2005 Domestic Audience Costs in International Relations_An Experimental Approach

7/28/2019 2005 Domestic Audience Costs in International Relations_An Experimental Approach

http://slidepdf.com/reader/full/2005-domestic-audience-costs-in-international-relationsan-experimental-approach 1/40

Domestic Audience Costs in International Relations:An Experimental Approach

Michael TomzStanford [email protected]

August 2005

Draft. Comments Welcome!

Abstract: When should threats and promises in international relations be viewed as credible?Recent theories suggest that credibility increases withdomestic audience costs: the penalty aleader would incur at home for making and then breaking foreign commitments. This paperprovides the first direct evidence of audience costs in military crises. The analysis, based onexperiments embedded in public opinion surveys, supports five conclusions: (1) audience costsexist under fairly general conditions; (2) these costs arise primarily because citizens areconcerned with the reputation of their leader/country; (3) such costs increase with the level of escalation; (4) audience costs are larger among certain demographic groups – Democrats, doves,isolationists, and the highly educated – than among others; and (5) the magnitude of costs variessystematically with contextual factors, especially power, interests, and political regime. These

findings help identify the conditions under which domestic audiences make commitmentscredible. At the same time, they demonstrate the promise of using experiments to answerpreviously intractable questions in the field of international relations.

Acknowledgements: This research was funded by a grant from Time Sharing Experiments in theSocial Sciences (TESS), an NSF Infrastructure Initiative. I am grateful to Skip Lupia, DianaMutz, and two anonymous reviewers for feedback on the grant proposal. Jim Fearon, JonKrosnick, Ken Schultz, Paul Sniderman, and Jessica Weeks provided helpful comments atvarious stages in the research.



1

How do leaders solve the problem of credibility in world affairs? Every day, it seems,

world leaders make threats and promises when dealing with other countries. They vow to take

military action or inflict economic sanctions against countries that refuse to meet certain

demands. They pledge to curtail pollution, uphold human rights, supply foreign aid, and lower

barriers to international trade and capital. Without a world government that compels leaders to

keep these kinds of commitments, many of which would be costly to carry out, why—and

when—should people take foreign leaders at their word?

The answer may lie at the intersection between foreign affairs and domestic politics. In a

seminal article, Fearon (1994) hypothesized that leaders would suffer “domestic audience costs”

if they issued a military threat and subsequently backed down. Citizens, he contended, would

think less of leaders who failed to follow through than of leaders who never threatened in the

first place. The prospect of losing support, or even office, gives leaders an incentive to avoid

making empty threats. In general, it is argued, the credibility of threats should increase with the

domestic political costs of threatening and not following through.

The concept of audience costs is now central to theories about the initiation and course of

military crises (e.g. Fearon 1994, 1997; Schultz 1999, 2001a; Smith 1998). Moreover, scholars

have incorporated the concept into models of alliance commitments (Gaubatz 1996; Smith

1996), foreign trade (Mansfield et al. 2002), economic sanctions (Dorussen and Mo 2001; Martin

1993), and international cooperation more generally (Leeds 1999; Lipson 2003; McGillivray and

Smith 2000).

Despite the prominence of audience costs in international relations theory, though, it is

still unclear whether—and under what conditions—such costs exist. Most empirical work on the

topic is indirect. Fearon (1994) conjectured that audience costs are lower in authoritarian



2

regimes than in democracies. Scholars have, therefore, checked for correlations between the

type of political regime and behavior during military crises (e.g. Eyerman and Hart 1996; Partell

and Palmer 1999). But such tests have ambiguous implications: it is not clear whether the

relationship between democracy and foreign policy is due to audience costs or some other

difference between the two types of regimes.

The alternative empirical strategy is to measure audience costs directly, but this approach

is vulnerable to problems of strategic selection bias (Baum 2004; Schultz 2001b). In theory,

leaders should take the prospect of audience costs into account when making foreign policy

decisions. Thus, in precisely those situations when the public would react harshly against

backing down, leaders will tend to avoid that path, leaving us little opportunity to observe the

public backlash. Audience costs will be largely invisible.1

This paper offers the first direct analysis of audience costs in a way that avoids problems

of strategic selection. The study is based on a set of experiments embedded in public opinion

surveys. In each experiment, the interviewer describes a military crisis. Some participants are

randomly assigned to a control group and told that the president does not get involved. Others

are randomly placed in a treatment condition in which the president escalates the crisis but

ultimately backs down. All participants in the between-subject design are then asked whether

they approve of the way the president handled the situation. I identify audience costs by

comparing approval ratings in the “stay out” and “back down” conditions. This approach not

1 Some researchers have quantified the effect of war on the tenure of leaders (e.g. Chiozza andGoemans 2004; Bueno de Mesquita and Siverson 1995; Goemans 2000). This research is centralto a new wave of work about the domestic effects of foreign policy, but does not settle the debateabout audience costs because it does not reveal what would have happened if the leader hadbacked down or had conceded without a fight.



3

only makes it possible to quantify audience costs, but it also overcomes the endogeneity

problems that contaminate observational data.

The remainder of the paper uses data from survey experiments to answer five questions.

First, do audience costs exist under fairly general conditions? Second, do such costs increase

with the level of escalation, thereby providing an effective signal in international military crises?

Third, do leaders who back down lose support from all voters, or does the erosion of support

occur within identifiable subsets of the electorate? Fourth, under what international

circumstances are audience costs likely to be largest, making escalation an especially credible

signal of intentions? Finally, what is the micro-level mechanismbehind audience costs in world

affairs?

1. Do Audience Costs Exist?

The first issue that needs systematic investigation is whether audience costs exist. To

make the goal of this inquiry completely transparent, Figure 1 depicts a simple model of crisis

bargaining with audience costs.2 In the model, two leaders from different countries disagree

about territory, policy, or some other item they both value. The interaction begins when Leader

One (L1) decides whether challenge Leader Two (L2) over the item, which has a normalized

value of 1. In armed crises, the challenge typically takes the form of an ultimatum: meet certain

demands or face military action.

[FIGURE 1 ABOUT HERE]

If L1 chooses not to challenge, she receives nothing and leaves her opponent with a

payoff of 1. If L1 challenges, on the other hand, her opponent has two options: concede or resist.

2 The model is adapted from Schultz (1999).



4

In the event of resistance, L1 can carry out the ultimatum, resulting in war with payoffswar1and

war2 for the two sides, or L1 can back down, in which case she incurs a domestic audience cost,

a, for having challenged before yielding the prize.

The literature contains three incompatible conjectures about the value of a, and we have

almost no evidence to settle the debate. Some analysts conjecture thata is positive—that citizens

would disapprove if their leader reneged on a public commitment (e.g. Fearon 1994; Guisinger

and Smith 2002; Smith 1998). Analysts offer two reasons for this kind of public response.

Perhaps citizens recognize the value of a good reputation and fear that hollow threats and

promises would undermine the country’s credibility, making it harder to bargain and cooperate

with other countries in the future. Or perhaps citizens perceive empty commitments as

dishonorable and deeply embarrassing. Either way, citizens would have grounds for penalizing

leaders who make foreign commitments and then renege.

