, 2009)

— the public will follow suit.

3Cue-taking models of public opinion about foreign policy do not limit themselves exclusively to party leaders ascue-givers — Dropp, Golby, and Feaver (2014) look at the cue-giving e↵ects of military generals, Hayes and Guardino(2011) and Murray (2014) at those of foreign leaders, Thompson (2006); Chapman (2011); Grieco et al. (2011) at theendorsement e↵ects of international institutions, and Pease and Brewer (2008) at that of Oprah Winfrey, but as wediscuss below, all of these cue-givers are su�ciently socially distant from individual members of the public that wecan think of a top-down logic as operating, even if the question of how publics weigh competing cues from multiplecue-givers remains an unanswered question. For an integration of the first two images, see Hu↵ and Schub (n.d.).

4

Thus, although the elite cue-taking school sees public opinion about the use of force as less

stochastic than the early postwar cynics did, their top-down take on the nature of public opinion is

perhaps no less pessimistic. Although proponents of these models take pains to point out that the

public “are not lemmings” (Berinsky, 2007, 975) and that relying on heuristic reasoning is neither

irrational nor inconsistent with fulfilling the requirements of democratic citizenship (Lupia and

McCubbins, 2000), the normative implications of these models are nonetheless somewhat saturnine

compared to their relatively jovian predecessors. If public opinion is driven from the top down, the

public’s ability to constrain their leaders in the manner anticipated by audience cost theory, for

example (Fearon, 1994; Levendusky and Horowitz, 2012), is limited, as the public is simply likely to

swallow whatever their elite cue-givers feed them. As Saunders (2015) argues, if public opinion about

foreign policy is truly as top-down as elite cue-taking theories suggest, “many domestic political

accounts of international relations have gotten the democratic audience wrong”, and IR scholars

should question whether the public belongs in our models of domestic politics at all.

2.1 Going beyond a top-down model

By reminding us that the nature of the information environment matters in the study of public

opinion, elite cue-taking models perform an invaluable service. And yet, there are three reasons why

we may wish to postpone throwing out the public with the bathwater.

First, elite cue-taking models are explicitly about a particular top-down causal mechanism,

rather than a simple correlation, yet many of the tests of top-down models of public opinion in

foreign policy rely on observational data where questions of directionality are di�cult to disentangle:

it could be the case that a correlation between party leaders’ statements and mass opinion is not

due to the public taking cues from party leaders, but from strategic politicians responding to the

wishes of their base; it could also be the case that both elites and attentive members of the public

rely on the same heuristics or anchor on the same values or orientations when processing information

about the world, and thus reach similar opinions simultaneously. If deeply-seated moral values shape

foreign policy preferences, for example (Kertzer et al., 2014), and Democrats and Republicans di↵er

on which moral values are important to them (Graham, Haidt, and Nosek, 2009), elites and masses

5

can polarize in tandem along partisan lines even without the former cueing the latter.

Experiments are better suited to showing cue-taking in action, but evidence here is mixed, such

that even proponents of elite cue theory in IR admit that “the existing literature is fragmented

with contradictory results” (Guisinger and Saunders, 2017, 2). Gelpi (2010) finds that events on the

ground consistently outperform elite cues in an experiment gauging support for the Iraq War, while

Levendusky and Horowitz (2012) find that elite party cues are surprisingly impotent in audience cost

experiments (but see Kertzer and Brutger 2016). Berinsky (2009, 118-123) finds partial support for

an elite cue model in an experiment regarding a hypothetical intervention in South Korea, but notes

that the hyper-polarized environment of the Iraq war — in which participants have already been

pre-treated with elite cues about the wisdom or folly of military interventions before they participate

in the experiment — makes for a harder test of the theory.

Second, the political behavior literature now has a more nuanced view of elite cues than many IR

scholars might realize, calling into question whether ordinary citizens are as easily bullied by the bully

pulpit as a top-down model of public opinion predicts (Edwards, 2003). Enns (2014) finds that elites

largely took cues about mass incarceration from an increasingly punitive mass public, rather than

the other way around, Saeki (2013) finds that legislators are more likely to undergo ideological shifts

in response to their voters than voters are in response to their legislators, Steenbergen, Edwards, and

de Vries (2007) finds that support for European integration is characterized by both top-down and

bottom-up cue-giving, and Messing and Westwood (2014) find that social endorsements outweigh

partisan sources in selective exposure. Similarly, Bullock (2011) demonstrates that when partisan

respondents in experiments are presented with policy information in addition to party cues, the e↵ect

of the former is as least as strong as the latter, showing that even strong partisans do not necessarily

automatically accept what their party leaders say; Boudreau and MacKenzie (2014) also find that

strong partisans are actually more, rather than less, likely to make use of policy information when

espousing judgments. Most relevant for us, both Druckman and Nelson (2003) and Klar (2014) find

that citizens’ conversations with one another can eliminate the e↵ects of elite rhetoric. Opinion

on foreign policy issues may abide by fundamentally di↵erent dynamics than opinion on domestic

ones, of course (though see Holsti and Rosenau 1996; Rathbun 2007), but these findings raise the

6

possibility that the e↵ects of elite partisan cues may be contextually contingent.

Third, the empirical record is filled with anomalies that purely top-down models of public opinion

about foreign policy have trouble explaining. If the public is simply taking cues from elites, there

should not be a “foreign policy disconnect” between the wishes of the former and the preferences of

the latter (Page and Bouton, 2007). Yet although there was relative elite consensus in the lead-up to

the Iraq War — and, in a content analysis of network news coverage in the eight months preceding

the war, Hayes and Guardino (2010, 61) find that “the voices of anti-war groups and opposition

Democrats were barely audible” — there was sizable domestic opposition to the war in a manner

that strictly top-down theories of public opinion have trouble explaining (Hayes and Guardino,

2011), just as they have trouble explaining why public support for torture rose when elite opposition

increased (Mayer and Armor, 2012). Additional evidence comes from outside the United States as

well: Kreps (2010) finds that against elite-driven theories of public opinion, the war in Afghanistan

was extremely unpopular in most of the countries that contributed troops to the mission, despite

the backing of foreign elites.

We suggest that some of these puzzles are perhaps less puzzling if we recognize that citizens do

not simply take cues from distant elites, but also bring their own predispositions to the table, and

can also take cues from one another. Despite the tendency of treating public opinion as the addi-

tive aggregation of individual and independently-administered responses to survey questions, public

opinion has a public quality (Sanders, 1999) stemming from the group context in which individuals

operate. In that sense, scholars of public opinion should not just be looking at micro-foundations,

but at meso-foundations: the social context and network in which citizens are embedded. Out-

side the study of political behavior, constructivist IR scholars have been making similar arguments,

pointing to the importance of the mass public “common sense” as an obstacle to elite hegemony

(e.g. Hopf, 2013). In an innovative study of the 1971 Bangladesh War, for example, Hayes (2012)

shows that Richard Nixon and Henry Kissinger’s attempts to cue the public to think of India as a

threat ultimately failed because the public saw India as a fellow democracy, and thus as inherently

nonthreatening. Public opinion proved to be uncueable. Many of our theories of norms in IR simi-

larly advance “bottom-up” models where societal groups are leading political elites, rather than the

7

other way around (e.g. Checkel, 1997; Fanis, 2011).

There are at least three reasons why scholars of public opinion in foreign a↵airs should think

seriously about meso-foundations and group context. First, groups and social networks are an

important source of information (Mutz, 1998). Although the prevailing information-based models

in American public opinion about foreign policy are purely elite-driven, information travels laterally

as well as top-down, and perceptions of the attitudes of our peers a↵ects both what we think,

and how certainly we think it (Visser and Mirable, 2004; Clarkson et al., 2013). If the power

of heuristic processing is a function of not only receiving information but also choosing whether

to accept it (Zaller, 1992), information from proximate peers is likely to amplify or dampen the

resonance of messages from distant elites, particularly given that Americans’ trust in government is

consistently lower than their trust in one another (Keele, 2007). Second, groups and social networks

are important sources of social influence (Milgram, 1974; Sinclair, 2012). Even when groups do

not explicitly coerce, the mere presence of a majority induces pressures towards conformity (Asch,

1951; Stein, 2013), particularly given the importance of group membership in defining who people

are and how they behave (Brewer and Brown, 1998; Smith, Seger, and Mackie, 2007). Third, and

relatedly, a rich body of research throughout the social sciences has documented that people behave

di↵erently in groups than they do as individuals (Hackman and Katz, 2010); late 19th- and early-

20th century scholars preoccupied with the “folly of the crowd” saw groups as more emotional and

impulsive than the individuals who comprise them (e.g. Le Bon, 1896), while an opposite body

of literature suggests that individual-level errors and irrationalities cancel each other out in groups

(Druckman, 2004), and a large literature on group polarization (Myers and Lamm, 1976; Friedkin,

1999) documents the extent to which groups adopt more extreme positions after deliberating than

the median stance amongst group members before deliberation takes place. Yet political scientists

have yet to appreciate how these meso-level e↵ects might play a role in public opinion about foreign

a↵airs.4

There are multiple pathways through which group cues could influence individuals. First, groups

can influence political behavior by explicitly or implicitly pushing social conformity. Second, groups

4Among the few exceptions we are aware of: Radziszewski (2013), which uses observational data to examine thee↵ects of discussion networks on Polish support for the Iraq War, and Todorov and Mandisodza (2004), which exploreshow second-order beliefs about American public opinion shape first-order foreign policy preferences.

8

can convey credible new information to group members about how other individuals view specific

policies; they thus let group members get a second opinion. Disentangling these e↵ects observation-

ally is very di�cult, so we turn to a series of five survey experiments to isolate the informational

e↵ect of group cues on support for war and peace in the absence of social pressure of conformity. In

the fourth and fifth experiments, we further test whether it is the information present in the social

cues, or the similarity of the cuegiver that drives the e↵ects of social cues. We believe in doing so,

we follow Mendelberg’s (2005) exhortation to bring “the group back into political psychology.”

To explore these meso-foundations of public opinion about foreign a↵airs, we designed five survey

experiments, fielded in three di↵erent studies. The first two experiments were embedded in a survey

fielded by Survey Sampling International (SSI) on a national sample of 1,035 registered voters in

the summer of 2014.5 The third experiment was administered to 1,446 American adults on Amazon

Mechanical Turk (MTurk) in the autumn of 2014. The fourth and fifth experiments were embedded

in a survey administered to 1,997 American adults on MTurk in the autumn of 2016. We describe

each in turn.

3 Experiments 1-2

3.1 Methods

At the beginning of the first study, participants completed a short questionnaire measuring their

militant assertiveness and internationalism — two key foreign policy orientations from the foreign

policy public opinion literature (e.g. Herrmann, Tetlock, and Visser, 1999), as well as a standard

battery of demographic and partisan characteristics. After subjects completed the opening ques-

tionnaire, they were presented with two foreign policy experiments presented in random order. In

each experiment, we presented participants with a fictional newspaper article — presented as real

— in which policymakers in Washington were debating a salient national security issue: a military

5SSI panels employ an opt-in recruitment method, after which panel participants are randomly selected for surveyinvitations, using population targets rather than quotas to produce a nationally diverse sample of registered voters.The experiment was embedded in a larger, unrelated survey, and participants were unaware of the content of thesurvey when they chose to participate. Because of the recruitment technique, the sample is nationally diverse, butnot a national probability sample; for other examples of recent political science research employing SSI samples, seeMalhotra and Margalit (2010); Kertzer and Brutger (2016).

9

pivot to Asia in response to increased threats from a rising China (China), and the deployment of

special forces units to combat terrorists in the Middle East (Terrorism).6 Examples of the stimulus

materials are shown in Appendix §1.

In each article, we manipulated two di↵erent factors. First, each article included an quote

from a member of Congress endorsing the policy proposal. For each participant, we randomly

assigned whether the endorsement in the article came from a Democrat (Democrat Endorsement) or

a Republican (Republican Endorsement). Since the persuasion literature emphasizes the importance

of source credibility (Lupia and McCubbins, 1998; Druckman, 2001; Pornpitakpan, 2004), in both

cases the speaker is described as a veteran member of Congress with established foreign policy

expertise. Second, we manipulated the emotionality of the argument put forth by the member of

Congress for the use of force, such that the Hot Cognition treatment argument was based on “gut”

feelings, while the Cold Cognition was based on “cool, cold logic.”7

After reading each article, participants were assigned into one of three groups: a Control group,

a Group Endorse condition, and a Group Oppose condition (see Appendix §1 for examples). In

both the Group Endorse and Group Oppose condition, participants were presented with a set of

results putatively illustrating the preferences of previous survey respondents, and told that “The

graph below shows the responses of people who have previously taken the survey.” Those in the

Group Endorse condition were told: “Those who answered the earlier questions on the survey like

you strongly supported” the policy proposal, and shown a bar graph where 74% of respondents were

in favor of the policy, whereas those in the Group Oppose condition were told that “Those who

answered the earlier questions on the survey like you strongly opposed” the policy proposal, and

shown a bar graph where 74% of respondents were opposed to the policy.

The nature of our social cue treatment builds upon a growing body of research which finds that

peer networks influence political behavior (Sinclair, 2012; Bond et al., 2012). Following Mann and

Sinclair (2013), we manipulate social cues using the language “like you” rather than selecting a

pre-defined reference group. In this way, the treatment lets participants define their own reference

6For examples of the importance of these issues on the contemporary American foreign policy agenda, see Ross(2012) and testimony by Seth G. Jones of the RAND Corporation on “Counterterrorism and the Role of SpecialOperations Forces” before the House Foreign A↵airs Committee, Subcommittee on Terrorism, Non-Proliferation, andTrade on April 8, 2014 http://www.rand.org/content/dam/rand/pubs/testimonies/CT400/CT408/RAND_CT408.pdf.

7See Appendix §1 for a broader discussion.

10

group, rather than assuming participants identify with other members of groups defined by particular

descriptive characteristics.8

1) Opening questionnaire

Demographics, partisanship, foreign policy orientations

Democratic endorsement

Republican endorsement

x

Hot cognition

Cold cognition

x

2) News article (either China or Terrorism)

Elite cue treatment Type of appeal

3) Group cue

Group endorse

Group oppose

Control

4) Main DVs

Willingness to use force, certainty, level of threat,

success

Group cue treatment

Repeat steps 2-4 for the other scenario (e.g. either China or Terrorism), in crossover experimental design

Figure 1: Study 1 Design: Experiments 1-2

Following the treatments, participants then answered questions related to their support for

using force in each scenario.9 Participants then proceeded to the next experiment (either Terrorism

or China), depending on which experiment they were randomly assigned to receive first. Thus,

Experiments 1-2 feature a modified crossover design. Participants who first received the China

experiment and the Emotional Appeal, Democratic Endorsement, and Group Endorse conditions,

for example, then received the Terrorism experiment, Logical Appeal, Republican Endorsement and

Group Oppose treatments.10 We summarize the study design in Figure 1, and present summary

statistics, sample characteristics, and randomization checks in Appendix §2.1.

3.2 Results

Do group-level cues influence foreign policy choices, and how do they compare to elite-level endorse-

ments? In Table 1 we explore the e↵ects of our treatments on support for the use of force. Across

8Unlike Mann and Sinclair (2013), the “like you” treatment here is in reference to how the other participantsanswered previous questions on the survey – the demographic questions and foreign policy orientation questions.Thus, the “like you” here deliberately refers both to people of similar demographic characteristics and to peoplewith similar foreign policy attitudes. One could imagine treatments in which we said “people of the same age anddemographic group as you”, but this potential treatment would involve imposing groups on subjects, rather thanletting participants define it themselves. See Experiments 4-5 for a modified version of the social cue treatment.

9We also measured the certainty of their opinions, the perceived likelihood of success of using force , and how muchof a threat they thought that the target of the policy shift (terrorism or China’s military) posed to US interests.

10In all of the results presented here, we control for order e↵ects. Those who received the group Control conditionin one experiment in Study 1 also had it for the other.

