Download Domestic Politics of International Compromise

Domestic Politics of International Compromise
Ryan Brutger⇤
October 14, 2016
Abstract: In an era of increasingly public diplomacy, conventional wisdom assumes
that leaders who compromise lose the respect of their constituents and damage their
reputations, which undermines the prospects for peace and cooperation. This is the first
paper to challenge this assumption and test how leaders can negotiate compromises and
avoid paying domestic costs. Drawing on theories of psychology, leadership, and partisanship, I argue that leaders reduce or eliminate domestic public constraints by exercising
“proposal power” and initiating compromises. Employing survey experiments to test
how public approval responds to leaders’ strategies across security and economic issues,
I find approval for compromise is conditioned by the partisanship of the audience and
the president, with Republican leaders having greater leeway to negotiate compromises
and audiences of Democrats being more supportive of compromise. These contributions
suggest that leaders who exercise proposal power have significant flexibility to negotiate
compromise settlements in international bargaining.
Acknowledgements: I would like to thank Omar Bashir, Thomas Christensen, Christina Davis,
In Song Kim, Yonatan Lupu, Helen Milner, Julia Morse, Tyler Pratt, Thomas Scherer, Kristopher Ramsay, Elizabeth Saunders, Jacob Shapiro, Yuki Shiraito, Jessica Weeks and participants
of Princeton’s graduate research seminar for valuable feedback on earlier drafts of this project. I
also thank Princeton’s Center for International Security Studies, the Fund for Experimental Social
Sciences, the Neihaus Center for Globalization and Governance, and the Mamdouha S. Bobst Center
for Peace and Justice for their financial support.
⇤ Assistant
Professor, University of Pennsylvania, Department of Political Science, Email: [email protected].
Web: http://web.sas.upenn.edu/brutger/.
Introduction
Hans Morgenthau (1948, 433) famously wrote that “No man who has taken such a stand before
the attentive eyes and ears of the world can in full public view agree to a compromise without
looking like a fool and a knave.” Yet even as he noted the risks of public compromise in the age
of “new” diplomacy, Morgenthau recognized compromise as essential to the maintenance of peace
and order in international relations.1 Almost seventy years after the publication of Politics Among
Nations, scholars continue to fear the e↵ects of public compromises. Indeed scholars have written
extensively on the costs of inconsistent rhetoric and e↵ects of changing policies on public opinion
(Fearon, 1994; Hummel, 2010; Kertzer and Brutger, 2016; Levy et al., 2015; Poole, 2007; Tomz
and Houweling, 2012). Although research shows that leaders are generally punished for acting
inconsistently, existing work has failed to examine how public opinion reacts to compromises and
its consequences for international negotiations. Since theories of political positioning (Hummel,
2010), diplomacy (Morgenthau, 1948; Yarhi-Milo, 2013), and international bargaining (Fearon, 1994;
Guisinger and Smith, 2002; Tarar and Leventouglu, 2013) rely on assumptions about how domestic
public opinion responds to leaders’ strategies, it is time to directly examine how compromise a↵ects
public opinion.
The idea that compromise and inconsistency generate domestic political costs is most prominently integrated into IR theory by Fearon’s (1994) seminal work on domestic audience costs. Since
Fearon wrote that backing down from threats generates domestic costs due to concerns about “credibility, face, and honor” (Fearon, 1994, 581), domestic audience costs have become common place
in crisis bargaining models (Haynes, 2012; Kurizaki, 2007; Tarar and Leventouglu, 2009). However,
theories invoking domestic concerns about reputation and public opinion extend far beyond audience costs (Chapman and Reiter, 2004; Cohen, 1995; Herrmann, Tetlock, and Visser, 1999; Jentleson
and Britton, 1998). Yet even as the influence of domestic public opinion on foreign policy has received significant attention, no study has systematically examined whether public opinion responds
to compromises in a manner that supports scholars’ extreme aversion to public compromises.