The second conjecture says thata is approximately zero, because citizens would not

punish leaders for breaking foreign commitments. Scholars who have advanced this hypothesis

point out that many citizens pay little attention to foreign policy, and others judge leaders based

on whether the job gets done. According to Brody (1994: 210), “the public seems to respond to

[foreign] policy outcomes, not to the means of achieving them.” If citizens focus mainly on final

outcomes, rather than the sequence of threats and promises in medias res, we would not expect

audience costs. Domestic approval ratings would simply reflect the fact that the leader lost the

territory, policy, or other item under dispute. It would hardly matter whether the leader made

international threats before conceding the prize.

Moreover, some add that even the most attentive citizens may forgive leaders for making

false commitments. After all, anyone who has played poker understands that bluffing can be an



5

optimal strategy (e.g. Gowa 1999; Ramsay 2003, 2004; Schultz 1999:237; Slantchev 2004). By

exaggerating their resources or resolve, successful bluffers can achieve higher payoffs than

would be possible with complete sincerity. At the same time, skilled bluffers understand the

need for flexibility, the need to fold under certain circumstances. “Why, then, would

constituents punish leaders whose bluffs are sometimes called” (Gowa 1999:26)?3

According to a third conjecture, the parameter amight actually benegativebecause

domestic audiences reward effort. On this view, citizens think more highly of leaders who try

before conceding than of leaders who forfeit the prize at the outset. Walt (1999:34), for

example, points out that citizens may “reward a leader who overreaches at first and then manages

to retreat short of war. Thus the British and French governments did not suffer domestic

audience costs when they backed down during the Rhineland crisis of 1936 or the Munich crisis

of 1938, because public opinion did not support going to war.” Although Walt never proves the

counterfactual—that British and French leaders would have been significantly less popular if

they had never threatened in the first place—his historical examples raise an interesting

possibility: perhaps leaders can gain points by escalating before giving up, rather than conceding

immediately.

The political windfall from moving toward war could be especially large when important

issues are at stake or power imbalances exist. In Walt’s example, voters did not favor war with

Germany. If British and French citizens had been more alarmed about German expansion, they

might have commended leaders who (at least) tried to stop the Germans with threats or shows of

force, while castigating leaders who stayed out completely. Power disparities could have a

similar effect: “leaders of small states may berewardedfor escalating crises with big states and

3 For further skepticism about audience costs, see Desch (2002:29-32).



6

then backing down, where they would be castigated for simply backing down. Standing up to a

‘bully’ may be praised even if one ultimately retreats” (Fearon 1994:580).

Do citizens typically respond with scorn, indifference, or praise when their leaders

commit without following through? Until we know, we cannot truly understand the effects of

committing before a domestic audience. At present, the idea of domestic audience costs remains

“an interesting and intuitively plausible conjecture about crisis bargaining, one well worth

further exploration. Until it is rigorously tested, however, there is no way of knowing how

significant the actual contribution really is”(Walt 1999:35).

The theoretical and empirical stakes in this debate are high, because evidence about

audience costs could shape the future direction of international relations research. If audience

costs exist under fairly general conditions, this discovery would provide—for the first time—

empirical microfoundations for a broad class of models in international security and political

economy. The discovery would also suggest profitable directions for new research, especially if

the domestic price of flip-flopping varies systematically with characteristics of the situation and

the audience. If, on the other hand, citizens show no stronger preference for leaders who avoid

commitments than for leaders who commit but subsequently renege, we will need to rethink how

leaders send signals and tie hands in world affairs.

2. A Direct Test of Audience Costs

To study whether audience costs exist, I embedded an experiment in a public opinion

survey, which was administered the survey to a random sample of 917 US adults during July

2004. All participants in the internet-based survey received an introductory script: “You will

read about a situation our country has faced many times in the past and will probably face again.



7

Different leaders have handled the situation in different ways. We will describe one approach

US leaders have taken, and ask whether you approve or disapprove.” (The full text of the

experiment appears in the appendix).

Participants then read about a foreign military crisis in which “A country sent its military

to take over a neighboring country.” Of course, such military situations vary along many

dimensions that could affect public opinion. Was the invading country democratic? Why did it

attack? How strong was the invading army? Would the invasion affect US interests? To help

participants reach an informed judgment, I supplied answers to these kinds of background

questions.

Rather than hold the background information constant, though, I randomly manipulated

four contextual variables: regime, motive, power, and interests. The country was lead by a

“dictator” in half the interviews and a “democratically elected government” in the other half.

The attacker sometimes had aggressive motives—it invaded “to get more power and

resources”—and sometimes invaded “because of a longstanding historical feud.” To manipulate

power, I informed half the participants that the attacker had a “strong military,” such that “it

would have taken a major effort for the United States to help push them out.” In the remaining

cases, the attacker had a “weak military” that the U.S. could repel without major effort. Finally,

a victory by the attacking country would either “hurt” or “not affect” the safety and economy of

the United States.

By crossing these four contextual variables, each with two possible values, I generated a

full 2x2x2x2 factorial design. In essence, each respondent was randomly assigned to one of 16

possible background situations. There were two distinct advantages to including such a wide

range of situations. First, the diversity of scenarios contributed to making the findings more



8

general by reducing the risk that idiosyncratic features of a particular crisis were driving the

results. Second, the variation opened opportunities to analyze how external conditions affect the

magnitude of audience costs, a topic I explore in Section 5.

After reading the background information, participants learned how the US president

handled the situation. Half the respondents were randomly assigned to a control group, in which

the president traveled down the “no action” branch of the game tree in Figure 1. Respondents

did not see the game tree. Instead, they were told: “The US president said the United States

would stay out of the conflict. The attacking country continued to invade. In the end, the US

president did not send troops, and the attacking country took over its neighbor.”

Remaining respondents were put in the treatment condition, in which the president issued

a military threat but failed to carry it out. In terms of Figure 1, the president challenged, the

opponent resisted, and the president backed down. I conveyed this sequence by stating: “The US

president said that if the attack continued, the US military would push out the invaders. The

attacking country continued to invade. In the end, the US president did not send troops, and the

attacking country took over its neighbor.”

Finally, respondents received a set of bullet points that recapitulated the entire scenario.

The bullet points summarized the invader’s political regime (democracyversusdictatorship),

military power (strongversusweak), and motive (more power versushistorical feud), and the

likely effect of the invasion on the safety and economy of the United States. The bullets also

reminded citizens of the path the president took: “The US president said the United States would

stay out of the conflict” OR “The US president said that if the attack continued, the US military

would push out the invaders.” The last three bullet points were common to all participants: the



9

attacking country continued to invade, the US president did not send troops, and the attacking

country took over its neighbor.