11

Table 1: Treatment E↵ects on Use of Force (OLS)

Dependent Variable: Support for Armed Force

China Terrorism China Terrorism China Terrorism

(1) (2) (3) (4) (5) (6)

Emotional Appeal 0.018 �0.009 0.010 �0.001 0.013 �0.003(0.017) (0.018) (0.015) (0.016) (0.015) (0.016)

Democrat Endorse �0.001 �0.035 �0.006 �0.029 �0.006 �0.029(0.017) (0.018) (0.015) (0.016) (0.015) (0.016)

Group Endorse 0.068⇤⇤⇤ 0.067⇤⇤⇤ 0.071⇤⇤⇤ 0.056⇤⇤⇤ 0.070⇤⇤⇤ 0.054⇤⇤⇤

(0.020) (0.022) (0.018) (0.019) (0.018) (0.019)

Group Oppose �0.055⇤⇤⇤ �0.061⇤⇤⇤ �0.065⇤⇤⇤ �0.060⇤⇤⇤ �0.066⇤⇤⇤ �0.064⇤⇤⇤

(0.020) (0.022) (0.018) (0.019) (0.018) (0.019)

Party ID 0.008⇤⇤ 0.013⇤⇤⇤ 0.009⇤⇤ 0.012⇤⇤⇤

(0.004) (0.004) (0.004) (0.004)

Militant Assertiveness 0.517⇤⇤⇤ 0.551⇤⇤⇤ 0.525⇤⇤⇤ 0.564⇤⇤⇤

(0.039) (0.042) (0.040) (0.042)

Internationalism 0.142⇤⇤⇤ 0.239⇤⇤⇤ 0.137⇤⇤⇤ 0.222⇤⇤⇤

(0.043) (0.045) (0.044) (0.046)

Controls X XN 1,035 1,021 1,034 1,020 1,031 1,017Adjusted R2 0.036 0.032 0.222 0.246 0.227 0.261⇤⇤p < .05; ⇤⇤⇤p < .01All regressions are OLS and control for the randomly assigned order of the experiments (China or Terrorism).

Controls include Male, Age, Education, Income, and White.

12

both experiments, we find that the group treatments strongly influence participants’ choices: par-

ticipants in the Group Endorse condition are significantly more likely to favor using force than those

in the Control condition, while subjects in the Group Oppose condition are significantly less likely

to support using force than those in the Control Condition. In comparison, our other treatments

have relatively weak and nonsignificant e↵ects: the e↵ect of a Democratic endorsement (Democratic

Endorse) reduces support for intervention, but only for the Terrorism experiment, and its negative

direction is noteworthy given that the literature on “party brand” and “against type” e↵ects would

predict that military missions would be more popular when endorsed by a Democrat than by a

Republican (e.g. Schultz, 2005; Trager and Vavreck, 2011). Additionally, the magnitude of the elite

cue is smaller and less significant than either of the group cues. Thus, we find strong support for

our claim that group cues are important factors in shaping foreign policy attitudes.

Participant-level characteristics matter too. In general, Republicans are significantly more likely

to favor intervention than Democrats across both experiments, but the substantive e↵ect of partisan-

ship is dwarfed by that of our two foreign policy orientations: hawks high in militant assertiveness

are far more likely to favor both pivoting to Asia and using special forces units to engage in coun-

terterrorism operations, as are internationalists who generally favor the US playing an active role

abroad. In this sense, these first set of results remind us that rather than just looking at the elite

partisan cues floating above citizens’ heads, we should also be looking at the core dispositions sitting

inside them, as well as the presence or absence of social cues from individuals’ peers. Substantively,

our results point to the under-explored e↵ects of social cues on support for the use of force. Rather

than cues only flowing from the top-down and swaying malleable voters about foreign policy, we

show that i) voters’ support for the use of force is consistent with their pre-existing value orienta-

tions (Militant Assertiveness and Internationalism), and ii) that voters are likely to take cues from

those who they feel share their own values and points of view.

Ultimately, though, elite cue theory predicts not just that people on average will respond to

statements di↵erently based on the political party of the cue-giver, but also that the e↵ect of the

cue depends on the partisanship of the recipient: participants who identify as Republicans should

respond to a Republican cue-giver di↵erently than participants who identify as Democrats. Yet when

13

we search for evidence of these heterogeneous treatment e↵ects on Table 2, we come away empty-

handed. Our results thus reconfirm our findings from Table 1 about the importance of group-level

cues in shaping public support.

Table 2: Is there a moderating e↵ect of party ID on support for the use of force? (OLS)

Dependent Variable: Support for Armed Force

China Terrorism China Terrorism China Terrorism

(1) (2) (3) (4) (5) (6)

Emotional Appeal 0.015 �0.006 0.010 �0.001 0.013 �0.003(0.016) (0.017) (0.015) (0.016) (0.015) (0.016)

Party ID 0.029⇤⇤⇤ 0.031⇤⇤⇤ 0.010⇤⇤ 0.020⇤⇤⇤ 0.011⇤⇤ 0.018⇤⇤⇤

(0.005) (0.006) (0.005) (0.005) (0.005) (0.005)

Democrat Endorse 0.039 �0.013 0.010 0.022 0.008 0.018(0.033) (0.035) (0.030) (0.032) (0.030) (0.032)

Party ID ⇥ Dem. Endorse �0.010 �0.006 �0.004 �0.013 �0.004 �0.012(0.007) (0.008) (0.007) (0.007) (0.007) (0.007)

Group Endorse 0.067⇤⇤⇤ 0.069⇤⇤⇤ 0.071⇤⇤⇤ 0.056⇤⇤⇤ 0.070⇤⇤⇤ 0.054⇤⇤⇤

(0.020) (0.021) (0.018) (0.019) (0.018) (0.019)

Group Oppose �0.053⇤⇤⇤ �0.063⇤⇤⇤ �0.065⇤⇤⇤ �0.060⇤⇤⇤ �0.066⇤⇤⇤ �0.064⇤⇤⇤

(0.020) (0.021) (0.018) (0.019) (0.018) (0.019)

Militant Assertiveness 0.516⇤⇤⇤ 0.556⇤⇤⇤ 0.524⇤⇤⇤ 0.569⇤⇤⇤

(0.039) (0.042) (0.040) (0.042)

Internationalism 0.142⇤⇤⇤ 0.241⇤⇤⇤ 0.137⇤⇤⇤ 0.224⇤⇤⇤

(0.043) (0.045) (0.044) (0.046)

Controls X XN 1,034 1,020 1,034 1,020 1,031 1,017Adjusted R2 0.074 0.078 0.222 0.248 0.226 0.262⇤⇤p < .05; ⇤⇤⇤p < .01All regressions are OLS and control for the randomly assigned order of the experiments (China or Terrorism first).

Controls include Male, Age, Education, Income, and White.

3.2.1 Were our elite cues overwhelmed by group cues?

An alternative explanation for the absence of evidence in favor of elite cues in Experiments 1-2 could

be that the group-level treatments are relatively strong, while the elite cue treatments are relatively

14

weak. We thus conducted two additional tests: first, testing for elite cue-taking only looking at the

treatment e↵ects among those participants who correctly answered the manipulation check for the

elite cue treatment, and second, testing for elite cue-taking by subsetting the data and restricting

our analysis solely to those participants who were in the group Control condition and thus did not

receive any group cues.

In Table 6 in Appendix §2.1, when we restrict the results to those who correctly pass the ma-

nipulation check (i.e. those who correctly identified the anonymous endorser in the scenario as a

Democrat or Republican), our core results remain unchanged: social cues (Group Endorse and Group

Oppose) and value orientation (Militant Assertiveness and Internationalism) influence voters, but

elite cues do not. In Table 5 in Appendix §2.1, we explore whether perhaps the meso-level treat-

ments are ‘swamping’ the e↵ects of elite endorsements, restricting our analysis to the group Control

condition (i.e. those who received no group cues in either the Terrorism or China experiments). We

find inconsistent results for the e↵ect of the Democratic Endorse condition — which now reduces

support for a pivot to Asia, rather than terrorism, although the e↵ect remains statistically and

substantively weak, and the partisanship ⇥ elite cue interaction remains nonsignificant.

Alternately, another possible explanation for the lack of results for our elite cues are that par-

tisanship moderates the e↵ect of the elite and group-level cues, whereupon our relatively simple

models above fail to capture the complex interplay between partisanship and elite and group-level

cues. We explore this question in Table 7 and Figure 6 in Appendix §2.1, which look at a richer

set of two-and three-way interactions between social cues, partisanship, and elite cues. The analysis

confirms our core results from Table 1. Elite cues and partisanship have weak and inconsistent re-

sults, and do not appear to moderate the much stronger and robust e↵ect of social cues on support

for force. Finally, in supplementary analyses in Appendix §2.1.1, we explore the e↵ects of elite and

social cues on certainty, threat perception, and perceived success, finding that group endorsements

systematically outweigh elite ones.

15

4 Experiment 3

One potential explanation of the findings of the previous study was that the social cue treatments

were simply stronger than the partisan elite cues. Both the elite and social cues were anonymous,

but the elite cue consisted of a single individual, whereas the social cue consisted of a group. In

this sense, the failure of an an endorsement by an anonymous, veteran Democratic or Republican

lawmaker to move respondents is notable, but there are other ways of thinking about elite partisan

cues as well. We thus conducted a third experiment on 1,446 American adults recruited in September

2014 from Amazon Mechanical Turk.

4.1 Method

1) Opening

questionnaire

Demographics,

partisanship,

foreign policy

orientations

Control

Dems support,

Reps oppose

Reps support,

Dems oppose

Elite

consensus

x

2) News article

(China)

Elite cue treatment

3) Group cue

Group endorse

Group oppose

Control

4) Main DVs

Willingness to use

force, certainty,

level of threat,

success

Group cue treatment

Figure 2: Study 2 Design: Experiment 3

The experiment mirrored its predecessor with two principal changes, based o↵ of the rising China

experiment from the previous study. First, given the weak and inconsistent e↵ects of the emotional

appeal in Experiments 1-2, we held the type of message constant in Experiment 3 and only used

the cold cognition message. Second, rather than manipulating elite partisan cues by manipulating

which party endorsed an aggressive foreign policy toward China, we manipulated the position of

both parties. A quarter of the participants were told that Democrats in Congress supported an

16

aggressive foreign policy toward China while Republicans in Congress opposed it; another quarter

were told that Republicans in Congress supported an aggressive foreign policy while Democrats

in Congress opposed it, and a final quarter were told that both Democrats and Republicans in

Congress supported the aggressive foreign policy. In this sense, the first two conditions depict a

polarized partisan environment, while the third displays elite consensus, which if elite cue theory is

correct, should display a “mainstreaming” e↵ect (Zaller, 1992). Finally, a quarter of participants

were in a control group, and were not given any information about elite endorsements, to provide a

baseline with which to compare the e↵ects of the other elite cues.11 Thus, as illustrated in Figure

2, the study design yields a 4 (Elite Cues) ⇥ 3 (Social Cues) fully-crossed factorial experiment.

4.2 Results

Table 3: Study 2: Treatment E↵ects

Support for Armed Force in China

(1) (2) (3)

Dem Support �0.029 �0.029 �0.028(0.020) (0.016) (0.016)

Repub Support �0.028 �0.042⇤⇤ �0.042⇤⇤

(0.020) (0.016) (0.017)Elite Consensus 0.037 0.033⇤⇤ 0.033⇤⇤

(0.020) (0.016) (0.016)Group Endorse 0.050⇤⇤⇤ 0.042⇤⇤⇤ 0.043⇤⇤⇤

(0.017) (0.014) (0.014)Group Oppose �0.078⇤⇤⇤ �0.065⇤⇤⇤ �0.065⇤⇤⇤

(0.017) (0.014) (0.014)Militant Assertiveness 0.665⇤⇤⇤ 0.668⇤⇤⇤

(0.028) (0.029)Internationalism 0.227⇤⇤⇤ 0.223⇤⇤⇤

(0.033) (0.033)Party ID 0.062⇤⇤ 0.064⇤⇤⇤

(0.024) (0.024)Controls XN 1,445 1,445 1,444Adjusted R2 0.043 0.363 0.362⇤⇤p < .05; ⇤⇤⇤p < .01All regressions are OLS and controls include Male, Age, and Education.

In Table 3 we present treatment e↵ects from Experiment 3. The results reinforce the findings

11The exact wording of the elite cues for Experiment 3 are presented in Appendix §1.

17

from Experiments 1-2 that social cues strongly influence support for the use of force. The Group

Endorse treatment significantly increases support for using force, and the Group Oppose treatments

significantly reduces support. Thus, even in the presence of elite cues, social cues exert a strong and

significant e↵ect on foreign policy attitudes.

Partisanship (Party ID) also strongly influences attitudes towards interventions, with Republi-

cans more in favor of shifting military resources towards China. Finally, as in the previous study, we

note the substantively large and statistically significant e↵ects of individuals’ foreign policy orien-

tations (Militant Assertiveness and Internationalism), which dwarf that of elite cues. These results

reinforce that ordinary citizens have stable foreign policy predispositions that strongly shape their

attitudes independent of the cues they receive from elites or other members of the public. In Ap-

pendix §2.2, we present a variety of robustness checks, showing that our results do not di↵er when

we subset among participants who passed the manipulation check, that the e↵ect of our cues are

not conditional on respondents’ partisanship, and so on.

5 Experiments 4-5

Experiments 1-3 show individuals are more likely to take cues about foreign policy from each other

than from political elites. Yet foreign policy is about more than just security; it is thus worth testing

whether we find similar patterns on economic issues. Additionally, Experiments 1-3 borrow from

Mann and Sinclair (2013) in utilizing social cues from individuals who answered previous survey

questions like the respondent. Although this avoids the problem of selecting a pre-defined reference

group for participants, it raises a number of questions, including about the mechanisms driving the

group cue: do social cues need to be from individuals “like” the respondent in order to shape foreign

policy views, or does simply knowing the views of other respondents more generally have the same

e↵ect? Are the power of social cues about the pull of homophily, or the appeal of getting a second

opinion? We thus fielded two additional experiments, on 1,997 American adults recruited via Amazon

Mechanical Turk, in September 2016. Experiments 4-5 mirrored their predecessors, with three

notable di↵erences. First, one of the experiments is about an international political economy (IPE)

issue: whether US citizens and corporations should continue to be subject to investor-state dispute

18

settlement from the International Centre for Settlement of Investment Disputes (ICSID). Second,

to disentangle the e↵ects of the social treatments, in addition to the “like you” treatments from

Experiments 1-3, we also include a revised version of the group endorse and group oppose treatments

that omit the “like you” language, simply reporting the views of generic survey participants. We

can thus compare the e↵ect of each type of social cue to one another to gain further leverage on the

mechanism responsible for the group cue e↵ects. Finally, since the two elite divided treatments in

Experiment 3 did not significantly di↵er from one another, we save statistical power by retaining

only one of them, a treatment in which Republicans support a policy, and Democrats oppose. Each

experiment is thus a 3 (Elite Cues) x 5 (Social Cues) fully-crossed factorial, illustrated in Figure 3.

1) Opening

questionnaire

Demographics,

partisanship,

foreign policy

orientations

Control

Elites divided

Elite

consensus

x

2) News article

(either China or ICSID)

Elite cue treatment

3) Group cue

Group endorse (1)

Group oppose (1)

Group oppose (2)

Control

4) Main DVs

Willingness to use

force, certainty

Group cue treatment

Group endorse (2)

Repeat steps 2-4 for the other scenario (e.g. either China or ICSID), in crossover design.Elites divided condition = Republicans support, Democrats oppose.

Figure 3: Study 3 Design: Experiments 4-5

We begin by simply comparing the “like you” group cues with their generic counterparts: as

we show in Appendix §2.3.1, there are no significant di↵erences between the “like you” coe�cients

and the generic coe�cients, a set of Davidson-MacKinnon J tests fails to find evidence that models

di↵erentiating each type of group cue significantly di↵ers from models that pool them together,

and a set of Wilcoxon rank-sum tests fails to find evidence that the distribution of the dependent

variable di↵ers across each type of group cue, further confirmed by visual inspection of the density

distributions. Since it appears that the social cues are not being driven by the “like you” language,

for simplicity we pool each type of group cue together for our subsequent analysis, presented in

19

Table 4.12

Table 4: Experiments 4-5 Results:

China ICSID

(1) (2) (3) (4) (5) (6)

Elite Divided �0.030 �0.026⇤⇤ �0.026 �0.043⇤⇤⇤ �0.043⇤⇤⇤ �0.042⇤⇤⇤

(0.016) (0.013) (0.013) (0.014) (0.014) (0.014)Elite Consensus 0.064⇤⇤⇤ 0.072⇤⇤⇤ 0.072⇤⇤⇤ 0.047⇤⇤⇤ 0.046⇤⇤⇤ 0.046⇤⇤⇤

(0.016) (0.013) (0.013) (0.014) (0.014) (0.014)Group Endorse 0.017 0.008 0.008 0.043⇤⇤⇤ 0.043⇤⇤⇤ 0.043⇤⇤⇤

(0.018) (0.015) (0.015) (0.016) (0.015) (0.015)Group Oppose �0.091⇤⇤⇤ �0.094⇤⇤⇤ �0.094⇤⇤⇤ �0.069⇤⇤⇤ �0.073⇤⇤⇤ �0.073⇤⇤⇤

(0.018) (0.015) (0.015) (0.016) (0.015) (0.015)Militant Assertiveness 0.655⇤⇤⇤ 0.651⇤⇤⇤ 0.0003 0.004

(0.028) (0.028) (0.029) (0.029)Internationalism 0.142⇤⇤⇤ 0.141⇤⇤⇤ 0.286⇤⇤⇤ 0.290⇤⇤⇤

(0.030) (0.031) (0.031) (0.031)Party ID 0.065⇤⇤⇤ 0.060⇤⇤ �0.016 �0.016

(0.023) (0.023) (0.024) (0.024)Controls X XN 1,997 1,997 1,994 1,997 1,997 1,994Adjusted R2 0.046 0.305 0.306 0.058 0.104 0.103⇤⇤p < .05; ⇤⇤⇤p < .01All regressions are OLS and control for the randomly assigned order of the experiments (China or ICSID first).