Rather than assuming that the public opposes compromise, this paper systematically examines
public attitudes toward compromise, focusing on three contributions. First, I argue that compromise
– where a state or leader agrees to concessions that are less than they previously threatened not to
1
See the “Four Prerequisites of Compromise” outlined in Morgenthau’s rules of diplomacy (Morgenthau, 1948, 441).
1
accept, but are short of backing down entirely – is viewed as a distinct political strategy that generates greater domestic support than other forms of political inconsistency. I find that compromise
does not systematically generate disapproval among domestic audiences. Using causal mediation
analysis, I further probe the foundations of support for compromise, finding that audiences believe
leaders preserve their reputations when negotiating compromises. Second, building from psychological and management theories of leadership, I introduce an original theory of “proposal power,”
whereby leaders who are perceived as initiating compromises, as opposed to accepting them, garner
greater public support. Proposal power allows leaders to frame the perception of negotiated settlements, such that leaders can avoid domestic punishment for baking down from threats and promises.
Lastly, I show that perceptions of compromise and proposal power are significantly influenced by
partisan politics. Rather than politics stopping at the water’s edge, partisanship plays a critical
role in shaping how audiences react to foreign policy compromises and conditions the diplomatic
incentives leaders face at home. These contributions show that leaders have significant flexibility to
negotiate compromise settlements on the international stage and, contrary to the prevailing wisdom,
public compromises do not jeopardize the respect of the public or the leader’s reputation.
Moving away from the assumption that audiences uniformly punish compromise is not just a
theoretical endeavor, but is important for capturing many real-world bargaining environments. On
the domestic stage, the importance of compromise was highlighted as the “fiscal cli↵” approached in
2012. Although many politicians had made ultimatums regarding spending cuts and tax increases,
a Gallup poll found that “Sixty-two percent of Americans would like to see federal government
leaders compromise on an agreement... more than twice the 25% who want leaders to stick to their
principles...” (Jones, 2012). On the international stage, President George W. Bush in the Six-Party
Talks negotiated what was described as a “wise compromise” with the North Koreans, even after
insisting “on complete, irreversible, and short-term disarmament as the only acceptable approach,”
which was not achieved in the negotiations (Bernard, 2007). More recently, this type of bargaining
scenario played out in the Syrian crisis, where President Obama made a public threat to the Syrian
leadership in 2012, but later chose to back down from that threat and instead accept a compromise.
The president’s threat was broadly publicized by major media outlets as a threat to use military
force, with the New York Times publishing the headline “Obama Threatens Force Against Syria”
(Landler, 2012) and the Wall Street Journal writing “President Threatens Military Response Against
Any Use of the Banned Arms” (Entous and Lee, 2012). Yet even after the administration reported
2
“definitive proof” that the Syrian regime had used chemical weapons (Entous, 2013), President
Obama was quick to embrace the idea of a compromise that avoided following through on his
threat of force (Borger and Roberts, 2013). Instead of following through on a military option,
the president supported Russia’s proposal to place Syria’s chemical weapons under international
control.2 Compromises such as President Obama’s in the Syrian crisis, President Bush’s in the
Six-Party Talks, and those on the domestic stage are not uncommon and thus deserve our attention
when evaluating the constraints under which leaders negotiate
Diplomacy, Compromise, and Public Opinion
Whether negotiating a trade agreement or contemplating military action, leaders often have an
incentive to push for a bigger piece of the pie. This means leaders will claim to have high demands
or initiate threats that misrepresent their true reservation price in negotiations in an e↵ort to gain
a better outcome. While asking for more may be a sound bargaining strategy in some situations, it
can also build expectations among domestic constituents that may make it harder for leaders to back
down from their demands or threats and instead accept compromises. This is why scholars of both
international relations (Fearon, 1994; Morgenthau, 1948) and American politics (Hummel, 2010)
argue that leaders pay a domestic cost for backing away from their stated positions. This should be
especially true in the age of modern diplomacy, where the media provides the public greater access
to information about their leader’s diplomatic positions (Potter and Baum, 2010, 2014; Slantchev,
2006), thus subjecting diplomacy to the “vice of publicity” (Morgenthau, 1948, 431).