I then asked respondents: “Do you approve, disapprove, or neither approve nor

disapprove of the way the US president handled the situation?” Those who approved or

disapproved were asked whether they held their view very strongly, or only somewhat. Those

who answered “neither” where prompted: “Do you lean toward approving of the way the US

president handled the situation, lean toward disapproving, or don't you lean either way?” The

answers to these questions implied seven levels of presidential approval, ranging from very

strong disapproval to very strong approval, and constitute the dependent variable in the study.

The experiment made it possible to measure audience costs cleanly and directly. By

design, the control and treatment groups differed in only one respect: whether the US president

escalated the crisis before letting the attacker take over its neighbor. For this reason, any

systematic difference in presidential approval was entirely due to path the president took, not to

variation in background conditions or the outcome of the crisis. Thus, I could quantify audience

costs by calculating approval in the “stay out” condition and subtracting approval in the “military

threat” condition.

Before computing audience costs, though, it seemed prudent to confirm that the “stay

out” and “military threat” groups were balanced on baseline covariates that could affect

presidential approval. I estimated a logistic regression in which the dependent variable was the

dichotomous treatment (stay out =0, military threat =1), and asked whether any demographic or

contextual variables predicted membership in the treatment group. Not one of the many

variables in the model – gender, age, education, income, attitudes toward internationalism and

the use of force, stakes for the United States, and the motive, power, and interests of the invader



10

– had a statistically significant effect on the probability of being in the treatment group. Based

on a likelihood ratio test, we cannot reject the hypothesis that the relationship between the

treatment andall baseline variables is zero.4

Having established that the treatment was random, I constructed an index of presidential

approval and used it to quantify audience costs. The index ran from 1 to 7, with higher values

indicating stronger approval of the way the president handled the situation. Responses spanned

the entire scale, with approximately 9 percent of citizens voicing the highest level of approval

and around 26 voicing the lowest possible level. Given that the attacking country took over its

neighbor, it is not surprising that more citizens disapproved than approved. The average score

was approximately 3.4, with a standard deviation of 2.0.

Table 1, which summarizes presidential approval by treatment group, provides strong

evidence of audience costs. The president who did not get involved scored 3.75 on the approval

scale, compared with 3.05 for the president who issued a threat but did not carry it out. The

difference between these two scores, 0.70, has an associated standard error of only 0.13, so we

can be highly confident (t-statistic 5.3) that the effect on approval did not arise by chance alone.

Moreover, a difference of this magnitude is substantively consequential: the cost of making an

empty threat in this experiment was greater than 10 percent of the total range of the approval

scale. Translated onto a typical 100-point scale, then, making the empty threat could cause

approval to drop by more than 10 percentage points.

[TABLE 1 ABOUT HERE]

4 The likelihood ratio test statistic, 5.34, is distributed chi-squared with 10 degrees of freedom. If all coefficients were zero, we would observe a test statistic that large with probability 0.87. Ialso estimated a logit model with a full set of first-order interactions. The LR test statistic had ap-value of 0.77.



11

Other analytical methods lead to the same conclusion. For example, I compared the

proportion of respondents who disapproved in the vatrious experiental conditions. When the

president stayed out, approximately 33 percent of the sample disapproved of the way the

president handled the situation, whereas 41 percent expressed an intermediate position (“neither

approve nor disapprove”) and the remainder approved. In contrast, 49 percent of the sample

disapproved when the president issued a “threat of force” but did not carry it out. Thus, even

with this mild treatment in which the president merely issued a verbal threat, disapproval swelled

by 16 points (standard error of 7 points). Other methods, including ordered probit regression,

propensity score matching, and tau-b calculations on the entire table all showed significant

audience costs.5

Overall, the experiment provides strong evidence that leaders who make threats and then

back down can suffer a significant drop in popularity at home. Other research has shown that

presidential approval in general, and opinions about foreign policy in particular, affect elections.

Moreover, leaders devote considerable effort to campaigning and speaking about the topic,

presumably because they regard it as politically important (e.g. Aldrich, Sullivan and Borgida

1989; Gelpi, Reifler and Feaver 2005). Of course, we cannot say with certainty how the fall in

popularity would affect the political fate of leaders. At a minimum, though, the experiment

documents a necessary—and heretofore unproven—condition for audience costs: that citizens

disapprove when leaders do not follow through.

This finding has important implications for international conflict. Precisely because it

would be costly to make threats and not carry them out, threats can be credible signals of resolve.

5 Propensity score matching can be helpful when treatment groups are not balanced on baseline covariatesthat could affect the behavior of interest (Imai 2005). That was not a concern in my study, though,because the “stay out” and “empty threat” groups had almost identical demographic profiles.



12

Thus, data from the experiment provide empirical foundations for signaling models in the

literature.

3. Do Costs Increase with the Level of Escalation?

The previous experiment established that even mild acts of escalation – making verbal

threats – expose leaders to potential audience costs. I now investigate whether audience costs are

even larger at higher levels of hostility. If so, leaders can send progressively stronger signals by

ratcheting the crisis to the next level. If, on the other hand, audience costs do not increase

continuously and strictly with the amount of escalation, leaders will need to find other ways to

convey resolve after they have taken the initial step of exchanging verbal threats.

The literature on militarized interstate disputed distinguishes three levels of escalation

prior to war (Jones, Bremer and Singer 1996). Threats, which are verbal indications of hostile

intent, typically take the form of ultimatums. The first experiment contained a standard threat:

the US president vowed that if the attack continued, the US military would push out the invaders.

According to data collected by Jones, Bremmer, and Singer, approximately 5 percent of

militarized interstate disputes did not progress beyond a threat to use force.

The next rung on the escalatory ladder is adisplayof force, defined as a military

demonstration without combat. Many government actions fall into this category: putting forces

on alert, mobilizing previously inactive forces, conducting military maneuvers, deploying ships

to a foreign region, violating foreign airspace, etc. Approximately 22 percent of known MIDs

have ended with this level of escalation.

Finally, theuseof force is defined as an active military operation against the foreign

target. In most historical examples, a country fires upon the armed forces, population, or



13

territory of another state. Force has also been used short of war to blockade countries, occupy

territory, seize materials and take prisoners. The vast majority of known MIDs since 1816 have

involved some use of force, and a relatively small share—around 2 percent—have led to what

the Correlates of War coders regard as full-blown interstate war.

Do leaders expose themselves to higher audience costs when they display or use force? I

investigated this question by extending the basic experiment from Section 2. The setup was

exactly the same as before. Participants read that a country invaded its neighbor, and

background information varied randomly along four dimensions: regime, motive, power, and

interests. The main innovation was to expand the set of presidential responses. In the previous

experiment, the president either stayed out or issued an empty threat. The follow-up experiment

included three additional presidential approaches.

In one of the new scenarios, the president displayed force before backing down.

Specifically, “The US president said that if the attack continued, the US military would push out

the invaders. He sent troops to the region and prepared them for war. The attacking country

continued to invade. In the end, the US president did not send our troops into battle, and the

attacking country took over its neighbor.”

In another scenario, the president not only threatened and deployed force, but also

“ordered US troops to destroy one of the invader’s military bases. US troops destroyed the base,

and no Americans died in the operation. The invasion still continued. In the end, the US

president did not order more military action, and the attacking country took over its neighbor.”