Controls include Male, Age, and Education.

The substantive e↵ects of the social cues and elite treatments presented in Table 4 provide several

important findings. First, compared to the previous experiments, we find stronger evidence in favor

of elite cues – particularly Elite Consensus, which bolsters support in both the China and ICSID

experiments. One reason may be because the study was fielded during the penultimate month of a

highly charged Presidential election campaign; supplementary analyses in Appendix §2.3.2 show that

our respondents displayed significantly higher baseline levels of partisan polarization here than in the

previous experiment. Second, despite the timing of the survey, as before, our largest e↵ects belong

to social cues, with the Group Oppose treatment strongly decreasing support in both the China

and ICSID experiments; the Group Endorse treatment also significantly raises support, but only

in the ICSID experiment. Third, similar to the previous experiments, foreign policy orientations

play statistically and substantively significant roles, although sensibly, military assertiveness is a

significant predictor of attitudes towards deploying naval forces in East Asia, but not on investor-

12See Appendix §2.3.1 for results disaggregated by type of social cue treatment.

20

state dispute mechanisms. In sum, our findings in Study 3 suggest that the e↵ect of social cues

are not domain specific. Social cues matter both for shaping the public’s attitudes towards security

policy (China), but also in IPE (ICSID), and their e↵ects do not seem to depend on them coming

from individuals who specifically share the same views as the respondent. Finally, supplementary

analyses in Appendix §2.3.2 o↵er further evidence in favor of our theoretical mechanisms, showing

that respondents who have less trust in government are significantly less sensitive to elite cues in the

China experiment, while Trump supporters are significantly less receptive to elite cues than Clinton

supporters are more generally.

Figure 4: Aggregating results across all 5 experiments

Elite cues

Social cues

Military assertiveness

Internationalism

Party ID

Experiment 5Experiment 4Experiment 3Experiment 2Experiment 1

Experiment 5Experiment 4Experiment 3Experiment 2Experiment 1

Experiment 5Experiment 4Experiment 3Experiment 2Experiment 1

Experiment 5Experiment 4Experiment 3Experiment 2Experiment 1

Experiment 5Experiment 4Experiment 3Experiment 2Experiment 1

0.0 0.2 0.4 0.6Effect size

Results are coe�cient estimates from regression models, with 95% confidence intervals calculated using B = 1500bootstraps; in addition to the treatments and orientations, the models also include demographic controls. To

facilitate comparability across studies, the plot presents the largest contrasts for each treatment. The results showthat social cues consistently exert a significant e↵ect (averaging +11.5%), while the e↵ect of elite cues is inconsistent

(averaging +4.2%), and foreign policy orientations generally outweigh party identification.

21

6 Conclusion

Public opinion is increasingly playing prominent role in IR scholarship: from theories of crisis bar-

gaining that abandon unitary actor assumptions and explicitly carve out a major role for domestic

publics (Fearon, 1994; Schultz, 2001; Slantchev, 2006; Tarar and Leventoglu, 2009), to the rise

of individual-level experiments exploring micro-foundations of public opinion towards world a↵airs

(Herrmann, Tetlock, and Visser, 1999; Tomz, 2007; Kertzer and McGraw, 2012; Wallace, 2013; Ren-

shon, 2015; Walsh, 2015). This prominence is all the more striking given that it was only 25 years

ago that political scientists were still asking whether leaders “waltz before a blind audience” on for-

eign a↵airs (Aldrich, Sullivan, and Borgida, 1989), and thus whether IR scholars might be justified

in bracketing the public altogether. Yet if elite cue-taking theories of public opinion are correct, and

the public passively digests whatever their leaders tell them, can publics constrain those that govern

them? If public opinion about foreign a↵airs is really just driven from the top down, should we even

bother looking for micro-foundations for foreign policy in public opinion at all?

We argued here that reports of the public’s passivity are somewhat exaggerated. Employing five

original survey experiments (the results of which are summarized in Figure 4), we found that the

e↵ect of elite cues was inconsistent, but that social cues exert important e↵ects, as do individuals’

general predispositions towards international a↵airs. We urge caution in dwelling on the substan-

tively larger e↵ect sizes for foreign policy orientations than cues here, since the orientations are real

traits our participants carry around with them, whereas the cues are one-shot treatments artificially

manipulated in an experimental context. Nonetheless, the fact that individuals do carry substan-

tively meaningful orientations towards foreign a↵airs around in their heads with them is precisely

what elite cue theory overlooks; our findings thus show that rather than simply being shaped from

the top down, public opinion is a function both of individuals’ social context, and their preexisting

attitudes towards the kind of role America should play in the world.13 Studying public opinion about

foreign a↵airs thus involves both micro- and meso-foundations. Our claim is not that elite cues are

irrelevant, but rather, that they only tell part of the story. In a sense, then, the results also remind

us what public opinion polls (and by extension, many of the survey experiments in IR) are missing:

13This echoes similar findings from American political behavior (e.g. Lewis-Beck, Helmut Norpoth and, and Weis-berg, 2009).

22

the public quality of public opinion (Sanders, 1999). Survey experiments in IR, like in political

science more generally, treat public opinion as the aggregation of individual surveys administered

in isolation. Methodologically, this isolation is crucial, since non-interference between units lets us

cleanly estimate causal e↵ects, but it also misses the social, deliberative dynamics that characterize

opinion formation in the wild. Experimental research able to bridge this gap in a naturalistic way

while also preserving our abilities to make causal inferences will move us considerably forward.

Although we believe our experimental results contribute to our understanding of the dynamics of

public opinion about foreign a↵airs, they are also open to a number of potential critiques suggesting

directions for future research. It could be that the e↵ects of our one-sided social cues are stronger in

the experiments than in the real world, where individuals are often in heterogeneous social contexts

(Klar, 2014). We believe this concern is overstated: given the presence of homophily in many social

networks, confirmation biases in information processing, and false consensus e↵ects (McPherson,

Smith-Lovin, and Cook, 2001; Nickerson, 1999; Krueger and Clement, 1994), we do not consider the

distribution of support in our treatments to be unrealistic. Nonetheless, future research should ex-

amine how mixed or competing social cues shape foreign policy preferences, whether people discount

cues from certain members of their social networks, as well as pinpointing the precise mechanisms

through which these cues exert their e↵ects.

A related concern could be that experimentally showing that individuals take cues from their

social context is di↵erent from showing that people take cues from their social networks in the

real world. In this regard, though, we should note that experimental methods have a clear advan-

tage compared to observational studies when it comes to testing the e↵ects of social cues, since

social networks are likely to confound the e↵ect of group cues with homophily. By showing that

experimentally-assigned group cues exhibit strong e↵ects, we provide strong evidence that social

cues play an important role in attitude formation.

We conclude with two broader implications of our findings. First, our results suggest that

people are perhaps more resistant to elite manipulation than some of the more pessimistic elite-

driven models of public opinion suggest. Indeed, although it may seem unsurprising to note that

general attitudes towards war and peace shape policy responses in specific instances, the fact that

23

individuals have these stable predispositions are what cynics like Almond and Lippmann were arguing

against. At the same time, however, if the inconsistent e↵ects of elite cues are normatively desirable,

the significant e↵ects of the group endorsement and opposition treatments show that citizens are

not entirely immune to social pressures. These social responses are particularly worth studying in

the age of new media, where both search engines like Google and social networks like Facebook

rely on complex algorithms to show users what they think they want to see, producing alternative

information environments whose implications for foreign policy opinion are not yet fully appreciated

(Bond et al., 2012; Zeitzo↵, Kelly, and Lotan, 2015). Our findings thus suggest that if we are truly

concerned about “manufacturing consent,” we should be worried less about the classic top-down

Chomskyite model where the media uncritically parrots what elites have to say, and more about

manipulation through fellow citizens: Rothschild and Malhotra (2014) show public opinion polls

can become self-fulfilling prophecies, while King, Pan, and Roberts (2016) suggest that the Chinese

government fabricates half a billion social media posts a year precisely because it understands the

power of social cues.

Finally, IR scholars have rightly begun to gather empirical evidence at the micro-level to test the

mechanisms that make our theories work (Kertzer, 2017). We would argue that our results should

encourage IR scholars to think seriously and systematically about meso-foundations as well. It is

striking, for example, that one of the central phenomena of interest for public opinion scholars of

foreign policy — the rally around the flag e↵ect — is inherently a collective phenomenon, but which

tends to be studied in an atomistic fashion. Future work in public opinion towards foreign policy

should therefore explore the broader group contexts in which individuals are embedded.

24

References

ABC News. 2014. “‘This Week’ Transcript: Sec. John Kerry and PM Benjamin Netanyahu.” http://abcnews.go.com/ThisWeek/week-transcript-sec-john-kerry-pm-benjamin-netanyahu/story?id=24632816&page=3.

Aldrich, John H., John L. Sullivan, and Eugene Borgida. 1989. “Foreign A↵airs and Issue Voting: DoPresidential Candidates ”Waltz Before A Blind Audience?”.” American Political Science Review83 (1): 123-141.

Almond, Gabriel A. 1950. The American People and Foreign Policy. New York: Harcourt, Braceand Company.

Asch, Solomon E. 1951. “E↵ects of group pressure on the modification and distortion of judgments.”In Groups, leadership and men, ed. H. Guetzkow. Pittsburgh: Carnegie Press.

Baum, Matthew A., and Tim Groeling. 2009. “Shot by the Messenger: Partisan Cues and PublicOpinion Regarding National Security and War.” Political Behavior 31 (2): 157-186.

Baum, Matthew A., and Tim Groeling. 2010. “Reality Asserts Itself: Public Opinion on Iraq andthe Elasticity of Reality.” International Organization 64 (3): 443-479.

Berinsky, Adam J. 2007. “Assuming the Costs of War: Events, Elites, and American Public Supportfor Military Conflict.” Journal of Politics 69 (4): 975-97.

Berinsky, Adam J. 2009. In Time of War: Understanding American Public Opinion from WorldWar II to Iraq. Chicago, IL: University of Chicago Press.

Bond, Robert M, Christopher J Fariss, Jason J Jones, Adam DI Kramer, Cameron Marlow, Jaime ESettle, and James H Fowler. 2012. “A 61-million-person experiment in social influence and politicalmobilization.” Nature 489 (7415): 295–298.

Boudreau, Cheryl, and Scott A. MacKenzie. 2014. “Informing the Electorate? How Party Cuesand Policy Information A↵ect Public Opinion about Initiatives.” American Journal of PoliticalScience 58 (1): 48-62.

Brewer, Marilynn B., and Rupert J. Brown. 1998. “Intergroup Relations.” In The Handbook of SocialPsychology, ed. Daniel T. Gilbert, Susan T. Fiske, and Gardner Lindzey. New York: McGraw-Hill.

Brody, Richard A. 1991. Assessing the President: The Media, Elite Opinion, and Public Support.Stanford: Stanford University Press.

Bullock, John G. 2011. “Elite Influence on Public Opinion in an Informed Electorate.” AmericanPolitical Science Review 105 (3): 496-515.

Chapman, Terrence L. 2011. Securing approval: domestic politics and multilateral authorization forwar. Chicago, IL: University of Chicago Press.

Chaudoin, Stephen, Helen V. Milner, and Dustin H. Tingley. 2010. “The Center Still Holds: LiberalInternationalism Survives.” International Security 35 (1): 75-94.

Checkel, Je↵rey T. 1997. “International Norms and Domestic Politics: Bridging the Rationalist-Constructivist Divide.” European Journal of International Relations 3 (4): 473-495.

Clarkson, Joshua J., Zakary L. Tormala, Derek D. Rucker, and Riley G. Dugan. 2013. “The malleableinfluence of social consensus on attitude certainty.” Journal of Experimental Social Psychology49 (6): 1019-1022.

Colaresi, Michael. 2007. “The Benefit of the Doubt: Testing an Informational Theory of the RallyE↵ect.” International Organization 61 (1): 99-143.

Converse, Philip E. 1964. “The nature and origin of belief systems in mass publics.” In Ideology andDiscontent, ed. David E. Apter. New York: Free Press.

Dropp, Kyle A., Jim Golby, and Peter Feaver. 2014. “Elite Military Support and the Use of Force.”Working paper.

Dropp, Kyle A., Joshua D. Kertzer, and Thomas Zeitzo↵. 2014. “The less Americans know aboutUkraine’s location, the more they want U.S. to intervene.” Monkey Cage.

Druckman, James N. 2001. “On the Limits of Framing E↵ects: Who Can Frame?” Journal ofPolitics 63 (4): 1041-1066.

25

Druckman, James N. 2004. “Political Preference Formation: Competition, Deliberation, and the(Ir)relevance of Framing E↵ects.” American Political Science Review 98 (4): 671-686.

Druckman, James N., and Kjersten Nelson. 2003. “Framing and Deliberation: How Citizens’ Con-versations Limit Elite Influence.” American Journal of Political Science 47 (4): 729-745.

Edwards, George C. 2003. On Deaf Ears: The Limits of the Bully Pulpit. New Haven, CT: YaleUniversity Press.

Eichenberg, Richard C. 2005. “Victory Has Many Friends: U.S. Public Opinion and the Use ofMilitary Force, 1981-2005.” International Security 30 (1): 140-177.

Enns, Peter K. 2014. “The Public’s Increasing Punitiveness and Its Influence on Mass Incarcerationin the United States.” American Journal of Political Science 58 (4): 857-872.

Fanis, Maria. 2011. Secular Morality and International Security: American and British Decisionsabout War. Ann Arbor, MI: University of Michigan Press.

Fearon, James D. 1994. “Domestic Political Audiences and the Escalation of International Disputes.”American Political Science Review 88 (3): 577-592.

Friedkin, Noah E. 1999. “Choice Shift and Group Polarization.” American Sociological Review 64 (6):856-875.

Gartner, Scott Sigmund. 2008. “The Multiple E↵ects of Casualties on Public Support for War: AnExperimental Approach.” American Political Science Review 102 (1): 95-106.

Gelpi, Christopher. 2010. “Performing on Cue? The Formation of Public Opinion Toward War.”Journal of Conflict Resolution 54 (1): 88-116.

Gelpi, Christopher, Peter D. Feaver, and Jason Reifler. 2009. Paying the Human Costs of War.Princeton: Princeton University Press.

Goren, Paul, Harald Schoen, Jason Reifler, Thomas Scotto, and William Chittick. 2016. “A UnifiedTheory of Value-Based Reasoning and U.S. Public Opinion.” Political Behavior Forthcoming.

Graham, Jesse, Jonathan Haidt, and Brian A. Nosek. 2009. “Liberals and Conservatives Rely onDi↵erent Sets of Moral Foundations.” Journal of Personality and Social Psychology 96 (5): 1029-1046.

Grieco, Joseph M., Christopher Gelpi, Jason Reifler, and Peter D. Feaver. 2011. “Let’s Get a SecondOpinion: International Institutions and American Public Support for War.” International StudiesQuarterly 55 (2): 563-583.

Groeling, Tim, and Matthew A. Baum. 2008. “Crossing the Water’s Edge: Elite Rhetoric, MediaCoverage, and the Rally-Round-the-Flag Phenomenon.” Journal of Politics 70 (4): 1065-1085.

Guisinger, Alexandra, and Elizabeth N. Saunders. 2017. “Mapping the Boundaries of Elite Cues:How Elites Shape Mass Opinion Across International Issues.” International Studies QuarterlyForthcoming.

Hackman, J. Richard, and Nancy Katz. 2010. “Group Behavior and Performance.” In Handbook ofSocial Psychology, ed. Susan T. Fiske, Daniel T. Gilbert, and Gardner Lindzey. John Wiley andSons.

Hayes, Danny, and Matt Guardino. 2010. “Whose Views Made the News? Media Coverage and theMarch to War in Iraq.” Political Communication 27: 59-87.

Hayes, Danny, and Matt Guardino. 2011. “The Influence of Foreign Voices on U.S. Public Opinion.”American Journal of Political Science 55 (4): 830-850.

Hayes, Jarrod. 2012. “Securitization, Social Identity, and Democratic Security: Nixon, India, andthe Ties That Bind.” International Organization 66 (1): 63-93.

Herrmann, Richard K., Philip E. Tetlock, and Penny S. Visser. 1999. “Mass Public Decisions toGo to War: A Cognitive-Interactionist Framework.” American Political Science Review 93 (3):553-573.

Holsti, Ole R. 1979. “The Three-Headed Eagle: The United States and System Change.” Interna-tional Studies Quarterly 23 (3): 339-359.

Holsti, Ole R. 1992. “Public Opinion and Foreign Policy: Challenges to the Almond-LippmannConsensus.” International Studies Quarterly 36 (4): 439-466.