The final scenario involved the use of force with limited US casualties: “20 Americans died in

the operation. The invasion still continued. In the end, the US president did not order more

military action, and the attacking country took over its neighbor.”



14

Two features of these extra scenarios make them especially appropriate for testing the

hypothesis that audience costs increase with the level of escalation. First, the new scenarios

differ only in the approach the president took. In all other respects, including background

circumstances and the outcome of the crisis, the extra scenarios are identical to each other and to

the “stay out” and “military threat” vignettes discussed earlier. Second, the more hostile

scenarios nest the less hostile ones. In particular, the vignette about the display of force includes

a threat to use force, and the vignette about the use of force refers to previous attempts to

threaten and display power. Any extra audience costs that we observe should, therefore, be due

to layering-on higher levels of escalation.

Using Knowledge Networks, I administered the extra scenarios to 1036 adults. By

design, I gave approximately 40 percent the “display of force” vignette and split the remaining

60 percent evenly between the two “use of force” scenarios. The extra interviews took place in

July and November 2004. To increase sample size and make sure answers from those two

months were comparable, I presented the “stay out” scenario to an additional 213 adults in

November 2004. Perhaps due to changes in world events, the stay out scenario was slightly

more popular in November (approval 4.00) than in July (approval 3.75). Although the difference

was only marginally significant, with a p-value of 0.13, I adjust for it in subsequent analyses.

Without controlling for this time effect, audience costs in the pooled sample would appear even

larger.

Table 2 summarizes the audience costs associated with each level of escalation. I

estimated the costs by regressing presidential approval on a constant term and mutually exclusive

dummy variables for military threat, display of force, use of force without US casualties, and use

of force with minimal US casualties. I also included a fixed effect for the month of the



15

interview, to control for differences in baseline approval between July and November. The key

quantities of interest, the coefficients on the escalation variables, represent deviations in

presidential approval relative to the omitted (base) scenario in which the president stayed out of

the conflict. These coefficients were estimated to be negative, meaning approval was lower

when the president escalated and backed down than when the president did not get involved. For

ease of presentation and comparability with Table 1, though, I present the costs as positive rather

than negative numbers. Thus, larger numbers in Table 2 mean bigger audience costs.


The estimates show three clear patterns. First, audience costs unambiguously exist in all

four scenarios. The estimated costs range from 0.5 to 1.5 points on an approval scale that runs

from 1 to 7. Expressed in terms of a 100-point scale, the decision to escalate and back down in a

military crisis could cause public opinion to move between 8 and 25 percentage points. The

estimates are not only substantively large but also highly precise. In all four cases, the ratio of

the point estimate to the standard error is at least 3.7. The 95 percent confidence intervals

around the estimated effects are, therefore, well to the right of zero. We can be nearly certain

that, in all four scenarios, citizens prefer the president who conceded immediately to the

president who conceded after upping the ante.

Second, audience costs do not appear to increase smoothly and continuously with the

level of escalation. Based on existing models of audience costs, we would expect the president

who displayed force to pay a higher price than the president who merely threatened to use it. In

our data, though, the costs in these two scenarios are similar. If anything, threatening force

exposes the president to higher costs than displaying costs, though the estimated difference is

statistically insignificant at conventional levels. (A Wald test of the hypothesis that the threat



16

and display scenarios involve identical audience costs has a p-value of 0.18.) In summary, the

experiment provides no evidence that potential audience costs increase as the president moves

from threatening to displaying force. This finding is somewhat surprising. If replicable, it will

have significant implications for empirical and theoretical work on military crises.

Third, although audience costs do not grow smoothly with the level of escalation, they do

exhibit a clear upward trend. Audience costs jump to 1.14 points when the president backs down

after using force without US casualties, and the costs increase by another 0.39 points when

Americans die in the operation. These changes are not only substantial but also statistically

significant at the 0.05 level or better. Thus, the experiment suggests two clear thresholds for the

accumulation of audience costs. The use of force exposes the president to much higher audience

costs than either threatening or displaying force, and the loss of American lives further raises the

price of escalating and then backing down.

4. How Are Audience Costs Distributed across the Electorate?

When the president escalates and backs down, does approval fall uniformly across the

electorate, or does the shift in public opinion occur mainly within identifiable demographic

groups? This question is important because different leaders have different constituencies. If the

distribution of audience costs is lumpy, then identical behavior may have distinct political

consequences depending on the support base of the leader in office. I explore this possibility by

slicing the data along demographic lines.

On a wide range of domestic and foreign policy issues, women tend to have different

preferences then men. Women show higher support for welfare, they tend to be more dovish and

isolationist in foreign affairs, and they are more protectionist on questions of international trade.



17

It therefore seems important to ask whether women and men have different views about

escalating and backing down. The gender distribution of the pooled sample (2163 respondents)

was fairly balanced, with 52 percent women and 48 percent men. To know whether gender

makes a difference, I computed audience costs for these two halves of the sample.

Interestingly, gender seems to have no effect on audience costs. In the pooled sample,

which includes all four types of escalation, the average audience cost among males was

approximately 0.87 points with a standard error of 0.14. The cost among females was 0.003

points smaller, a difference statistically and substantively indistinguishable from zero (see Table

3). I next asked whether women and men react differently to specific types of escalation. In

fact, men and women responded identically not just in the aggregate, but also to each scenario

separately. In a multivariate regression, the interactions between gender and the four levels of

escalation (threat of force, display of force, use of force without casualties, use of force without

casualties) were jointly insignificant, having an F4,2152 statistic of 0.20 and an associated p-value

of 0.94.


I next asked whether audience costs increase with the education of the respondent. A

large body of literature suggests that the most sophisticated members of the electorate think and

behave differently from those with less knowledge of public affairs. They tend to package their

views in more coherent ways, and they have a better understanding of the costs and benefits of

government policies. In the area of international finance, Tomz (2004) shows that highly

educated citizens are most sensitive to the reputational consequences of defaulting on loans from

foreigners. I expected to find the same pattern in military affairs. The most educated citizens are

best able to recognize how escalating and backing down could undermine national credibility,



18

with potentially adverse consequences for future interactions with countries. For this reason,

audience costs should be larger among more educated segments of the population.

To test this hypothesis I computed audience costs for respondents with and without

college degrees (see Table 3). Relative to staying out, the president who escalated and backed

down lost 1.38 approval points among citizens who had completed college and 0.71 points

among those who had not. The implied difference, 0.67 points on a seven-point approval scale,

was both substantively and statistically significant. I next checked whether the difference held

for each type of escalation. In the data, citizens with and without college degrees responded

about the same when the president backed down from a mere threat,

6

but attitudes diverged

sharply at higher levels of escalation. The difference in audience costs between the two

educational groups ranged from 0.62 points (standard error 0.29) when the president displayed

force to 1.16 points (standard error 0.32) when American soldiers died. Thus, on average and

within all but one of the escalatory scenarios, highly educated citizens are most sensitive to “flip

flopping” in a military crisis.

In addition to emphasizing demographic variables like gender and education, previous

research has suggested that citizens view foreign policy through two interpretative frameworks:

hawk versus dove and interventionism versus isolationism (e.g. Holsti 1996; Wittkopf 1990). I

hypothesized that audience costs would be larger among doves and isolationists. The reason is

straightforward. When the president stays out, these “soft-line” constituencies should express

higher approval than “hard-line” constituencies. Moreover, when the president threatens or takes

military action, soft-liners should object more strongly than hard-line counterparts. Thus, the

6 Costs were only 0.14 points larger (standard error 0.29) in the college-degree subsample,



19

difference in approval between the stay-out and escalation scenarios should be much larger for

doves and isolationists.

To measure hawkishness versus dovishness, I asked respondents how much they agreed

or disagreed with the following statement: “The use of military force only makes problems

worse.” I used a second statement to assess internationalism: “The United States needs to play

an active role in solving conflicts around the world.” In both cases, participants indicated

whether they agreed strongly, agreed somewhat, neither agreed nor disagreed, disagreed

somewhat, or disagreed strongly. In the analysis below, I focus on respondents who felt strongly

one way or the other.

Table 3 confirms that audience costs were indeed larger among soft-liners. Strong doves

(those who strongly agreed that the use of military force only makes problems worse) docked the

president by nearly 1.8 points for escalating and backing down. That estimate is tantamount to

thirty percentage points on a standard 100-point scale. In contrast, strong hawks assessed an

average audience cost of approximately 0.8 points. Most of the striking difference between these

two groups stemmed from baseline attitudes about the president who stayed out. Strong doves

gave that president an approval score of 4.5, whereas strong hawks felt the president deserved a

3.6. Contrary to expectation, hawks and doves had quite similar opinions about leaders who

escalated and backed down.7 Thus, doves exhibited larger audience costs because they felt better

about leaders who stayed out, not because they felt worse about leaders who escalated and

backed down.

Table 3 also presents audience costs for two levels of internationalism. This additional

analysis seemed warranted because the hawkishness and internationalism correlated at only 0.25

7 Hawks assigned an approval score of 2.8, whereas doves assigned a score of 2.7. In our data,the difference was not statistically significant at conventional levels.



20

when both were measured on 5-point scales. Notwithstanding this difference, a now-familiar

pattern emerged: audience costs were noticeably larger among strong isolationists (1.72) than

among strong internationalists (1.02). As before, the difference arose almost entirely from

distinct opinions about the “stay out” president, who was far more popular among isolationists

than among internationalists. In contrast, internationalists and isolationists offered

approximately the same assessment of leaders who escalated and backed down. It seems likely

that the two groups reached the same conclusion for different reasons, though: soft liners disliked

the president for escalating, whereas hardliners disliked the president for backing down. I leave

this as a topic for future research.

Finally, Table 3 gives audience costs as a function of political party affiliation. On

average, strong Democrats penalized the president by 1.12 points for escalating and backing

down, whereas strong Republicans lowered their support by only 0.53 points. The previous

analysis of hawks versus doves helps explain this systematic difference. In our sample, strong

democrats are 4 times more likely to be strong doves than strong hawks, and strong republicans

are 40 times more likely to take a strong hawkish position than a strong dovish one. If

republicans tend to be hawks, and if hawks have smaller audience costs, it makes sense that

audience costs should be smaller among Republican constituents than among Democratic ones.8

These findings have potentially important implications for the credibility of signals in

military crises. If presidents depend mainly on support from constituents who share their

partisan affiliation, then escalation should be a more informative signal of resolve under

Democratic presidents than under Republican ones. After all, Democratic presidents tend to rely

8 Strong Republicans are also 4 times more likely to be strong internationalists than to be strongisolationists, whereas strong Democrats are split evenly between internationalism andisolationism.



21

on the subset of citizens who are most likely to impose audience costs, so they should pay the

biggest price for escalating and backing down. It follows that an act of escalation conveys more

information when it comes from a Democratic president than from a Republican one.

The findings also have implications for cross-country analysis. Escalation should be

most informative when done by countries with especially dovish, isolationist, or educated

electorates. These high-audience-cost democracies should need fewer escalatory steps to signal

their intentions, should be less likely to launch limited military probes, and should be less likely

to back down in crises they actually enter. By sending highly credible signals of resolve, these

countries will be especially successful at persuading opponents to back down.

5. How Do Audience Costs Vary with International Circumstances?

The experiments not only shed light on the existence and incidence of audience costs, but

they also reveal how the international context affects the costs of escalating and backing down.

In the experiments, I varied not only the behavior of the president but also the circumstances

under which the crisis took place. The invading countries differed in power, motive, and

political regime, and US interests were at stake in some scenarios but not in others. Until now I

have averaged over this contextual variation. In this section I test hypotheses about the effects of

context while integrating over demographic groups and escalation levels.

I hypothesized, in particular, that audience costs would vary inversely with the national

interests of the escalating state. Citizens are most likely to demand action when their own safety

and welfare are at risk. Consequently, the president who says out should be less popular when

the invasion would hurt their own country than when it would not. If citizens have similar

opinions about the president who escalated and backed down, regardless of circumstances, but



22

have a lower view of leaders who stay out when the national interest is at stake, then audience

costs will move inversely with the national interest.

Table 4 provides some support for this hypothesis. On average, the cost of escalating and

backing down was 1.07 approval points (on the 7-point scale) when victory by the foreign power

would not affect the United States, compared with 0.68 points when the safety and economy of

the United States would suffer. The resulting gap of 0.39 points between these two scenarios

was both politically consequential and statistically distinguishable from zero. Further analysis

reveals that, on average, citizens approved less of the president when the invasion would hurt US

interests. This was the case not only when the president stayed out, but also when the president

escalated and backed down.


Table 4 also shows that audience costs are larger when the crisis involves a strong

adversary than when the crisis involves a weak one. It is not yet clear, however, whether this

finding would generalize to other countries and time periods. Both militarily and economically,

the United States is the strongest country in the world. As Fearon (1994) speculates, the

relationship might be reversed in smaller countries, where leaders could win approval (or suffer

smaller costs) for escalating against bullies than for escalating against weak states.9

Finally, Table 4 shows that audience costs do not seem to vary with the motive or

political regime of the adversary. I estimate somewhat higher costs when the US confronts a

democracy, and when the US confronts a country that does not clearly have aggressive motives.

9 It is not possible to test this conjecture with a survey of US citizens, but I found support for it ina separate study of audience costs in Argentina.



23

The differences in this survey are fairly small, though, and are not statistically distinguishable

from zero at conventional levels of confidence.

A natural extension of this analysis would be to check for higher-order interaction effects,

rather than taking each contextual variable in isolation. Preliminary tests suggest that audience

costs are nearly zero when the United States confronts a strong, aggressive dictator whose

behavior could have adverse consequences for the United States. If this profile seems to describe

Iraq under Saddam Hussein, we may have one reason why the United States, through acts of

escalation, did not convince Hussein to back down in the first and second Gulf Wars.