26

Holsti, Ole R. 2004. Public Opinion and American Foreign Policy. Revised edition ed. Ann Arbor,MI: University of Michigan Press.

Holsti, Ole R, and James N Rosenau. 1996. “Liberals, populists, libertarians, and conservatives: Thelink between domestic and international a↵airs.” International Political Science Review 17 (1):29–54.

Hopf, Ted. 2013. “Common Sense Constructivism and Hegemony in World Politics.” InternationalOrganization 67 (2): 317-354.

Hu↵, Connor, and Robert Schub. N.d. “The Inter-Temporal Tradeo↵ in Mobilizing Support for War.”Working paper. https://connordhu↵.files.wordpress.com/2016/10/intertemporaltradeo↵.pdf.

Hurwitz, Jon, and Mark Pe✏ey. 1987. “How Are Foreign Policy Attitudes Structured?” AmericanPolitical Science Review 81 (4): 1099-1120.

Jentleson, Bruce W. 1992. “The Pretty Prudent Public: Post Post-Vietnam American Opinion onthe Use of Military Force.” International Studies Quarterly 36 (1): 49-74.

Jones, Je↵rey M. 2014. “Americans’ Reaction to Middle East Sit-uation Similar to Past.” http://www.gallup.com/poll/174110/americans-reaction-middle-east-situation-similar-past.aspx.

Keele, Luke. 2007. “Social Capital and the Dynamics of Trust in Government.” American Journalof Political Science 51 (2): 241-254.

Kennan, George F. 1951. American Diplomacy, 1900-1950. Chicago: University of Chicago Press.Kertzer, Joshua D. 2013. “Making Sense of Isolationism: Foreign Policy Mood as a MultilevelPhenomenon.” Journal of Politics 75 (1): 225-240.

Kertzer, Joshua D. 2017. “Microfoundations in international relations.” Conflict Management andPeace Science Forthcoming.

Kertzer, Joshua D., and Kathleen M. McGraw. 2012. “Folk Realism: Testing the Microfoundationsof Realism in Ordinary Citizens.” International Studies Quarterly 56 (2): 245-258.

Kertzer, Joshua D., Kathleen Powers, Brian C. Rathbun, and Ravi Iyer. 2014. “Do moral valuesshape foreign policy preferences?” Journal of Politics 76 (3): 825-840.

Kertzer, Joshua D., and Ryan Brutger. 2016. “Decomposing Audience Costs: Bringing the AudienceBack into Audience Cost Theory.” American Journal of Political Science 60 (1): 234-249.

King, Gary, Jennifer Pan, and Margaret E. Roberts. 2016. “How the Chinese Government FabricatesSocial Media Posts for Strategic Distraction, not Engaged Argument.” Working paper.

Klar, Samara. 2014. “Partisanship in a Social Setting.” American Journal of Political Science 58 (3):687-704.

Kreps, Sarah. 2010. “Elite Consensus as a Determinant of Alliance Cohesion: Why Public OpinionHardly Matters for NATO-led Operations in Afghanistan.” Foreign Policy Analysis 6 (3): 191-215.

Krueger, Joachim, and Russell W. Clement. 1994. “The truly false consensus e↵ect: An ineradicableand egocentric bias in social perception.” Journal of Personality and Social Psychology 67 (4):596-610.

Le Bon, Gustave. 1896. The Crowd: A Study of the Popular Mind. New York: Macmillan.Levendusky, Matthew S., and Michael C. Horowitz. 2012. “When Backing Down is the Right Deci-sion: Partisanship, New Information, and Audience Costs.” Journal of Politics 74 (2): 323-338.

Lewis-Beck, Michael S., William G. Jacoby Helmut Norpoth and, and Herbert F. Weisberg. 2009.The American voter revisited. Ann Arbor, MI: University of Michigan Press.

Lippmann, Walter. 1955. Essays in the Public Philosophy. Boston: Little, Brown.Lupia, Arthur, and Mathew D. McCubbins. 1998. The Democratic Dilemma. Cambridge, UK: Cam-bridge University Press.

Lupia, Arthur, and Mathew D. McCubbins. 2000. “The Institutional Foundations of Political Com-petence: How Citizens Learn What They Need to Know.” In Elements of Reason: Cognition,Choice, and the Bounds of Rationality, ed. Arthur Lupia, Mathew D. McCubbins, and Samuel L.Popkin. Cambridge: Cambridge University Press.

Malhotra, Neil, and Yotam Margalit. 2010. “Short-Term Communication E↵ects or Longstanding

27

Dispositions? The Public’s Response to the Financial Crisis of 2008.” Journal of Politics 72 (3):852-867.

Mann, Christopher B., and Betsy Sinclair. 2013. “Voters Like You: Social Group Cuesin Behavior and Opinion.” Working paper. http://politics.as.nyu.edu/docs/IO/29116/SocialGroupings_MannSinclair_03.23.13.pdf.

Mayer, Jeremy D., and David J. Armor. 2012. “Support for torture over time: interrogating theAmerican public about coercive tactics.” The Social Science Journal 49 (4): 439-446.

McDermott, Rose, Dustin Tingley, Jonathan Cowden, Giovanni Frazzetto, and Dominic D. P. John-son. 2009. “Monoamine oxidase A gene (MAOA) predicts behavioral aggression following provo-cation.” Proceedings of the National Academy of Sciences 106 (7): 2118-2123.

McPherson, Miller, Lynn Smith-Lovin, and James M Cook. 2001. “Birds of a Feather: Homophilyin Social Networks.” Annual Review of Sociology 27: 415-444.

Mendelberg, Tali. 2005. “Bringing the Group Back Into Political Psychology: Erik H. Erikson EarlyCareer Award Address.” Political Psychology 26 (4): 637-650.

Messing, Solomon, and Sean J. Westwood. 2014. “Selective Exposure in the Age of Social Media:Endorsements Trump Partisan Source A�liation When Selecting News Online.” CommunicationResearch 41 (8): 1042-1063.

Milgram, Stanley. 1974. Obedience to authority: an experimental view. New York: Harper & Row.Morgenthau, Hans. 1948. Politics Among Nations. New York: Knopf.Mueller, John E. 1971. “Trends in Popular Support for the Wars in Korea and Vietnam.” AmericanPolitical Science Review 65 (2): 358-375.

Murray, Shoon. 2014. “Broadening the Debate about War: The Inclusion of Foreign Critics inMedia Coverage and Its Potential Impact on US Public Opinion.” Foreign Policy Analysis 10 (4):329-350.

Mutz, Diana C. 1998. Impersonal Influence: How Perceptions of Mass Collectives A↵ect PoliticalAttitudes. Cambridge, UK: Cambridge University Press.

Myers, David G., and Helmut Lamm. 1976. “The Group Polarization Phenomenon.” PsychologicalBulletin 83 (4): 602-627.

Nickerson, Raymond S. 1999. “How we know—and sometimes misjudge—what others know: Imput-ing one’s own knowledge to others.” Psychological bulletin 125 (6): 737–759.

Page, Benjamin I., and Marshall M. Bouton. 2007. The Foreign Policy Disconnect: What Americanswant from our leaders but don’t get. Chicago, IL: University of Chicago Press.

Page, Benjamin I., and Robert Y. Shapiro. 1992. The Rational Public: Fifty Years of Trends inAmericans’ Policy Preferences. Chicago: University of Chicago Press.

Pease, Andrew, and Paul R. Brewer. 2008. “The Oprah Factor: The E↵ects of a Celebrity Endorse-ment in a Presidential Primary Campaign.” The International Journal of Press/Politics 19 (3):386-400.

Pew Research Center. 2014. “Hamas Seen as More to Blame Than Israel for Current Violence:Deep Partisan Divide in Reactions to Mideast Fighting.” http://www.people-press.org/files/2014/07/7-28-14-Israel-Hamas-Release.pdf.

Pornpitakpan, Chanthika. 2004. “The Persuasiveness of Source Credibility: A Critical Review ofFive Decades’ Evidence.” Journal of Applied Social Psychology 34 (2): 243-281.

Radziszewski, Elizabeth. 2013. “Interpersonal Discussions and Attitude Formation on Foreign Policy:the Case of Polish Involvement in the Iraq War.” Foreign Policy Analysis 9 (1): 103-123.

Rathbun, Brian C. 2007. “Hierarchy and Community at Home and Abroad: Evidence of a Com-mon Structure of Domestic and Foreign Policy Beliefs in American Elites.” Journal of ConflictResolution 51 (3): 379-407.

Rathbun, Brian C., Joshua D. Kertzer, Jason Reifler, Paul Goren, and Thomas J. Scotto. 2016.“Taking Foreign Policy Personally: Personal Values and Foreign Policy Attitudes.” InternationalStudies Quarterly 60 (1): 124-137.

Renshon, Jonathan. 2015. “Losing Face and Sinking Costs: Experimental Evidence on the Judgment

28

of Political and Military Leaders.” International Organization 69 (3): 659-695.Rosenau, James N. 1965. Public Opinion and Foreign Policy: An Operational Formulation. NewYork: Random House.

Ross, Robert S. 2012. “Problem with the Pivot: Obama’s New Asia Policy Is Unnecessary andCounterproductive, The.” Foreign A↵. 91: 83.

Rothschild, David, and Neil Malhotra. 2014. “Are public opinion polls self-fulfilling prophecies?”Research and Politics 1 (2).

Saeki, Manabu. 2013. “The Myth of the Elite Cue: Influence of Voters’ Preferences on the USCongress.” Public Opinion Quarterly 77 (3): 755-782.

Sanders, Lynn M. 1999. “Democratic Politics and Survey Research.” Philosophy of the Social Sci-ences 29 (2): 248-280.

Saunders, Elizabeth N. 2015. “War and the Inner Circle: Democratic Elites and the Politics of UsingForce.” Security Studies 24 (3): 466-501.

Schultz, Kenneth A. 2001. “Looking for Audience Costs.” Journal of Conflict Resolution 45 (1):32-60.

Schultz, Kenneth A. 2005. “The Politics of Risking Peace: Do Hawks or Doves Deliver the OliveBranch?” International Organization 59 (1): 1-38.

Sinclair, Betsy. 2012. The Social Citizen: Peer Networks and Political Behavior. Chicago: Universityof Chicago Press.

Slantchev, Branislav L. 2006. “Politicians, the Media, and Domestic Audience Costs.” InternationalStudies Quarterly 50 (2): 445-477.

Smith, Eliot R., Charles R. Seger, and Diane M. Mackie. 2007. “Can Emotions Be Truly Group Level?Evidence Regarding Four Conceptual Criteria.” Journal of Personality and Social Psychology93 (3): 431-446.

Steenbergen, Marco R., Erica E. Edwards, and Catherine E. de Vries. 2007. “Who’s Cueing Whom?Mass-Elite Linkages and the Future of European Integration.” European Union Politics 8 (1):13-35.

Stein, Randy. 2013. “The pull of the group: Conscious conflict and the involuntary tendency towardsconformity.” Consciousness and Cognition 22 (3): 788-794.

Tarar, Ahmer, and Bahar Leventoglu. 2009. “Public Commitment in Crisis Bargaining.” Interna-tional Studies Quarterly 53 (3): 817-839.

Thompson, Alexander. 2006. “Coercion Through IOs: The Security Council and the Logic of Infor-mation Transmission.” International Organization 60 (1): 1-34.

Todorov, Alexander, and Anesu N. Mandisodza. 2004. “Public Opinion on Foreign Policy: TheMultilateral Public That Perceives Itself as Unilateral.” Public Opinion Quarterly 68 (3): 323-348.

Tomz, Michael. 2007. “Domestic Audience Costs in International Relations: An Experimental Ap-proach.” International Organization 61 (4): 821-40.

Trager, Robert F., and Lynn Vavreck. 2011. “The Political Costs of Crisis Bargaining: PresidentialRhetoric and the Role of Party.” American Journal of Political Science 55 (3): 526-545.

Visser, Penny S., and Robert R. Mirable. 2004. “Attitudes in the Social Context: The Impact ofSocial Network Composition on Individual-Level Attitude Strength.” Journal of Personality andSocial Psychology 87 (6): 779-795.

Wallace, Geo↵rey P.R. 2013. “International Law and Public Attitudes Toward Torture.” Interna-tional Organization 67 (1): 105-140.

Walsh, James Igoe. 2015. “Precision Weapons, Civilian Casualties, and Support for the Use ofForce.” Political Psychology 36 (5): 507-523.

Wittkopf, Eugene R. 1990. Faces of internationalism: Public Opinion and American foreign policy.Duke University Press.

Zaller, John R. 1992. The Nature and Origins of Mass Public Opinion. Cambridge: CambridgeUniversity Press.

29

Zeitzo↵, Thomas, John Kelly, and Gilad Lotan. 2015. “Using social media to measure foreign policydynamics An empirical analysis of the Iranian–Israeli confrontation (2012–13).” Journal of PeaceResearch 52 (3): 368–383.

30

A Bottom-Up Theory of Public Opinion about

Foreign PolicySupplementary Appendix

Contents

1 Examples of stimulus materials 2

Table 1: Type of Appeal: Experiments 1-2 . . . . . . . . . . . . . . . . . . . . . . . . 3Figure 1: Group endorse cue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3Figure 2: Group oppose cue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4Table 2: Elite partisan cue treatments for Experiment 3 . . . . . . . . . . . . . . . . 4Figure 3: China scenario with cold cognition treatment . . . . . . . . . . . . . . . . . 5Figure 4: Terrorism scenario with hot cognition treatment . . . . . . . . . . . . . . . 6Table 3: Elite cue treatments for Experiments 4-5 . . . . . . . . . . . . . . . . . . . 7

2 Supplementary analyses and robustness checks 8

2.1 Study 1 (SSI): Experiments # 1-2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8Table 4: Randomization check . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8Table 5: Summary statistics and sample characteristics . . . . . . . . . . . . . . . . . 9Figure 5: Substantive e↵ects of treatments and main dispositional variables . . . . . 10Table 6: Do elite endorsements matter in the absence of group treatments? . . . . . 11Table 7: Results for only those who passed the manipulation check . . . . . . . . . . 12Table 8: No evidence of elite cue x partisanship x group cue interactions . . . . . . . 13Figure 6: No evidence of elite cue x partisanship x group cue interactions . . . . . . 142.1.1 The e↵ects of social cues on certainty and associated beliefs . . . . . . . . . . 15Table 9: E↵ects on perceptions of certainty, success, and threat . . . . . . . . . . . . 16

2.2 Study 2 (Amazon MTurk): Experiment #3 . . . . . . . . . . . . . . . . . . . . . . . 17Table 10: Randomization check . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17Table 11: Summary statistics and sample characteristics . . . . . . . . . . . . . . . . 18Table 12: Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19Figure 7: E↵ect of elite consensus magnified by social cues . . . . . . . . . . . . . . . 20Table 13: Results for only those who passed the manipulation check . . . . . . . . . 21

2.3 Study 3 (Amazon MTurk): Experiments #4-5 . . . . . . . . . . . . . . . . . . . . . . 22Table 14: Randomization check: China . . . . . . . . . . . . . . . . . . . . . . . . . . 22Table 15: Randomization check: ICSID . . . . . . . . . . . . . . . . . . . . . . . . . 232.3.1 Comparison of group cue treatments . . . . . . . . . . . . . . . . . . . . . . . 23Table 16: Rank-sum tests comparing the two types of group cues . . . . . . . . . . . 25Figure 8: Density distributions of group cues . . . . . . . . . . . . . . . . . . . . . . 26Table 17: Summary statistics and sample characteristics . . . . . . . . . . . . . . . . 27Table 19: Study 3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 292.3.2 Explaining variation in the e�cacy of elite cues . . . . . . . . . . . . . . . . . 31Figure 5: Substantive e↵ects of treatments and main dispositional variables . . . . . 322.3.3 Subgroup analysis by trust and vote choice . . . . . . . . . . . . . . . . . . . 33Table 21: Elite cues are three times stronger for Clinton supporters than Trump

supporters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 342.4 Salience of foreign policy during survey periods . . . . . . . . . . . . . . . . . . . . . 35

1

1 Examples of stimulus materials

Type of Appeal

Previous research has found that emotions and emotional appeals can influence political behavior —

including perception of threat (Lerner et al., 2003), ingroup cohesion (Zeitzo↵, 2014), rally ‘round the

flag e↵ects (Aday, 2010; Lambert et al., 2010), and voter persuasion (Brader, 2005). Furthermore,

Kahneman (2011) argues that cognition occurs in two modes — an impulsive, “hot” cognition,

and a slower, e↵ortful (“cold”) type of thinking. Since we were interested in how di↵erent partisan

endorsements and group cues influence foreign policy opinion, in Experiments 1-2 we also investigate

the possibility that that di↵erent appeals — a colder, cognitive message (Cold Cognition Treatment),

and a hotter, a↵ect-laden one (Hot Cognition Treatment) — may change how subjects process the

various endorsements.