6. Why Do Citizens Disapprove?

To deepen the research, I next probed the micro-level mechanism: whydo citizens

disapprove of leaders who escalate but do not follow through? Evidence comes from a separate

survey of 347 citizens in June and July 2004.10

It many respects, this small survey resembled the larger one I have discussed so far.

Citizens were told about a situation in which a country invaded its neighbor. Some learned that

the president stayed out, whereas others learned that the president escalated the crisis but did not

follow through. In all cases, the vignette ended with the attacking country taking over its

neighbor.

This survey differed from the main instrument in two ways, however. Most importantly,

this extra survey asked citizens to explain the opinions they expressed. After indicating, in a

closed-end format, the degree to which they approved of the way the president handled the

situation, participants received a followup: “Could you please type a few sentences telling us

why you [disapprove very strongly OR disapprove somewhat strongly OR …] of the way the US

10 The study dates were June 11-17, 2004 (N=131) and July 2-14, 2004 (N=216).



24

president handled the situation? Participants entered their answers directly into a text box, which

allowed me to analyze each respondent’s account in his or her own words.

For manageability, this study of motivations contained fewer experimental manipulations

than the main instrument. The president either stayed out, or he displayed force before backing

down. Thus, the survey did not contain cases in which the president issued a mere empty threat,

or when the president used force to destroy an enemy base. Moreover, the survey presented a

narrower set of background conditions: the attacking country was always described as having a

strong military, and citizens did not receive information about the motives or political regime of

the invader. I did randomize the “interest” variable, however: citizens were told that an invasion,

if left unchecked would either hurtor not affect the safety and economy of the United States.

This study, like the larger experiment, found strong evidence of audience costs. The

president who stayed out achieved an approval score of 3.74, while the president who escalated

and backed down rated only 2.73 in the public’s eye. Thus, the president who changed course

incurred an audience cost of 1 point on the 7-point scale. This estimate is approximately five its

standard error and has an associated confidence interval from .60 to 1.43. Thus, the experiment

corroborated our main findings about the existence of audience costs.11

More tellingly, the survey revealedwhyaudience costs exist. In the study, 185 citizens

considered a scenario in which the president escalated and backed down. Of these, 105

disapproved—either strongly or somewhat—of the way the president handled the situation. Why

did they view the president’s behavior so negatively? Some did not say, and a few

misunderstood the follow-up question or provided an answer that was impossible to classify.

11 This is a somewhat larger cost than the value of 0.52 we observed on average in Table 2. Whythis difference? The text was a bit shorter, so backing down might have appeared starker.Moreover, the adversaryin this study always had a strong military. As noted in section 5,audience costs increase with the military strength of the adversary.



25

Nevertheless, 87 of the 105 people clearly explained why they disapproved of the way the

president behaved.

I classified the 87 open-ended responses into three categories. The first category (16

percent) included people who thought the president should have pushed out the invaders, not

because the president had made a prior commitment, but simply because it was the right thing to

do. Some said the United States had a moral obligation to protect the victims of aggression;

others pointed out that the safety and economy of the United States would suffer if the invader

took over its neighbor. Presumably, the people who answered this way would have objected just

as much, and for the same reasons, if the president had stayed out. In fact, citizens in the control

group, where the president neither threatened nor showed force, explained their disapproval in

precisely these terms. Because the reasons apply equally to all scenarios in which the president

failed to push back the invaders, they cannot be the source of audience costs.

The next category (12 percent) contained citizens who disliked the fact that the president

escalated in the first place. Some contended that it was not America’s responsibility to solve

other countries’ problems (“I do not feel that the USA should be the police for the world. We

should not have sent troops in this situation.”) Others argued the US government should focus

on its own citizens (“The US has enough problems of our own at this time. We have people that

are homeless and hungry. We should take care of our own first.”). These responses help us

understand an important empirical regularity: audience costs tend to be higher among doves,

isolationists, and Democrats than among other demographic groups.

The vast majority (72 percent) of respondents gave a third type of reason for

disapproving: the president behaved inconsistently. Typically, they complained that

inconsistency would hurt the reputation and credibility of the country. As one citizen noted: “It



26

was inconsistent. If you say that you are going to do something, you need to do it or else you lose

your credibility. It would have been better to ignore the situation completely than to make a

public commitment and then not carry it out.” Another respondent explained: “When a

President says something, in this case that he will push back the invading country, he must

follow through or lose credibility in the world community. He sent troops and when the threat

didn't work, he allowed the invasion to continue. That is a terrible precedent to set.”

Interestingly, not all respondents in the third category emphasized the reputational costs

of inconsistency. Two participants complained that the president had wasted money by

deploying troops but not using them, and one committed a sunk-cost fallacy (“If we spend the

money to prepare the troops and then send them over there, let them do the job they went to do.

Why waste money that could be used for something else such as healthcare? Do what you went

there to do and pullout.”) Moreover, eight participants inferred that the president had not shown

sufficient foresight (“United States President must not have truly thought things through”), or

they found the president’s behavior puzzling (“Why would he have troops there to help and not

do anything to help? That seems wrong.”)

Overall, though, the open-ended responses were most consistent with reputation-based

theories of audience costs. As in the model by Guisinger and Smith (2002), citizens seem to act

as reputational watchdogs that disapprove of—and perhaps even remove—leaders who sully

their reputation by making commitments and not following through.

7. Conclusions

This paper has offered the first direct study of audience costs in a way that avoids

problems of strategic selection. The analysis, based on a set of experiments embedded in public



27

opinion surveys, supports five conclusions: (1) audience costs exist under fairly general

conditions; (2) these costs arise primarily because citizens are concerned with the reputation of

their leader/country; (3) such costs increase with the level of escalation; (4) audience costs are

larger among certain demographic groups – Democrats, doves, isolationists, and the highly

educated – than among others, and (5) the magnitude of costs varies systematically with

contextual factors, especially power, interests, political regime.

These findings have both substantive and methodological implications for the study of

international relations. Substantively, they confirm that domestic audiences can enhance the

credibility of international commitments by punishing leaders who say one thing but do another.

This discovery was far from preordained. If citizens had focused on foreign policy outcomes

rather than processes, if they had rewarded leaders for trying before conceding, or if they had not

cared about their country’s reputation, audience costs would not have emerged. The fact that

audience costs arose consistently, across a wide range of experimental conditions, counts as

strong evidence that domestic actors can contribute to foreign credibility. Consequently, the

experiments in this paper supply microfoundations for many leading theories in international

security and political economy.12

At the same time, the experiments help specify the conditions under which the gain in

credibility is likely to be greatest. In military affairs, for example, audience costs should be

largest—and escalation should be most informative—when the country has a relatively dovish or

educated population, or when the head of state depends on constituents with those proclivities.

In contrast, escalation conveys less information about resolve when the national interest is at

12 For further evidence on this point, see Tomz (2004), who finds that domestic audiences aresensitive to the reputational costs of reneging on international debt contracts.



28

stake or when the adversary is an aggressive dictator, because inaction under those circumstances

would be nearly as unpopular as making commitments and then backing down.

Through the use of public opinion polls and census data, it should be possible to classify

countries (and key constituencies such as political parties or interest groups) according to their

foreign policy attitudes and demographic makeup. These variables, in turn, could become

predictors in future models of international crises. Where potential audience costs are high, e.g.

in countries with dovish or educated electorates, leaders should to be more circumspect about

issuing threats, but they should also more successful in convincing opponents to back down once

a crisis has started. The analysis also suggests new ways to understand how standard variables

such as power, stakes, regime type, and motive affect audience costs and, therefore, the outcome

of crises.

Finally, this paper demonstrates the promise of using public opinion surveys with

embedded experiments to answer central questions in the field of international relations.

Scholars increasingly recognize that foreign policy emerges from two-level games in which

leaders respond not only to foreign adversaries but also to domestic constituents. Survey-based

experiments can contribute to this emerging research agenda by measuring domestic preferences

and beliefs directly, while at the same time overcoming vexing problems of strategic selection.



29

Figure 1: A Crisis Bargaining Model with Audience Costs

No action

Challenge

(0, 1)

Concede

Resist

(1, 0)

Standfirm

Back

down(-a, 1)

(war1,war2)

L1

L2

L1



30

Table 1: The Domestic Political Cost of Making Empty Threats (Numbers in the table are presidential approval scores on a scale from 1 to 7)

Mean Std Err

Stay out 3.75 0.09 3.56 - 3.93

Empty threat 3.05 0.09 2.87 - 3.23

Difference 0.70 0.13 0.44 - 0.96

95% CI

Note: Estimates based on 437 cases of stay out and 477 casesof back down.



31

Table 2: Audience Costs at Four Levels of Escalation(Numbers in the table are the estimated loss of approval)

Estimate Std Err

Threat of force 0.70 0.13 0.44 - 0.96

Display of force 0.52 0.14 0.25 - 0.80

Use w/o US casualties 1.14 0.18 0.79 - 1.49Use w/ US casualties 1.53 0.17 1.20 1.87

95% CI

Note: Estimates based on a regression with 2163 observations. The regression included fixed effects to control for the timewhen the survey was administered.



32

Table 3: Audience Costs in Subsets of the Electorate(Numbers in the table are the estimated loss of approval)

Difference (std error)

Men 0.87 (0.14)

Women 0.86 (0.13)

College degree 1.38 (0.19)

No college degree 0.71 (0.11)

Strong dove 1.79 (0.39)

Strong hawk 0.79 (0.29)

Strong isolationist 1.72 (0.33)

Strong internationalist 1.02 (0.32)

Strong Democrat 1.12 (0.22)

Strong Republican 0.53 (0.25)

0.70 (0.48)

0.59 (0.33)

0.67 (0.22)

1.00 (0.48)

Estimate (std error)

0.00 (0.19)

Note: Estimates based on regressions with 2163 observations.Regressions included fixed effects to control for the time whenthe survey was administered.



33

Table 4: Audience Costs Conditional on Context(Numbers in the table are the estimated loss of approval)

Difference (std error)

Interests

Not affect the U.S. 1.07 (0.13)Would hurt the US 0.68 (0.13)

Power

Strong military 1.10 (0.13)

Weak military 0.64 (0.13)

Motive

Historical feud 0.95 (0.13)

More power 0.80 (0.13)

Regime

Democracy 0.95 (0.13)

Dictatorship 0.79 (0.13)

Estimate (std error)

0.39 0.18

0.45 (0.19)

0.14 (0.19)

0.16 (0.19)

Note: Estimates based on regressions with 2163 observations.Regressions included fixed effects to control for the time whenthe survey was administered.



34

References

Aldrich, John H., John L. Sullivan, and Eugene Borgida. 1989. “Foreign Affairs and IssueVoting: Do Presidential Candidates “Waltz Before a Blind Audience?” American PoliticalScience Review83:123-141.

Baum, Matthew A. 2004. “Going Private: Public Opinion, Presidential Rhetoric, and theDomestic Politics of Audience Costs in U.S. Foreign Policy Crises.” J ournal of ConflictResolution48, no. 5 (October): 603-31.

Brody, Richard A. 1994. “Crisis, War, and Public Opinion: The Media and Public Support forthe President.” Taken by Storm: The Media, Public Opinion, and US Foreign Policy in the Gulf War, eds. W. Lance Bennett and David L. Paletz, pp. 210-27. Chicago: Univ. of Chicago Press.

Bueno de Mesquita, Bruce and Randolph M. Siverson. 1995. “War and the Survival of PoliticalLeaders: A Comparative Study of Regime Types and Political Accountability.” American

Political Science Review89, no. 4 (December): 841-55.

Chiozza, Giacomo and H. E. Goemans. 2004. “International Conflict and the Tenure of Leaders:Is War Still Ex Post Inefficient?” American journal of Political Science48, no. 3 (July): 604-19.

Desch, Michael C. 2002. “Democracy and Victory: Why Regime Type Hardly Matters.”International Security27, no. 2 (Fall): 5-47.

Dorussen, Han and Jongryn Mo. 2001. “Ending Sanctions: Audience Costs and Rent-Seeking asCommitment Strategies.” J ournal of Conflict Resolution45, no. 4: 395 - 426.

Eyerman, Joe and Robert A. Hart, Jr. 1996. “An Empirical Test of the Audience CostProposition: Democracy Speaks Louder than Words.” J ournal of Conflict Resolution40, no. 4(December): 597-616.

Fearon, James D. 1994. “Domestic Political Audiences and the Escalation of InternationalDisputes.” American Political Science Review88, no. 3 (September): 577-92.

Fearon, James D. 1997. “Signaling Foreign Policy Interests: Tying Hands versus SinkingCosts.” J ournal of Conflict Resolution41, no. 1 (February): 68-90.

Gaubatz, Kurt Taylor. 1996. “Democratic States and Commitment in International Relations.”International Organization50, no. 1 (Winter): 109-39.

Gelpi, Christopher, Jason Reifler and Peter Feaver. 2005. “Iraq the Vote: Retrospective andProspective Foreign Policy Judgments on Candidate Choice and Casualty Tolerance.” WorkingPaper, Duke Univ.

Goemans, H. E. 2000. “Fighting for Survival: The Fate of Leaders and the Duration of War.” J ournal of Conflict Resolution44, no. 5 (October): 555-79.



35

Gowa, Joanne. 1999. Ballots and Bullets: The Elusive Democratic Peace. Princeton, NJ:Princeton Univ. Press.

Guisinger, Alexandra and Alastair Smith. 2002. “Honest Threats: The Interaction of Reputaiton

and Political Institutions in International Crises.” J ournal of Conflict Resolution46, no. 2 (April2002): 175-200.Holsti, Ole R. 1996. Public Opinion and American Foreign Policy. Ann Arbor: Univ. of Michigan Press.