In each of the two experiments, subjects were randomly shown a map (Cold Cognition Treat-

ment), or a picture that was found to be threatening (Hot Cognition Treatment).1 The argument

put forth by the Democrat or Republican elite policymaker in our experiment also varied depending

on the treatment. Table 1 shows how these appeals varied by appeal type (Cold Cognition or Hot

Cognition) and scenario (China or Terrorism).

1A pre-test on 100 American adults recruited using Amazon Mechanical Turk was used to select threatening andmore neutral stimuli. Pre-test results confirmed the images used in the Hot Cognition treatment significantly increasedfearful and threatening perceptions compared to the Cold Cognition treatment. This is similar to the manipulationused in Gadarian (2014). For a helpful guide to developing emotional manipulations in political science experiments,see Albertson and Gadarian (2016).

2

Table 1: Type of Appeal: Experiments 1-2

Scenario Emotional LogicalChina “It’s not rocket sci-

ence. China is tryingto bully the US, andbullies only respond toforce. My gut tells mewe need to shift mil-itary resources to theregion to send a signaland protect our inter-ests.”

“China is using its mil-itary to expand it’s in-fluence. Cool, coldlogic dictates that weneed to shift militaryresources to the regionto send a signal andprotect our interests.”

Terrorism “It’s not rocket sci-ence. Terrorists aretrying to kill Ameri-cans, my gut tells mewe should use our mili-tary to get them overthere before they at-tack us.”

“Terrorists are usingthese countries as abase of operations.Cool, cold logic dic-tates that we shoulduse our military toneutralize the terroristthreat over there.”

Figure 1: Group Endorse Cue

The graph below shows the responses of people who have previously taken the survey. Those whoanswered the earlier questions on the survey like you strongly supported sending US special forces

into foreign countries to go after terrorists.

3

Figure 2: Group Oppose Cue

The graph below shows the responses of people who have previously taken the survey. Those whoanswered the earlier questions on the survey like you strongly opposed sending US special forces

into foreign countries to go after terrorists.

Table 2: Elite partisan cue treatments for Experiment 3

Cue Wording

Control [blank ]Dem. Support, Repub. Oppose Republicans and Democrats in

Congress are divided on the is-sue. Republicans strongly sup-port shifting US military re-sources to the region, whileDemocrats oppose such a move,and call for diplomatic e↵orts in-stead.

Repub. Support, Dem. Oppose Democrats and Republicans inCongress are divided on the is-sue. Democrats strongly supportshifting US military resources tothe region, while Republicans op-pose such a move, and call fordiplomatic e↵orts instead.

Both Support Both Republicans andDemocrats in Congress areunited on the issue, and stronglysupport shifting US militaryresources to the region.

4

Figure 3: China Scenario with Cold Cognition Treatment

5

Figure 4: Terrorism Scenario with Hot Cognition Treatment

6

Table 3: Elite cue treatments for Experiments 4-5

Cue ICSID Scenario Wording China Scenario Wording

Control Those who support ICSID arguethat it protects investments andguarantees a transparent legalprocess for resolving disputes.Others have argued that ICSIDtilts the playing field further infavor of big multinational corpo-rations, and that disputes withforeign investors should be han-dled by the existing American le-gal system.

Some have argued that the USshould increase its naval pres-ence to deter China from furtherprovocative acts in the SouthChina Sea. Others have arguedthat such a move is a risky choicethat may escalate tensions evenfurther, and have instead calledfor diplomacy.

Elite Divided Democrats and Republicans inCongress are divided on the is-sue. Republicans strongly sup-port using ICSID for investor-state disputes, while Democratsare opposed, calling for disputeswith foreign investors to be han-dled by the existing American le-gal system.

Democrats and Republicans inCongress are divided on theissue. Republicans stronglysupport increasing US navalpresence in the region, whileDemocrats oppose such a move,and call for diplomatic e↵orts in-stead.

Elite Consensus Those who support ICSID arguethat it protects investments andguarantees a transparent legalprocess for resolving disputes.Others have argued that ICSIDtilts the playing field further infavor of big multinational corpo-rations, and that disputes withforeign investors should be han-dled by the existing American le-gal system. Democrats and Re-publicans in Congress are united.Both Democrats and Republi-cans strongly support using IC-SID for investor-state disputes.

Some have argued that the USshould increase its naval pres-ence to deter China from furtherprovocative acts in the SouthChina Sea. Others have ar-gued that such a move is a riskychoice that may escalate tensionseven further, and have insteadcalled for diplomacy. Democratsand Republicans in Congress areunited. Both Democrats andRepublicans strongly support in-creasing US naval presence in theregion.

7

2 Supplementary analyses and robustness checks

2.1 Study 1 (SSI): Experiments # 1-2

Table 4: Randomization Check on Treatments (Logit)

Dependent Variable: Assignment to Treatment

Emotional Appeal Democrat Endorse Group Endorse Group Oppose

(1) (2) (3) (4)

Male �0.150 �0.093 0.047 0.004(0.128) (0.128) (0.136) (0.135)

White �0.090 0.092 �0.185 �0.025(0.176) (0.176) (0.184) (0.185)

Age 0.002 0.002 �0.007 0.001(0.004) (0.004) (0.004) (0.004)

Education �0.019 �0.031 0.013 0.016(0.035) (0.035) (0.037) (0.037)

Income 0.037 �0.019 0.041 �0.001(0.032) (0.032) (0.034) (0.034)

Party ID 0.021 �0.011 0.043 �0.033(0.032) (0.032) (0.034) (0.034)

Militant Assertiveness 0.165 0.176 �0.479 0.574(0.336) (0.337) (0.359) (0.356)

Internationalism 0.244 0.344 �0.038 0.280(0.367) (0.367) (0.391) (0.388)

N 1,031 1,031 1,031 1,031AIC 1,443.144 1,443.422 1,317.756 1,339.891

Results from study 1. ⇤p < .1; ⇤⇤p < .05; ⇤⇤⇤p < .01.

8

Tab

le5:

SummaryStatisticsan

dSam

ple

Characteristics

NNA

Min

Max

Median

Mean

Std.Dev.

Description

Male

1035

00

10.000

0.480

0.500

White

1035

00

11.000

0.829

0.377

White,

not

Hispan

ic.

Age

1035

018

8952.000

50.028

15.944

Education

1035

01

96.000

6.022

1.953

From

nohighschoo

lto

grad

uate

degree.

Mean(6)is

Associate’s

degree.

Income

1032

31

103.000

3.719

2.118

Hou

sehold

income.

From

less

than

20,000

USD

tomorethan

200,000USD.Mean

isbetween

35,000

USD

and75,000

USD.

Party

ID1034

11

74.000

3.909

2.196

7-point

scale;

1(Stron

gDem

o-crat)to

7(Stron

gRepublican).

Militan

tAssertiveness

1035

00

10.500

0.509

0.205

Militan

tAssertiveness

scale

(Herrm

ann,Tetlock,an

dVisser,

1999;

Kertzer

and

McG

raw,

2012);normalized

to0-1.

Internationalism

1035

00

10.563

0.544

0.179

Internationalism

scale

(Her-

rman

n,

Tetlock,

and

Visser,

1999;

Kertzer

and

McG

raw,

2012);normalized

to0-1.

Arm

edForce

ChinaScenario

1035

00

10.478

0.466

0.271

Supportforsendingmilitaryre-

sources

toAsia.

Con

tinu

ous0-10

normalized

to0-1.

0(Stron

gly

Oppose)

to1(Stron

glySupport).

Arm

edForce

Terrorism

Scenario

1021

140

10.500

0.488

0.287

SupportforsendingUS

Special

Forcesto

figh

tterrorism.

Con

-tinu

ous0-10

scalenormalized

to0-1.

0(Stron

gly

Oppose)

to1

(Stron

glySupport).

Resultsfrom

study1.

9

Figure

5:Substan

tive

e↵ects

oftreatm

ents

andmaindispositional

variab

les

a) C

hina

Eff

ect

siz

e

-0.2

0.0

0.2

0.4

0.6

Em

otio

na

l A

pp

ea

l

De

mo

cra

t E

nd

ors

e

Gro

up

En

do

rse

Gro

up

Op

po

se

Pa

rty I

D

Milita

nt

Asse

rtiv

en

ess

Inte

rna

tio

na

lism

Ba

sic

mo

de

l

No

gro

up

tre

atm

en

t

Pa

sse

d m

an

ipu

latio

n c

he

ck

b) T

erro

rism

Eff

ect

siz

e

-0.2

0.0

0.2

0.4

0.6

Ba

sic

mo

de

l

No

gro

up

tre

atm

en

t

Pa

sse

d m

an

ipu

latio

n c

he

ck

This

figure

plots

thesu

bstantivee↵

ects

from

thebasictrea

tmen

te↵

ectmodels(see

models1and2in

Table

??)in

black

,aseries

ofmodelsestimatedsolely

onth

esu

bsetofparticipants

whodid

notreceiveagroupcu

e(see

models3an

d4in

Table

6)in

dark

grey,

andasetofmodelsestimatedsolely

onth

ose

participants

who

passed

theelitecu

emanipulationch

eck(see

models1and2in

Table

3in

Appen

dix

§2.1)in

lightgrey.

Allth

reedispositionalva

riablesare

rescaledfrom

0-1,so

that

thee↵

ectestimate

represents

thee↵

ectofgoingfrom

theminim

um

toth

emaxim

um

level

ofea

chva

riable

(e.g.from

strongDem

ocratto

strongRep

ublica

n,etc.).

Theresu

ltsremain

consisten

tth

roughout:

groupcu

essignifica

ntlya↵ectparticipants’views,

whileth

eelitepartisancu

ehasnoe↵

ect,

andth

ee↵

ectofpartisansh

ipis

wea

ker

thanth

atofgen

eralforeignpolicy

orien

tations.

10

Tab

le6:

DoElite

Endorsements

Matterin

theAbsence

ofGroupTreatments?(O

LS)

Dep

endentVariable:SupportforArm

edForce

China

Terrorism

China

Terrorism

China

Terrorism

China

Terrorism

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Emotional

Appeal

0.027

�0.015

0.025

�0.004

0.028

�0.017

0.025

�0.005

(0.029)

(0.033)

(0.027)

(0.030)

(0.029)

(0.033)

(0.027)

(0.030)

Dem

ocratEndorse

�0.051⇤

�0.021

�0.039

�0.029

�0.048

0.037

�0.062

0.060

(0.029)

(0.033)

(0.027)

(0.029)

(0.060)

(0.069)

(0.056)

(0.061)

Party

ID0.007

0.003

0.023⇤

⇤0.033⇤

⇤⇤0.005

0.015

(0.007)

(0.008)

(0.009)

(0.011)

(0.009)

(0.010)

Party

ID⇥

Dem

.Endorse

�0.00003

�0.015

0.006

�0.022

(0.013)

(0.015)

(0.013)

(0.014)

Militan

tAssertiveness

0.496⇤

⇤⇤0.621⇤

⇤⇤0.497⇤

⇤⇤0.628⇤

⇤⇤

(0.067)

(0.074)

(0.067)

(0.074)

Internationalism

0.137⇤

0.194⇤

⇤0.138⇤

0.197⇤

⇤

(0.072)

(0.080)

(0.072)

(0.079)

Con

trols

XX

XX

N338

334

336

332

338

334

336

332

Adjusted

R2

0.009

�0.006

0.201

0.235

0.039

0.021

0.199

0.239

⇤ p<

.1;⇤⇤p<

.05;

⇤⇤⇤ p

<.01

This

regressiononly

looksatth

esu

bsetofresp

onden

tswhodid

notreceiveeith

erth

eGroupEndorseorGroupOppo

seco

ndition.Allregressionsare

OLSand

controlforth

erandomly

assigned

ord

erofth

escen

arios(C

hinaScenarioorTerrorism

Scenariofirst).Controls

includeMale,Age,Education,In

come,

andW

hite.

11

Table 7: Results for only those participants who passed manipulation check

Dependent Variable: Support for Armed Force

China Terrorism China Terrorism

(1) (2) (3) (4)

Emotional Appeal 0.016 �0.008 0.016 �0.008(0.017) (0.018) (0.017) (0.018)

Democrat Endorse �0.003 �0.029 �0.001 0.022(0.017) (0.018) (0.035) (0.037)

Group Endorse 0.088⇤⇤⇤ 0.054⇤⇤ 0.088⇤⇤⇤ 0.054⇤⇤

(0.021) (0.022) (0.021) (0.022)

Group Oppose �0.053⇤⇤⇤ �0.060⇤⇤⇤ �0.053⇤⇤⇤ �0.060⇤⇤⇤

(0.020) (0.022) (0.020) (0.022)

Party ID 0.007⇤ 0.010⇤⇤ 0.008 0.017⇤⇤

(0.004) (0.005) (0.006) (0.007)

Party ID X Democrat Endorse �0.0004 �0.013(0.008) (0.008)

Militant Assertiveness 0.530⇤⇤⇤ 0.589⇤⇤⇤ 0.530⇤⇤⇤ 0.596⇤⇤⇤

(0.045) (0.050) (0.045) (0.050)

Internationalism 0.151⇤⇤⇤ 0.226⇤⇤⇤ 0.151⇤⇤⇤ 0.229⇤⇤⇤

(0.048) (0.053) (0.048) (0.053)

Controls X X X XN 824 766 824 766Adjusted R2 0.229 0.260 0.229 0.261

Results from study 1. ⇤p < .1; ⇤⇤p < .05; ⇤⇤⇤p < .01

12

Table 8: No evidence of elite cue x partisanship x group cue interactions

Dependent Variable: Support for Armed Force

China Terrorism

(1) (2)

Emotional Appeal 0.012 �0.003(0.015) (0.016)

Democrat Endorse �0.070 0.059(0.055) (0.057)

Group Endorse �0.035 0.041(0.054) (0.055)

Group Oppose �0.103⇤⇤ �0.052(0.052) (0.056)

Party ID 0.004 0.017⇤

(0.009) (0.010)

Militant Assertiveness 0.518⇤⇤⇤ 0.570⇤⇤⇤

(0.040) (0.042)

Internationalism 0.138⇤⇤⇤ 0.220⇤⇤⇤

(0.044) (0.046)

Democrat Endorse X Group Endorse 0.137⇤ �0.051(0.077) (0.078)

Democrat Endorse X Party ID 0.008 �0.022⇤

(0.012) (0.013)

Group Endorse X Party ID 0.019 �0.001(0.012) (0.013)

Democrat Endorse X Group Oppose 0.098 �0.072(0.075) (0.080)

Group Oppose X Party ID 0.004 0.002(0.012) (0.013)

Democrat Endorse X Group Endorse X Party ID �0.020 0.022(0.017) (0.017)

Democrat Endorse X Group Oppose X Party ID �0.014 0.007(0.017) (0.018)

Controls X XN 1,031 1,017Adjusted R2 0.227 0.263

Results from study 1. ⇤p < .1; ⇤⇤p < .05; ⇤⇤⇤p < .01. See Figure 6 for a visualization of the e↵ects.

13

Figure 6: No evidence of elite cue x partisanship x group cue interactions

(a) China

Dem

Rep

Oppose

Contro

lEndors

e

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Elite C

ueG

roup C

ues

Support

Democrats

Dem

Rep

Oppose

Contro

lEndors

e

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Elite C

ueG

roup C

ues

Support

Republicans

(b) Terrorism

Dem

Rep

Oppose

Contro

lEndors

e

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Elite C

ueG

roup C

ues

Support

Democrats

Dem

Rep

Oppose

Contro

lEndors

e

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Elite C

ueG

roup C

ues

Support

Republicans

Illustrating the substantive e↵ects from Table 8 from Study 1, we find no evidence that the impact of elite cues ismoderated by group cues, or that the impact of elite cues is moderated by partisanship. Given the nature of thetheory being tested (in which the e↵ect of elite cues is conditional on partisanship, but also may be conditional onsocial cues), it is necessary to estimate a fully-saturated three-way interaction model (Braumoeller, 2004). The top

row depicts the treatment e↵ects in the China scenario for Democrats and Republicans, respectively, while thebottom row does the same for the treatment e↵ects in the Terrorism scenario. Importantly, we see the same

“staircase” pattern across all panels, showing the consistent e↵ects of the group cue treatments, irrespective of theelite cue or respondents’ own partisan a�liation.

14

2.1.1 The e↵ects of social cues on certainty and associated beliefs

Although our main interest is in testing how these di↵erent types of cues mobilize support for the

use of force, we also included a number of additional questions to ascertain how certain participants

were about their position on using force, how successful they thought the use of force would be at

achieving its goal, and how much of a threat they perceived from from the scenarios described. These

measures, the results for which are displayed in Table 9, are of interest in as much as they allow us

to observe not just how much they supported a given mission, but the potential mechanisms through

which the treatments shape judgments, and the broader architecture of participants’ beliefs. In the

previous set of analyses, we saw that the emotional appeal lacked significant e↵ects on mobilizing

support; here we find that participants in the hot cognition (Emotional Appeal) treatment were more

likely to perceive a threat from China, and also more likely to perceive the pivot to Asia as being

successful. Interestingly, we find that group endorsements have a stronger e↵ect in terms of increasing

certainty, perceived likelihood of success, and threat perception than group opposition does – the

Group Endorse treatment significantly increases participants’ certainty about their decision in the

China scenario, increases perceptions of the likelihood of success in both the China scenario and the

Terrorism scenario, and increases threat perception in the China scenario. This asymmetry between

group endorsements and group opposition is of theoretical interest, and merits future study.