Imai, Kosuke. 2005. “Do Get-Out-the-Vote Calls Reduce Turnout? The Importance of StatisticalMethods for Field Experiments.” American Political Science Review99:283-300.

Jones, Daniel M., Stuart A. Bremer, and J. David Singer. 1996. "Militarized Interstate Disputes,1816-1992: Rationale, Coding Rules, and Empirical Patterns." Conflict Management and PeaceScience15, no. 2: 163-215.

Leeds, Brett Ashley. 1999. “Domestic Political Institutions, Credible Commitments, andInternational Cooperation.” American Journal of Political Science43, no. 4 (October): 979-1002.

Lipson, Charles. 2003. Reliable Partners: How Democracies Have Made a Separate Peace.Princeton, NJ: Princeton Univ. Press.

Mansfield, Edward D., Helen V. Milner, and B. Peter Rosendorff. 2002. “Why DemocraciesCooperate More: Electoral Control and International Trade Agreements.” InternationalOrganization56, no. 3 (Summer): 477-513.

Martin, Lisa L. 1993. “Credibility, Costs, and Institutions: Cooperation on EconomicSanctions.” World Politics45, no. 3 (April): 406-32.

McGillivray, Fiona and Alastair Smith. 2000. “Trust and Cooperation Through Agent-SpecificPunishments.” International Organization54, no. 4 (Autumn): 809-24.

Partell, Peter J. and Glenn Palmer. 1999. “Audience Costs and Interstate Crises: An EmpiricalAssessment of Fearon’s Model of Dispute Outcomes.” International Studies Quarterly43: 389-405.

Ramsay, Kristopher W. 2003. “Crisis Bargaining and Representative Democracy.” PaperPresented at the Meetings of the Midwest Political Science Association, Chicago (April).

Ramsay, Kristopher W. 2004. “Politics at the Water’s Edge: Crisis Bargaining and ElectoralCompetition.” J ournal of Conflict Resolution48, no. 4 (August): 459-86.

Schultz, Kenneth A. 1998. “Domestic Opposition and Signaling in International Crises.”American Political Science Review92, no. 4 (December): 829-44.



36

Schultz, Kenneth A. 1999. “Do Democratic Institutions Constrain or Inform? Contrasting TwoInstitutional Perspectives on Democracy and War.” International Organization53, no. 2(Spring): 233-66.

Schultz, Kenneth A. 2001a.Democracy and Coercive Diplomacy. New York: CambridgeUniversity Press.

Schultz, Kenneth A. 2001b. “Looking for Audience Costs.” J ournal of Conflict Resolution45,no. 1 (February): 32-60.

Slantchev, Branislav L. 2004. “The Watchful Eye: Information Transmission and PoliticalFailure.” Working Paper, Univ. of California San Diego, September 9.

Smith, Alastair. 1996. “To Intervene or Note to Intervene: A Biased Decision.” J ournal of Conflict Resolution40, no. 1 (March): 16-40.

Smith, Alastair. 1998. “International Crises and Domestic Politics.” American Political ScienceReview92, no. 3. (September): 623-638.

Tomz, Michael. 2004. “Interests, Information, and the Domestic Politics of InternationalAgreements.” Working paper, Stanford University.

Walt, Stephen M. 1999. “Rigor or Rigor Mortis? Rational Choice and Security Studies,”International Security23, no. 4 (Spring): 5-48

Wittkopf, Eugene R. 1990. Faces of Internationalism: Public Opinion and American ForeignPolicy. Durham: Duke Univ. Press.



37

Appendix: Text of the Experiment

The following questions are about US relations with other countries around the world. You willread about a situation our country has faced many times in the past and will probably face again.Different leaders have handled the situation in different ways. We will describe one approach

US leaders have taken, and ask whether you approve or disapprove.

A country sent its military to take over a neighboring country. The attacking country was led bya [dictator, who invaded OR democratically elected government, which invaded] [to get morepower and resources OR because of a longstanding historical feud.] The attacking country had a[strong military, so it would OR weak military, so it would not] have taken a major effort for theUnited States to help push them out. A victory by the attacking country would [hurt OR notaffect] the safety and economy of the United States.

P1: The US president said the United States would stay out of the conflict. The attackingcountry continued to invade. In the end, the US president did not send troops, and the attacking

country took over its neighbor.

P2: The US president said that if the attack continued, the US military would push out theinvaders. The attacking country continued to invade. In the end, the US president did not sendtroops, and the attacking country took over its neighbor.

P3: The US president said that if the attack continued, the US military would push out theinvaders. He sent troops to the region and prepared them for war. The attacking countrycontinued to invade. In the end, the US president did not send our troops into battle, and theattacking country took over its neighbor.

P4: The US president said that if the attack continued, the US military would push out theinvaders. He sent troops to the region and prepared them for war. The attacking countrycontinued to invade. The president then ordered US troops to destroy one of the invader’smilitary bases. US troops destroyed the base, and no Americans died in the operation. Theinvasion still continued. In the end, the US president did not order more military action, and theattacking country took over its neighbor.

P5: The US president said that if the attack continued, the US military would push out theinvaders. He sent troops to the region and prepared them for war. The attacking countrycontinued to invade. The president then ordered US troops to destroy one of the invader’smilitary bases. US troops destroyed the base, and 20 Americans died in the operation. Theinvasion still continued. In the end, the US president did not order more military action, and theattacking country took over its neighbor.

To summarize,

• The attacking country had a [strong OR weak] military, was led by a [dictator ORdemocratically elected government], and invaded [to get more power and resources ORbecause of a longstanding historical feud].



38

• A successful invasion would [hurt OR not affect] the safety and economy of the UnitedStates.

[If P1, include these additional bullet points]

• The US president said the United States would stay out of the conflict.

• The attacking country continued to invade.• The US president did not send troops.


• The US president said that if the attack continued, the US military would push out theinvaders.

• The attacking country continued to invade.

• The US president did not send troops.


• The US president said that if the attack continued, the US military would push out the

invaders. He sent troops to the region and prepared them for war.• The attacking country continued to invade.

• The US president did not send our troops into battle.


• The US president said that if the attack continued, the US military would push out theinvaders. He sent troops to the region and prepared them for war.


• The president then ordered US troops to destroy a military base.

• US troops destroyed the base, and no Americans died.

• The invasion still continued.

• The US president did not order more military action.


• The US president said that if the attack continued, the US military would push out theinvaders. He sent troops to the region and prepared them for war.


• The president then ordered US troops to destroy a military base.

• US troops destroyed the base, and 20 Americans died.

• The invasion still continued.

• The US president did not order more military action.

[Final bullet point for all conditions]

• The attacking country took over its neighbor.

Do you approve, disapprove, or neither approve nor disapprove of the way the US presidenthandled the situation? [If approve]: Do you approve very strongly, or only somewhat?



[If dispprove]: Do you disapprove very strongly, or only somewhat? [If neither]: Do you leantoward approving of the way the US president handled the situation, lean toward disapproving,or don't you lean either way?

Documents

2005 Domestic Audience Costs in International Relations_An Experimental Approach