As before, participant-level characteristics exert the largest impact. Consistent with the psy-

chological literature on the relationship between conservatism, uncertainty avoidance, and threat

management (Jost et al., 2007), we see that across both scenarios Republicans express more cer-

tainty about their responses than Democrats do, and also perceive higher levels of threat. Once

again, though, the substantively largest contributions to the model come from participants’ prior

foreign policy orientations: participants high in military assertiveness — who tend to believe in

the e�cacy of the use of force – are far more likely to believe the missions will be successes than

their dovish counterparts; internationalists are similarly optimistic compared to isolationists. Al-

though we see similarly sensible results for militant assertiveness with respect to threat perceptions

— hawks are more likely to perceive a threat in both the China and Terrorism scenarios — we see

that internationalists are actually less rather than more likely to perceive a threat posed by a rising

China, reflecting the presence of multiple “faces” of internationalism: a military internationalism

eager to deploy force abroad, and a cooperative internationalism that sees opportunities for gains

from trade and mutual cooperation (Wittkopf, 1990; Holsti, 2004).

15

Table 9: E↵ects on Perceptions of Certainty, Success, and Threat (OLS)

Certainty of Action Likelihood of Success Threat Posed

China Terrorism China Terrorism China Terrorism

(1) (2) (3) (4) (5) (6)

Emotional Appeal 0.018 0.001 0.028⇤ �0.025 0.036⇤⇤ �0.013(0.016) (0.017) (0.014) (0.015) (0.015) (0.015)

Democrat Endorse 0.001 0.007 0.011 �0.027⇤ �0.003 �0.015(0.016) (0.017) (0.014) (0.015) (0.015) (0.015)

Group Endorse 0.056⇤⇤⇤ 0.012 0.057⇤⇤⇤ 0.049⇤⇤⇤ 0.040⇤⇤ 0.021(0.020) (0.021) (0.018) (0.018) (0.019) (0.018)

Group Oppose 0.016 0.021 �0.025 �0.025 0.002 �0.012(0.020) (0.021) (0.018) (0.019) (0.019) (0.018)

Party ID 0.012⇤⇤⇤ 0.014⇤⇤⇤ 0.003 0.002 0.008⇤ 0.013⇤⇤⇤

(0.004) (0.004) (0.004) (0.004) (0.004) (0.004)

Militant Assertiveness �0.019 0.024 0.543⇤⇤⇤ 0.565⇤⇤⇤ 0.426⇤⇤⇤ 0.397⇤⇤⇤

(0.044) (0.046) (0.039) (0.040) (0.042) (0.039)

Internationalism 0.013 0.007 0.124⇤⇤⇤ 0.206⇤⇤⇤ �0.094⇤⇤ 0.021(0.048) (0.050) (0.042) (0.044) (0.045) (0.043)

Controls X X X X X XN 1,031 1,031 1,031 1,031 1,031 1,031Adjusted R2 0.028 0.024 0.217 0.246 0.141 0.163

Results from study 1. ⇤p < .1; ⇤⇤p < .05; ⇤⇤⇤p < .01

16

2.2 Study 2 (Amazon MTurk): Experiment #3

Table 10: Randomization Check on Treatments (Logit)

Dem Support Rep Support Elite Consensus Group Endorse Group Oppose

(1) (2) (3) (4) (5)

Militant Assertiveness �0.239 0.180 �0.037 0.372 �0.501⇤

(0.302) (0.305) (0.298) (0.282) (0.279)Internationalism 0.322 0.380 0.005 0.137 �0.235

(0.347) (0.351) (0.342) (0.323) (0.317)Party ID �0.139 0.225 0.155 �0.050 0.079

(0.256) (0.257) (0.251) (0.237) (0.235)Male 0.059 0.224⇤ �0.115 0.083 0.075

(0.125) (0.128) (0.123) (0.117) (0.115)Age 0.002 �0.004 �0.004 0.006 0.002

(0.006) (0.006) (0.006) (0.005) (0.005)Education �0.085⇤ 0.050 0.019 0.035 0.003

(0.048) (0.049) (0.048) (0.045) (0.044)Constant �0.887⇤⇤ �1.727⇤⇤⇤ �0.987⇤⇤⇤ �1.383⇤⇤⇤ �0.521

(0.366) (0.377) (0.364) (0.345) (0.337)N 1,444 1,444 1,444 1,444 1,444AIC 1,639.504 1,603.000 1,662.231 1,807.159 1,847.190

Results from study 2. ⇤p < .1; ⇤⇤p < .05; ⇤⇤⇤p < .01

17

Tab

le11:SummaryStatisticsan

dSam

ple

CharacteristicsforStudy2

NNA

Min

Max

Median

Mean

Std.Dev.

Description

Male

1444

10

11.00

0.494

0.500

Age

1445

018

7557.00

34.117

11.022

Education

1445

01

85.000

4.223

1.271

From

nohighschoo

l(1)to

grad

-uatedegree(8).

Mean(4)is

As-

sociate’sdegree.

Party

ID1445

11

73.000

3.363

1.545

7-point

scale;

1(Stron

gDem

o-crat)to

7(Stron

gRepublican).

Militan

tAssertiveness

1445

00

10.375

0.406

0.213

Militan

tAssertiveness

scale

(Herrm

ann,Tetlock,an

dVisser,

1999;

Kertzer

and

McG

raw,

2012);normalized

to0-1.

Internationalism

1445

00

10.625

0.591

0.184

Internationalism

scale

(Her-

rman

n,

Tetlock,

and

Visser,

1999;

Kertzer

and

McG

raw,

2012);normalized

to0-1.

Arm

edForce

ChinaScenario

1445

00

10.378

0.394

0.273

Supportforsendingmilitaryre-

sources

toAsia.

Con

tinu

ous0-10

normalized

to0-1.

0(Stron

gly

Oppose)

to1(Stron

glySupport).

18

Table 12: Study 2: Results

(1) (2) (3) (4) (5)

Dem Support �0.029 �0.090⇤⇤⇤ �0.059⇤⇤ �0.087⇤⇤ �0.058⇤⇤

(0.020) (0.033) (0.027) (0.035) (0.029)Rep Support �0.028 �0.061⇤ �0.044 �0.045 �0.026

(0.020) (0.034) (0.028) (0.037) (0.030)Elite Consensus 0.037⇤ �0.028 �0.005 �0.012 0.009

(0.020) (0.033) (0.027) (0.035) (0.029)Group Endorse 0.050⇤⇤⇤ �0.015 �0.004 0.014 0.011

(0.017) (0.035) (0.029) (0.037) (0.031)Group Oppose �0.078⇤⇤⇤ �0.139⇤⇤⇤ �0.079⇤⇤⇤ �0.125⇤⇤⇤ �0.075⇤⇤

(0.017) (0.034) (0.028) (0.037) (0.031)Dem Support ⇥ Group Endorse 0.099⇤⇤ 0.065 0.076 0.051

(0.049) (0.040) (0.053) (0.044)Dem Support ⇥ Group Oppose 0.096⇤⇤ 0.036 0.067 0.021

(0.047) (0.039) (0.052) (0.043)Rep Support ⇥ Group Endorse 0.045 0.023 0.014 0.009

(0.050) (0.041) (0.053) (0.044)Rep Support ⇥ Group Oppose 0.061 �0.011 0.046 �0.016

(0.048) (0.040) (0.053) (0.044)Elite Consensus ⇥ Group Endorse 0.114⇤⇤ 0.092⇤⇤ 0.083 0.074⇤

(0.049) (0.040) (0.052) (0.043)Elite Consensus ⇥ Group Oppose 0.090⇤ 0.029 0.062 0.020

(0.048) (0.039) (0.051) (0.043)Dem Support ⇥ Party ID 0.002 0.013

(0.042) (0.035)Rep Support ⇥ Party ID 0.033 0.055

(0.044) (0.036)Elite Consensus ⇥ Party ID 0.041 0.052

(0.042) (0.034)Group Endorse ⇥ Party ID 0.077⇤ 0.048

(0.045) (0.037)Group Oppose ⇥ Party ID 0.023 0.024

(0.044) (0.036)Dem Support ⇥ Group Endorse ⇥ Party ID �0.058 �0.040

(0.064) (0.053)Dem Support ⇥ Group Oppose ⇥ Party ID �0.064 �0.060

(0.062) (0.051)Rep Support ⇥ Group Endorse ⇥ Party ID �0.032 �0.019

(0.063) (0.052)Rep Support ⇥ Group Oppose ⇥ Party ID 0.003 �0.013

(0.062) (0.051)Elite Consensus ⇥ Group Endorse ⇥ Party ID �0.073 �0.059

(0.061) (0.051)Elite Consensus ⇥ Group Oppose ⇥ Party ID �0.043 �0.039

(0.060) (0.050)Militant Assertiveness 0.667⇤⇤⇤ 0.671⇤⇤⇤

(0.029) (0.029)Internationalism 0.222⇤⇤⇤ 0.219⇤⇤⇤

(0.033) (0.033)Male �0.010 �0.009

(0.012) (0.012)Age �0.0003 �0.0003

(0.001) (0.001)Education 0.006 0.006

(0.005) (0.005)Party ID 0.064⇤⇤⇤ 0.023 �0.017

(0.024) (0.029) (0.024)Constant 0.409⇤⇤⇤ 0.448⇤⇤⇤ �0.007 0.454⇤⇤⇤ 0.015

(0.017) (0.023) (0.040) (0.024) (0.038)N 1,445 1,445 1,444 1,445 1,444Adjusted R2 0.043 0.045 0.363 0.068 0.364

Note: ⇤p < .1; ⇤⇤p < .05; ⇤⇤⇤p < .01. Treatment e↵ects are in relation to the elite cue and group cue controls.

19

0.0

0.2

0.4

Endorse Control Oppose

Group Cue

Su

ppo

rt

Elite CueControl

Dem Support

Rep Support

Consensus

Figure 7: Polarized elite cues lack significant e↵ects, while the e↵ect of elite consensus is magnifiedby social cues. The figure visualizes the results from model 3 of Table 12 in Appendix §2.2. Barsrepresent 95% confidence intervals.

20

Table 13: Results for only those participants who passed the manipulation check

(1) (2) (3) (4) (5)

Dem Support �0.030 �0.089⇤⇤⇤ �0.056⇤⇤ �0.066 �0.048(0.020) (0.033) (0.027) (0.059) (0.049)

Rep Support �0.031 �0.063⇤ �0.043 �0.103 �0.113⇤⇤

(0.021) (0.034) (0.028) (0.064) (0.053)Elite Consensus 0.040⇤ �0.024 �0.003 �0.048 �0.055

(0.020) (0.033) (0.027) (0.059) (0.049)Group Endorse 0.050⇤⇤⇤ �0.012 0.004 �0.054 �0.035

(0.018) (0.035) (0.029) (0.065) (0.054)Group Oppose �0.073⇤⇤⇤ �0.137⇤⇤⇤ �0.076⇤⇤⇤ �0.146⇤⇤ �0.110⇤⇤

(0.017) (0.034) (0.028) (0.061) (0.051)Dem Support ⇥ Group Endorse 0.096⇤ 0.057 0.104 0.062

(0.050) (0.041) (0.089) (0.074)Dem Support ⇥ Group Oppose 0.096⇤⇤ 0.033 0.126 0.078

(0.048) (0.040) (0.087) (0.072)Rep Support ⇥ Group Endorse 0.038 0.014 0.028 0.029

(0.050) (0.041) (0.093) (0.078)Rep Support ⇥ Group Oppose 0.068 �0.007 0.002 �0.014

(0.050) (0.041) (0.091) (0.075)Elite Consensus ⇥ Group Endorse 0.112⇤⇤ 0.087⇤⇤ 0.153⇤ 0.157⇤⇤

(0.049) (0.041) (0.089) (0.074)Elite Consensus ⇥ Group Endorse 0.093⇤ 0.030 0.117 0.093

(0.049) (0.040) (0.086) (0.071)Dem Support ⇥ Party ID �0.051 �0.026

(0.123) (0.102)Rep Support ⇥ Party ID 0.112 0.173

(0.135) (0.113)Elite Consensus ⇥ Party ID 0.074 0.128

(0.123) (0.103)Group Endorse ⇥ Party ID 0.118 0.093

(0.136) (0.113)Group Oppose ⇥ Party ID 0.049 0.081

(0.134) (0.111)Dem Support ⇥ Group Endorse ⇥ Party ID �0.020 �0.008

(0.192) (0.160)Dem Support ⇥ Group Oppose ⇥ Party ID �0.102 �0.107

(0.189) (0.157)Rep Support ⇥ Group Endorse ⇥ Party ID �0.011 �0.043

(0.196) (0.163)Rep Support ⇥ Group Oppose ⇥ Party ID 0.129 0.016

(0.197) (0.164)Elite Consensus ⇥ Group Endorse ⇥ Party ID �0.122 �0.174

(0.190) (0.158)Elite Consensus ⇥ Group Oppose ⇥ Party ID �0.107 �0.156

(0.182) (0.152)Militant Assertiveness 0.659⇤⇤⇤ 0.657⇤⇤⇤

(0.029) (0.029)Internationalism 0.225⇤⇤⇤ 0.226⇤⇤⇤

(0.033) (0.033)Male �0.012 �0.012

(0.012) (0.012)Age �0.0003 �0.0002

(0.001) (0.001)Education 0.005 0.006

(0.005) (0.005)Party ID 0.073⇤⇤⇤ 0.118 �0.009

(0.025) (0.084) (0.071)Constant 0.406⇤⇤⇤ 0.444⇤⇤⇤ �0.011 0.395⇤⇤⇤ 0.020

(0.017) (0.023) (0.041) (0.042) (0.049)N 1,399 1,399 1,398 1,399 1,398Adjusted R2 0.041 0.043 0.357 0.073 0.360

Results from study 2. Note: ⇤p < .1; ⇤⇤p < .05; ⇤⇤⇤p < .01. Treatment e↵ects in relation to the elite & group cue controls.

21

2.3 Study 3 (Amazon MTurk): Experiments #4-5

Table 14: Randomization check: China

Elite Divided Elite Consensus Group Endorse Group Oppose

(1) (2) (3) (4)

Military assertiveness �0.080 �0.300 0.172 0.059(0.247) (0.248) (0.238) (0.237)

Internationalism 0.550⇤⇤ �0.077 0.353 �0.283(0.269) (0.267) (0.258) (0.256)

Party ID 0.135 0.092 0.058 �0.202(0.204) (0.205) (0.197) (0.196)

Male �0.045 0.022 0.028 0.035(0.096) (0.097) (0.093) (0.092)

Age �0.005 0.007 �0.003 �0.003(0.004) (0.004) (0.004) (0.004)

Education 0.017 �0.081⇤⇤ �0.055 0.050(0.035) (0.036) (0.034) (0.034)

Constant �0.910⇤⇤⇤ �0.503⇤ �0.413 �0.270(0.277) (0.274) (0.265) (0.264)

N 1,994 1,994 1,994 1,994AIC 2,548.103 2,534.239 2,686.388 2,702.630⇤p < .1; ⇤⇤p < .05; ⇤⇤⇤p < .01

22

Table 15: Randomization check: ICSID

Elite Divided Elite Consensus Group Endorse Group Oppose

(1) (2) (3) (4)

Military assertiveness �0.103 0.164 0.059 0.172(0.247) (0.247) (0.237) (0.238)

Internationalism 0.122 �0.071 �0.283 0.353(0.266) (0.267) (0.256) (0.258)

Party ID 0.345⇤ �0.380⇤ �0.202 0.058(0.204) (0.205) (0.196) (0.197)

Male 0.071 �0.070 0.035 0.028(0.096) (0.096) (0.092) (0.093)

Age 0.007 0.002 �0.003 �0.003(0.004) (0.004) (0.004) (0.004)

Education �0.017 �0.007 0.050 �0.055(0.035) (0.035) (0.034) (0.034)

Constant �1.071⇤⇤⇤ �0.565⇤⇤ �0.270 �0.413(0.275) (0.274) (0.264) (0.265)

N 1,994 1,994 1,994 1,994AIC 2,543.611 2,550.857 2,702.630 2,686.388⇤p < .1; ⇤⇤p < .05; ⇤⇤⇤p < .01

2.3.1 Comparison of group cue treatments

Studies 1 and 2 build on Mann and Sinclair (2013) by using a social cue treatment that presents

respondents with the responses from other survey respondents who answered the previous survey

respondents like them. By using the “like you” language, we avoid the problem of selecting a pre-

defined reference group for participants, thereby letting participants define the relevant comparison

point for themselves rather than assuming they identify with other members of groups defined by

particular descriptive characteristics (as would be the case in treatments that emphasized what other

respondents of the same gender, or who resided in the same town, thought).

However, it also raises four sets of questions. First, it raises questions about the mundane

realism of the treatment: although participants are often presented with polling data summarizing

the views of others, they are rarely so micro-directed as to only reflect the responses of others

“like them”. When news articles present polling results, for example, the survey results presented

rarely varies depending on the individual reader! Second, it raises questions about the construct

validity: although many social networks tend towards homophily (e.g. McPherson, Smith-Lovin,

and Cook, 2001; Freelon, Lynch, and Aday, 2015), this tendency is far from universal (Huckfeldt,

Mendez, and Osborn, 2004; Gentzkow and Shapiro, 2011). Third, it raises questions about the

23

mechanisms driving the treatment e↵ects. The interpretation advanced in the main text is that the

group cues are social cues, operating by presenting information about the beliefs of other societal

actors. A more individualistic interpretation, however, might be that the results are being driven by

the words “like you”, which may produce pressures for attitudinal consistency, in which respondents

who have already expressed a certain set of political attitudes express viewpoints similar to those

of other individuals who also happen to share these attitudes. Fourth, it raises questions about the

comparability between the social cue and the elite cues, since the elite cues are not presented in the

form of responses of “elites like you.”

Thus, for experiments 4-5 in Study 3, we employ two types of group cues. As before, we include

both a group endorse cue, and a group oppose cue, in which participants are presented with a set of

survey marginals, along with a group control, in which no social cues are presented. Here, however,

we include two types of each group cue: a pair of treatments were the survey marginals are the views

of respondents who answered the previous set of survey responses like them, and a more generic social

cue where the survey marginals are simply presented as the views of other survey respondents. By

comparing these two sets of treatments, we can determine whether the results are being driven by

the “like you” wording.

We carry out this analysis in four steps. First, in Table 18 we run an OLS model with both

the original social cure (group endorse “like you” and group oppose “like you” and the generic

group endorse, and the generic group oppose (omitting the “like you phrasing”). From a visual

inspection, and a formal test of the equality coe�cients (F -test), we find no statistical di↵erence

in the coe�cients between the original (“like you”) social cues and the generic social cues. This

suggests that the e↵ects of our social cues are not driven by the wording “like you.” Second, we

estimate a series of Davidson-MacKinnon J tests to compare a model that includes a separate set

of indicator variables for each of the four social cues (group endorse “like you”, group oppose “like

you”, the generic group endorse, and the generic group oppose), and a model that pools the type of

social cue together (a pooled group endorse, and a pooled group oppose); it systematically fails to

find evidence that one model is better than the other.2

Third, we conduct a simple visual test, plotting the density distributions of our dependent

variable of interest for each experiment, conditioning on elite and group cues. If the results are

being driven by the “like you” language, we should see systematically di↵erent findings between

2For China, in an additive specification: t = �0.772, p < 0.44 for the full model, and t = 0.625, p < 0.532 for thepooled model; in an interactive specification: t = 1.094, p < 0.27 for the full model, and t = 1.506, p < 0.13 for thepooled model. For ICSID, in an additive specification: t = 1.031, p < 0.30 for the full model, t = 0.622, p < 0.53 forthe pooled model; in an interactive specification: t = �2.754, p < 0.006 for the full model, t = 2.873, p < 0.004 forthe pooled model. Thus, for three of the four tests, we fail to find evidence that the model fits significantly di↵er; forthe last test, we find they di↵er, but that neither one outperforms the other. The results from the rank-sum tests,and a visual inspection of Figure 8 confirm this pattern.

24

the salmon- and turquoise-colored distributions in each panel in Figure 8. Instead, we see that the

two sets of treatments track together, only deviating slightly in the elite consensus x group oppose

condition in the ICSID experiment (the middle panel in the bottom row of Figure 8(b).

Table 16: Rank-sum tests comparing the two types of group cues

Elite cue Social cue China experiment ICSID experimentControl Endorse p < 0.597 p < 0.743Consensus Endorse p < 0.773 p < 0.354Divided Endorse p < 0.903 p < 0.534Control Oppose p < 0.245 p < 0.869Consensus Oppose p < 0.697 p < 0.040Divided Oppose p < 0.606 p < 0.754

Fourth, we conduct more formal counterparts to the visual tests from above by estimating a

series of Wilcoxon rank-sum tests that explicitly compares each of the two distributions. The test

results further buttress the findings from the visual test, in that of the twelve comparisons being

made, only the distributions in the elite consensus x group oppose condition in the ICSID experiment

significantly di↵er from one another (p < 0.04). Given the sheer number of comparisons, and the

overall pattern of the distributions, we thus simplify the analyses presented in the main text by

pooling the generic social cue and “like you” social cue together.

25

Figure 8: Density distributions of group cues

(a) China experiment

Elite Control Elite Consensus Elite Divided

Gro

up

En

do

rse

Gro

up

Op

po

se

0 25 50 75 100 0 25 50 75 100 0 25 50 75 100

Support

Density Type of group cue

Like You

Generic

(b) ICSID experiment

Elite Control Elite Consensus Elite Divided

Gro

up

En

do

rse

Gro

up

Op

po

se

0 25 50 75 100 0 25 50 75 100 0 25 50 75 100

Support

Density Type of group cue

Like You

Generic

The overlapping density plots confirm the results of the Vuong test and rank sum tests described above, showingthat the two types of group cues have similar e↵ects to one another, such that we pool them in the main analysis in

the text. The findings thus suggest that the group cue treatment e↵ects are not being driven by the “like you”language used in Studies 1 and 2.

26

Tab

le17:SummaryStatisticsan

dSam

ple

CharacteristicsforStudy3

NNA

Min

Max

Median

Mean

Std.Dev.

Description

Male

1994

30

11.000

0.503

0.500

Age

1997

018

9033.000

35.878

11.443

Education

1997

01

85.000

4.230

1.366

From

nohighschoo

l(1)to

grad

-uatedegree(8).

Mean(4)is

As-

sociate’sdegree.

Party

ID1997

00

10.333

0.413

0.274

7-point

scale

normalized

tolie

between

0an

d1;

0(Stron

gDem

ocrat)

to1(Stron

gRepub-

lican).

Militan

tAssertiveness

1997

00

10.375

0.405

0.214

Militan

tAssertiveness

scale

(Herrm

ann,Tetlock,an

dVisser,

1999;

Kertzer

and

McG

raw,

2012);normalized

to0-1.

Internationalism

1997

00

10.625

0.591

0.194

Internationalism

scale

(Her-

rman

n,

Tetlock,

and

Visser,

1999;

Kertzer

and

McG

raw,

2012);normalized

to0-1.

Arm

edForce

ChinaScenario

1997

00

10.467

0.462

0.292

Supportforsendingmilitaryre-

sources

toAsia.

Con

tinu

ous0-10

normalized

to0-1.

0(Stron

gly

Oppose)

to1(Stron

glySupport).

SupportforIC

SID

1445

00

10.522

0.508

0.262

SupportforallowingUScitizens

be

subject

toIC

SID

.Con

tinu

-ou

s0-10

normalized

to0-1.

0(Stron

glyOppose)

to1(Stron

gly

Support).

27

Table 18: Study 3 Results (disaggregated by social cue)

China ICSID

(1) (2) (3) (4) (5) (6)

Elite Divided �0.031 �0.026⇤⇤ �0.026 �0.044⇤⇤⇤ �0.043⇤⇤⇤ �0.042⇤⇤⇤

(0.016) (0.013) (0.013) (0.014) (0.014) (0.014)Elite Consensus 0.063⇤⇤⇤ 0.072⇤⇤⇤ 0.071⇤⇤⇤ 0.047⇤⇤⇤ 0.046⇤⇤⇤ 0.046⇤⇤⇤

(0.016) (0.013) (0.013) (0.014) (0.014) (0.014)Group Endorse (“Like You”) 0.011 0.004 0.003 0.045⇤⇤ 0.045⇤⇤ 0.045⇤⇤

(0.020) (0.017) (0.017) (0.018) (0.018) (0.018)Group Endorse 0.022 0.011 0.013 0.042⇤⇤ 0.040⇤⇤ 0.040⇤⇤

(0.020) (0.017) (0.017) (0.018) (0.018) (0.018)Group Oppose (“Like You”) �0.094⇤⇤⇤ �0.094⇤⇤⇤ �0.095⇤⇤⇤ �0.064⇤⇤⇤ �0.069⇤⇤⇤ �0.068⇤⇤⇤

(0.020) (0.017) (0.017) (0.018) (0.018) (0.018)Group Oppose �0.088⇤⇤⇤ �0.093⇤⇤⇤ �0.093⇤⇤⇤ �0.075⇤⇤⇤ �0.077⇤⇤⇤ �0.077⇤⇤⇤

(0.020) (0.017) (0.017) (0.018) (0.018) (0.018)Militant Assertiveness 0.655⇤⇤⇤ 0.651⇤⇤⇤ 0.0004 0.004

(0.028) (0.028) (0.029) (0.029)Internationalism 0.142⇤⇤⇤ 0.141⇤⇤⇤ 0.286⇤⇤⇤ 0.289⇤⇤⇤

(0.030) (0.031) (0.031) (0.031)Party ID 0.065⇤⇤⇤ 0.060⇤⇤ �0.016 �0.015

(0.023) (0.023) (0.024) (0.024)Controls X X

F -test of equality between coe�cients on Social Cue Treatment Versions (1) (“Like You”) vs. (2) (Generic)Group Endorse Treatments 0.585 0.688 0.567 0.838 0.795 0.789Group Oppose Treatment 0.761 0.964 0.932 0.557 0.659 0.578N 1,997 1,997 1,994 1,997 1,997 1,994R2 0.049 0.308 0.310 0.060 0.107 0.108Adjusted R2 0.045 0.305 0.305 0.057 0.103 0.102⇤⇤p < .05; ⇤⇤⇤p < .01All regressions are OLS and control for the randomly assigned order of the scenarios (China Scenario or ICSID

Scenario first). Controls include Male, Age, and Education.

28

Table 19: Study 3 Results

China ICSID

(1) (2) (3) (4) (5) (6)

Elite Divided �0.030⇤ �0.074⇤⇤ �0.050 �0.043⇤⇤⇤ �0.011 �0.007(0.016) (0.036) (0.031) (0.014) (0.032) (0.031)

Elite Consensus 0.064⇤⇤⇤ 0.020 0.029 0.047⇤⇤⇤ 0.093⇤⇤⇤ 0.088⇤⇤⇤

(0.016) (0.034) (0.029) (0.014) (0.031) (0.030)Group Endorse 0.017 �0.008 �0.008 0.043⇤⇤⇤ 0.066⇤⇤ 0.065⇤⇤

(0.018) (0.030) (0.026) (0.016) (0.027) (0.027)Group Oppose �0.091⇤⇤⇤ �0.138⇤⇤⇤ �0.134⇤⇤⇤ �0.069⇤⇤⇤ �0.027 �0.031

(0.018) (0.030) (0.026) (0.016) (0.027) (0.026)Order �0.010 �0.008 �0.018 0.012 0.011 0.009

(0.013) (0.013) (0.011) (0.011) (0.011) (0.011)Military assertiveness 0.651⇤⇤⇤ 0.004

(0.028) (0.029)Internationalism 0.143⇤⇤⇤ 0.288⇤⇤⇤

(0.031) (0.031)Party ID 0.061⇤⇤⇤ �0.017

(0.023) (0.024)Male 0.025⇤⇤ �0.006

(0.011) (0.011)Age 0.0004 �0.0003

(0.0005) (0.001)Education �0.0002 �0.001

(0.004) (0.004)Divided ⇥ Endorse 0.048 0.025 �0.029 �0.033

(0.044) (0.037) (0.039) (0.038)Divided ⇥ Oppose 0.060 0.036 �0.050 �0.053

(0.043) (0.037) (0.039) (0.038)Consensus ⇥ Endorse 0.027 0.023 �0.037 �0.035

(0.042) (0.036) (0.038) (0.037)Consensus ⇥ Oppose 0.084⇤⇤ 0.086⇤⇤ �0.078⇤⇤ �0.072⇤

(0.042) (0.036) (0.038) (0.037)Constant 0.486⇤⇤⇤ 0.513⇤⇤⇤ 0.112⇤⇤⇤ 0.511⇤⇤⇤ 0.485⇤⇤⇤ 0.342⇤⇤⇤

(0.018) (0.025) (0.038) (0.016) (0.023) (0.039)N 1,997 1,997 1,994 1,997 1,997 1,994R2 0.048 0.051 0.313 0.060 0.062 0.110⇤p < .1; ⇤⇤p < .05; ⇤⇤⇤p < .01

29

Table 20: Study 3 Results among those who passed the manipulation check

cSupport iSupport

(1) (2) (3) (4) (5) (6)

Elite Divided �0.041⇤⇤ �0.084⇤⇤ �0.048 �0.048⇤⇤⇤ �0.016 �0.018(0.016) (0.037) (0.031) (0.015) (0.035) (0.034)

Elite Consensus 0.071⇤⇤⇤ 0.033 0.044 0.063⇤⇤⇤ 0.113⇤⇤⇤ 0.111⇤⇤⇤

(0.016) (0.035) (0.030) (0.015) (0.032) (0.031)Group Endorse 0.015 �0.007 �0.008 0.043⇤⇤⇤ 0.066⇤⇤ 0.066⇤⇤

(0.018) (0.030) (0.025) (0.017) (0.027) (0.027)Group Oppose �0.098⇤⇤⇤ �0.138⇤⇤⇤ �0.134⇤⇤⇤ �0.069⇤⇤⇤ �0.027 �0.031

(0.018) (0.030) (0.025) (0.017) (0.027) (0.026)Order �0.013 �0.012 �0.017 0.010 0.008 0.005

(0.013) (0.013) (0.011) (0.012) (0.012) (0.012)Military Assertiveness 0.683⇤⇤⇤ �0.016

(0.030) (0.031)Internationalism 0.133⇤⇤⇤ 0.307⇤⇤⇤

(0.032) (0.033)Party ID 0.051⇤⇤ �0.002

(0.024) (0.026)Male 0.024⇤⇤ �0.014

(0.011) (0.012)Age 0.0003 �0.0003

(0.001) (0.001)Education 0.001 0.001

(0.004) (0.004)Divided ⇥ Endorse 0.045 0.013 �0.031 �0.033

(0.045) (0.038) (0.042) (0.041)Divided ⇥ Oppose 0.061 0.033 �0.049 �0.048

(0.045) (0.038) (0.041) (0.040)Consensus ⇥ Endorse 0.028 0.023 �0.041 �0.040

(0.044) (0.037) (0.039) (0.038)Consensus ⇥ Oppose 0.067 0.072⇤ �0.086⇤⇤ �0.086⇤⇤

(0.044) (0.037) (0.040) (0.039)Constant 0.491⇤⇤⇤ 0.515⇤⇤⇤ 0.107⇤⇤⇤ 0.512⇤⇤⇤ 0.486⇤⇤⇤ 0.330⇤⇤⇤

(0.019) (0.025) (0.038) (0.017) (0.023) (0.041)N 1,837 1,837 1,835 1,758 1,758 1,755R2 0.057 0.058 0.332 0.069 0.071 0.123⇤p < .1; ⇤⇤p < .05; ⇤⇤⇤p < .01

30

2.3.2 Explaining variation in the e�cacy of elite cues

As the discussion in the main text indicates, although social cues display the strongest results across

all five studies, the e↵ects of elite cues are inconsistent, with weak or non-significant e↵ects in

Experiments 1-3, and stronger e↵ects in Experiments 4-5. That we find such inconsistent e↵ects

for elite cues is not unusual: Bullock (2011), for example, laments that the magnitude of variation

across experimental studies of the e↵ects of elite cues “makes generalization di�cult.”

Guisinger and Saunders (2017) o↵er a pair of mechanisms that might be able to account for this

variation, suggesting that the e↵ect of elite cues depends on two characteristics of the pre-existing

distribution of opinion. The first concerns ceiling e↵ects: if a high proportion of the sample already

agrees with the policy, there is less room for elite endorsements to bolster support.3 Of course,

there’s no reason why ceiling e↵ects should implicate elite cues in particular: if social cues can exert

a significant e↵ect in Experiment 3 but elite cues cannot, ceiling e↵ects are unlikely to blame. The

second concerns underlying polarization: partisan cues should exert stronger e↵ects on issues where

the underlying level of polarization is high.

We can test both of these hypotheses here. Following Guisinger and Saunders (2017), we first

calculated the baseline level of support for each policy in Experiments 3-5 (as measured by the

mean level of support of respondents in the elite control x group control condition).4 As shown

in the left-hand panel of Figure 9, contrary to their findings, Experiments 4-5, where elite cues

have stronger e↵ects, actually feature a higher level of baseline support, rather than a lower level; if

anything, the green and blue distributions for Experiments 4-5 are to the right of the red distribution

for Experiment 3, though the magnitudes are small. In this sense, there is little reason to suspect

ceiling e↵ects are artificially dampening the e↵ect of elite cues in Experiment 3.

In the right-hand panel of Figure 9, we calculate the baseline level of polarization among par-

ticipants in the group and elite control conditions for Experiments 3-5, dropping the independents

from each sample, and calculating the di↵erence between the average level of support for each pol-

icy among Republicans, minus the average level of support for each policy among Democrats, the

distributions of which are calculated here using B = 1500 bootstraps. Positive values thus indicate

policies more popular among Republicans than Democrats, and values further away from 0 indicate

greater degrees of polarization. Here, we find that the green and blue distributions representing

Experiments 4 and 5 show significantly more partisan polarization among respondents than the red

distribution representing Experiment 3. Consistent with Guisinger and Saunders (2017), then, this

3Guisinger and Saunders (2017) frame the mechanism as “the share of the population not already in alignmentwith elite opinion”, but since elite opinion in their study reflects the content of the elite cue being manipulated, thetwo are functionally equivalent.

4We focus on Experiments 3-5, because Experiments 1-2 do not have an elite control condition, precluding thepossibility of obtaining a baseline measurement free of cues.

31

Figure

9:Exp

erim

ents

4-5displayahigher

levelof

polarization,butnot

alower

levelof

baselinesupport

Bas

elin

e le

vel o

f sup

port

in g

roup

and

elit

e co

ntro

l

Density

-0.2

0.0

0.2

0.4

0.6

0.8

1.0

1.2

Bas

elin

e le

vel o

f pol

ariz

atio

n in

gro

up a

nd e

lite

cont

rol

-0.3

-0.2

-0.1

0.0

0.1

0.2

0.3

Follow

ingGuisinger

andSaunders(2017),

thetw

opanelsin

this

figure

test

twodi↵eren

tex

planationsforwhyth

ee↵

ectof

elitecu

esis

stronger

inExperim

ents

4-5

thanExperim

ent3.Theleft-handpanel

plots

thebaselinelevel

ofsu

pport

amongparticipants

inth

egroupandeliteco

ntrol,in

Experim

ent3(a

Chinaex

perim

ent,

show

nin

red),

Experim

ent4(a

revised

Chinaex

perim

ent,

show

nin

),andExperim

ent5(theIC

SID

experim

ent,

show

nin

blue);th

eth

reeden

sity

distributionsare

show

nalon

gwithvertica

llines

den

otingth

emea

nofea

chdistribution.Theplotsh

owsth

atifanyth

ing,th

ereis

ahigher

baselinelevel

ofsu

pport

inExperim

ents

4-5,su

chth

atth

ewea

ker

resu

ltsin

Experim

ent3ca

nnotbedueto

aceilinge↵

ect.

Theright-handpanel

plots

thebaselinelevel

ofpartisanpolariza

tionforea

chof

theth

reestudies(w

ithExperim

ent3in

red,Experim

ent4in

green

,andExperim

ent5in

blue,

asbefore;th

efurther

each

distributionis

from

theblack

dash

edve

rtical

linein

thecenterofth

efigure,th

ehigher

thelevelsofpolariza

tion.Thefigure

thussh

owsth

atExperim

ents

4-5

displaysignifica

ntlyhigher

leve

lsof

polariza

tionth

anExperim

ent3,su

ggestingwhyelitecu

esdisplaystronger

e↵ects

inth

efinaltw

oex

perim

ents.

32

higher degree of polarization suggests one reason why elite cues display stronger e↵ects in these two

experiments, which were fielded at the end of September during a highly polarizing Presidential

election. Indeed, at the end of the survey, one of our participants remarked how unusual it was for

Democrats and Republicans to be united on any given issue, rendering the elite consensus treatment

more costly than it would have been had the study been fielded further away from election day.

2.3.3 Subgroup analysis by trust and vote choice

Finally, Experiments 4-5 also allow us to o↵er further evidence in favor of our theoretical mechanisms.

One of our central critiques of top-down models of public opinion in foreign policy are that cues are

the most persuasive when they come from cuegivers you trust, and in an era when more Americans

are turning away from party politics (Krupnikov and Klar, 2016), trust in government is abysmally

low (Keele, 2007), and the most notable political events of the past year consist of populist anger

against the political establishment (whether manifested by Brexit, Donald Trump steamrolling his

way to the Republican nomination over the ardent objections of GOP elites, and so on), it seems

plausible that people might take cues from actors other than partisan political elites.

To seek additional evidence exploring our theoretical mechanisms, in Experiments 4-5 we in-

cluded a standard measure of trust in government borrowed from the American National Election

Survey (a sample item: “How much of the time do you think you can trust the government in Wash-

ington do do what is right—just about always, most of the time, or only some of the time?”). Given

that we were fielding our study a month out from a rather hotly contested Presidential election whose

contours have been shaped by anti-establishment sentiment, we also included a standard measure of

vote choice borrowed from a Bloomberg Politics poll (“If the general election were held today, and

the candidates were Hillary Clinton for the Democrats, Donald Trump for the Republicans, Gary

Johnson for the Libertarian Party, or Jill Stein for the Green Party, for whom would you vote?”). If

our theoretical story is correct, we should expect (i) respondents with less trust in government to be

less swayed by the elite cues in our experiment, and (ii) Donald Trump supporters to be less swayed

by the elite cues in our experiment, given the anger many of them tend to report about established

politicians on both sides in Washington.

Table 21 presents the results from a set of linear regression models estimating the e↵ects of our

elite and social cue treatments for both the China and ICSID experiments, while also controlling

for the order in which the experiments were fielded to account for any potential order e↵ects. The

first four columns in the table subset the sample by median-splitting the respondents into those

who express a low level of trust in government, compared to those who report a high level of trust

in government. The last four columns in the table subset the sample into those who reported

33

Table 21: Elite cues are three times stronger for Clinton supporters than Trump supporters

China ICSID China ICSIDLow trust High trust Low trust High trust Clinton Trump Clinton Trump

(1) (2) (3) (4) (5) (6) (7) (8)

Elite division �0.024 �0.047⇤ �0.037⇤⇤ �0.068⇤⇤ �0.071⇤⇤⇤ 0.041 �0.090⇤⇤⇤ 0.057⇤⇤

(0.019) (0.027) (0.016) (0.026) (0.020) (0.031) (0.018) (0.029)Elite consensus 0.056⇤⇤⇤ 0.112⇤⇤⇤ 0.060⇤⇤⇤ 0.003 0.088⇤⇤⇤ 0.044 0.061⇤⇤⇤ 0.056⇤

(0.018) (0.029) (0.016) (0.027) (0.020) (0.031) (0.018) (0.029)Group endorse 0.004 0.058⇤ 0.036⇤⇤ 0.074⇤⇤⇤ 0.015 0.059⇤ 0.050⇤⇤ 0.014

(0.021) (0.030) (0.018) (0.029) (0.023) (0.034) (0.020) (0.032)Group oppose �0.096⇤⇤⇤ �0.065⇤⇤ �0.085⇤⇤⇤ �0.019 �0.089⇤⇤⇤ �0.062⇤ �0.052⇤⇤⇤ �0.114⇤⇤⇤

(0.021) (0.031) (0.018) (0.028) (0.023) (0.034) (0.020) (0.032)Order �0.014 0.0001 0.016 �0.011 �0.011 �0.007 0.001 0.006

(0.015) (0.023) (0.013) (0.022) (0.017) (0.025) (0.015) (0.023)Constant 0.478⇤⇤⇤ 0.502⇤⇤⇤ 0.493⇤⇤⇤ 0.572⇤⇤⇤ 0.464⇤⇤⇤ 0.568⇤⇤⇤ 0.549⇤⇤⇤ 0.463⇤⇤⇤

(0.022) (0.031) (0.019) (0.030) (0.024) (0.035) (0.021) (0.033)N 1,512 481 1,512 481 1,003 543 1,003 543Adjusted R2 0.036 0.097 0.066 0.043 0.087 0.028 0.099 0.048

Max elite cue +8.0% +15.9% +9.7% +7.1% +15.8% +4.4% +15.1% +5.6%⇤p < .1; ⇤⇤p < .05; ⇤⇤⇤p < .01

an intention to vote for Hillary Clinton, and those who reported an intention to vote for Donald

Trump.5 For the China experiment, respondents with high levels of trust in government appear

to be more sensitive to elite cues than individuals with low trust in government; comparing the

elite consensus condition to the control, individuals who are high in trust in government display a

treatment e↵ect roughly two times larger than individuals who are low in trust in government, a

statistically significant di↵erence (p < 0.037). For the ICSID experiment, the di↵erences in treatment

e↵ects displayed between individuals with low and high trust are not statistically significant. When

we subset the results by vote choice, we find even more striking results. Here, a direct comparison

of the coe�cients is somewhat more complex, because the treatment e↵ects have di↵erent meanings

based on the subsample (e.g. for Trump supporters, the elite division treatment involves their

party being in favor and the outparty being opposed, while for Clinton supporters, the elite division

treatment involves their party being opposed and the outparty being in favor). Thus, we instead

calculate the maximum e↵ect of elite cues within each subsample, by estimating the largest contrast

for each (thus, for Clinton supporters, the max elite cue e↵ect is between elite division and elite

consensus; for Trump supporters, the max elite cue e↵ect is between the elite control condition and

the elite consensus condition). Here, we find that elite cues display a maximum e↵ect 3.6 times

bigger in the China experiment for Clinton supporters than for Trump supporters, and 2.7 times

bigger in the ICSID experiment for Clinton supporters than for Trump supporters.6 These results

5Johnson and Stein supporters are dropped from the analysis due to their small cell sizes.6Although it is plausible we would see larger cues for Trump supporters if we had a condition where elite Republicans

were explicitly opposed to the policy, given the magnitude of the other e↵ect sizes, it is unlikely such a treatmentwould su�ciently narrow the gap; in the China experiment, for example, in order for elite cues to exert as large an

34

thus o↵er additional evidence in favor of the theoretical account we present here.

2.4 Salience of foreign policy during survey periods

A possible concern about our study is perhaps voters are not necessarily ignorant of foreign a↵airs,

but simply that these issues are less central to most citizens’ daily lives. To measure how salient

foreign policy issues were during the period of our survey we turn to the polling aggregator web-

site PollingReport.7 Going back to July of 2015-November 2016, terrorism or national security

consistently ranked among the most important issues facing Americans, behind only the economy.8

• CNN/ORC poll from July 22- 25, 2015 ranked terrorism 3rd (12%) and foreign policy 5th

(10%)

• Quinnipiac University. July 23-28, 2015 ranked terrorism 3rd (12%) and foreign policy 4th

(9%)

• ABC News/Washington Post Poll. November 16-19, 2015 ranked terrorism 2nd (29%)

• CBS News Poll. April 8-12, 2016 ranked terrorism/Islamic extremism/ISIS 2nd (9%)

• NBC News/Wall Street Journal Poll May 15-19, 2016 ranked 2nd (21%)

• ABC News/Washington Post Poll. Sept. 5-8, 2016 terrorism/national security ranked 2nd

(19%)

• CBS News/New York Times Poll. Sept. 9-13, 2016 national security, terrorism ranked 2nd

(29%)

• CBS News/New York Times Poll. Oct. 28-Nov. 1, 2016 national security, terrorism ranked

2nd (28%)

Furthermore, a Gallup poll from January 21-25 2016, showed that both Democrats (82%) and

Republicans (92%) ranked terrorism and national security as “extremely” or “very important.”9

In this sense, although foreign policy issues may not be the sole concern of the mass public, it

nonetheless looms larger than some of the more pessimistic takes of public opinion in foreign policy

would allege.

e↵ect for Trump supporters as they do for Clinton supporters, the the e↵ect of the Republican oppose/Democratssupport treatment compared to the elite control would have to be at least �11.4%; in comparison, the e↵ect of elitedivision compared to the elite control is +4.1%, and the e↵ect of elite consensus compared to the elite control is+4.4%.

7See http://www.pollingreport.com/prioriti.htm.8These contain every poll where national security or terrorism were mentioned.9See http://www.gallup.com/poll/188918/democrats-republicans-agree-four-top-issues-campaign.aspx

35

References

Aday, Sean. 2010. “Leading the charge: Media, elites, and the use of emotion in stimulating rallye↵ects in wartime.” Journal of Communication 60 (3): 440–465.

Albertson, Bethany, and Shana Kushner Gadarian. 2016. “Did that Scare You? Tips on CreatingEmotion in Experimental Subjects.” Political Analysis Forthcoming.

Brader, Ted. 2005. “Striking a responsive chord: How political ads motivate and persuade voters byappealing to emotions.” American Journal of Political Science 49 (2): 388–405.

Braumoeller, Bear F. 2004. “Hypothesis testing and multiplicative interaction terms.” Internationalorganization 58 (04): 807–820.

Bullock, John G. 2011. “Elite Influence on Public Opinion in an Informed Electorate.” AmericanPolitical Science Review 105 (3): 496-515.

Freelon, Deen, Marc Lynch, and Sean Aday. 2015. “Online Fragmentation in Wartime: A Longitu-dinal Analysis of Tweets about Syria, 2011-2013.” Annals of the American Academy of Politicaland Social Science 659 (1): 166-179.

Gadarian, Shana Kushner. 2014. “Scary pictures: How terrorism imagery a↵ects voter evaluations.”Political Communication 31 (2): 282–302.

Gentzkow, Matthew, and Jesse M. Shapiro. 2011. “Ideological Segregation Online and O✏ine.”Quarterly Journal of Economics 126 (4): 1799-1839.

Guisinger, Alexandra, and Elizabeth N. Saunders. 2017. “Mapping the Boundaries of Elite Cues:How Elites Shape Mass Opinion Across International Issues.” International Studies QuarterlyForthcoming.

Herrmann, Richard K., Philip E. Tetlock, and Penny S. Visser. 1999. “Mass Public Decisions toGo to War: A Cognitive-Interactionist Framework.” American Political Science Review 93 (3):553-573.

Holsti, Ole R. 2004. Public Opinion and American Foreign Policy. Revised edition ed. Ann Arbor,MI: University of Michigan Press.

Huckfeldt, Robert, Jeanette Morehouse Mendez, and Tracy Osborn. 2004. “Disagreement, Am-bivalence, and Engagement: The Political Consequences of Heterogeneous Networks.” PoliticalPsychology 25 (1): 65-95.

Jost, John T., Jaime L. Napier, Hulda Thorisdottir, Samuel D. Gosling, Tibor P. Palfai, and BrianOstafin. 2007. “Are Needs to Manage Uncertainty and Threat Associated With Political Conser-vatism or Ideological Extremity?” Personality and Social Psychology Bulletin 33 (7): 989-1007.

Kahneman, Daniel. 2011. Thinking, Fast and Slow. Macmillan.Keele, Luke. 2007. “Social Capital and the Dynamics of Trust in Government.” American Journalof Political Science 51 (2): 241-254.

Kertzer, Joshua D., and Kathleen M. McGraw. 2012. “Folk Realism: Testing the Microfoundationsof Realism in Ordinary Citizens.” International Studies Quarterly 56 (2): 245-258.

Krupnikov, Yanna, and Samara Klar. 2016. Independent Politics: How American Disdain for PartiesLeads to Political Inaction. New York, NY: Cambridge University Press.

Lambert, Alan J., Laura D. Scherer, John Paul Schott, Kristina R. Olson, Rick K. Andrews,Thomas C. O’Brien, and Alison R. Zisser. 2010. “Rally E↵ects, Threat, and Attitude Change.”Journal of Personality and Social Psychology 98 (6): 886-903. 10.1037/a0019086.

Lerner, Jennifer S, Roxana M Gonzalez, Deborah A Small, and Baruch Fischho↵. 2003. “E↵ectsof Fear and Anger on Perceived Risks of Terrorism A National Field Experiment.” PsychologicalScience 14 (2): 144–150.

Mann, Christopher B., and Betsy Sinclair. 2013. “Voters Like You: Social Group Cuesin Behavior and Opinion.” Working paper. http://politics.as.nyu.edu/docs/IO/29116/

SocialGroupings_MannSinclair_03.23.13.pdf.McPherson, Miller, Lynn Smith-Lovin, and James M Cook. 2001. “Birds of a Feather: Homophilyin Social Networks.” Annual Review of Sociology 27: 415-444.

36

Wittkopf, Eugene R. 1990. Faces of internationalism: Public Opinion and American foreign policy.Duke University Press.

Zeitzo↵, Thomas. 2014. “Anger, Exposure to Violence, and Intragroup Conflict: A “Lab in theField” Experiment in Southern Israel.” Political Psychology 35 (3): 309–335.

37