Scholarship on international negotiations and bargaining perpetuates the longstanding idea that
leaders jeopardize their domestic standing and face public ridicule if they back down from their
bargaining positions under the watchful eye of domestic publics (Morgenthau, 1948). The most
prominent theory of foreign policy accountability is audience costs theory, but recent research has
moved beyond traditional audience costs theory and shown that domestic publics hold their leaders
accountable not only for threats, but also public promises on the international stage (Levy et al.,
2015). From these theories, the IR literature generally assumes that once leaders publicly take a
2
The compromise agreement regarding chemical weapons was inspired by a comment from Secretary of State, John Kerry, which was “not intended to be a diplomatic opening.” However, before
Kerry’s return flight to Washington had even landed, Russia had officially proposed a settlement
where Syria would relinquish control of its chemical weapons (Gearan, DeYoung, and Englund,
2013).
3
stand, they will pay a domestic cost in the form of lower public approval or lost votes, for any form
of inconsistency, such as agreeing to a compromise where they fail to follow through on their original
threat or promise.
The implications of domestic public opinion costs for international diplomacy and bargaining are
two-fold. First, as Morgenthau (1948) pointed out, international diplomacy requires compromise and
subjecting international negotiations to the whim of public opinion jeopardizes international peace
and security by undermining the potential for compromise. The importance of compromise for peace
and security is emphasized by Huth (1988) and Leng (1993) who find that reciprocating firm-butflexible strategies, where leaders pursue diplomatic compromises, are the most e↵ective bargaining
strategies for deterrence and avoiding war. Pursuing a compromising strategy can also provide the
necessary time to break a diplomatic stalemate (Huth, 1988, 13) and may be viewed as an act
of restraint, which could lead to a downward spiral of the conflict and a more peaceful outcome
(Kydd, 1997, 387) For such compromise strategies to succeed, leaders must have the capacity to give
ground in negotiations, and thus leaders need to believe they can make concessions and compromises
without fear of their domestic constituents retaliating against them. Second, domestic constraints
can serve a useful role in international bargaining by signaling information and acting as a credible
signal of leaders’ resolve (Chapman, Urpelainen, and Wolford, 2013; Fearon, 1994; Mo, 1995). Yet
even though they can enhance bargaining leverage at times, domestic constraints can also reduce
the potential win-set of agreements, which can undermine international cooperation and prospects
for peace.3
Reconsidering Theories of Compromise
To understand why and how audiences react to their leaders’ bargaining strategies, I first revisit
the most prominent explanations for why audiences punish leaders who act inconsistently on the
international stage and then, building upon insights from psychology, management studies, and
theories of partisanship, develop a new foundation from which to understand public opinion of
compromise.
Prominent explanations of why audiences punish leaders for inconsistency invoke concerns about
3
Stasavage (2004) and Fingleton and Raith (2001) each show that public negotiations can results
in incentives and constraints that reduce the likelihood of agreement.
4
the leader’s or the country’s reputation (Brutger and Kertzer, 2015; Fearon, 1994; Guisinger and
Smith, 2002; Levy et al., 2015) or the leader’s competence (Gelpi and Grieco, 2015; Smith, 1998). As
Tomz notes, audiences “believe that hollow threats and promises undermine the country’s reputation;
that empty commitments are dishonorable and embarrassing; or that inconsistency is evidence of
incompetence” (Tomz, 2007). While these mechanisms may condition approval for leaders who back
down from threats and do nothing, I argue that compromise is viewed in a distinct manner, such
that these beliefs are tempered, or completely absent, when leaders choose not to follow through
on a threat or promise and instead negotiate a settlement, especially when the leader proposes the
settlement.
In contrast to IR’s general aversion to compromise strategies, scholars of psychology and management studies find that concessions and compromise yield positive outcomes in terms of how
audiences feel about negotiations and their perception of the negotiator. While IR scholars argue
that signaling resolve is important in bargaining (McManus, 2014), management scholars have found
that not o↵ering concessions “can create the impression that a focal negotiator does not act ‘in good
faith,’ which in turn can result in deadlock...” (Hu↵meier and Freund, 2014). In their study of
bargaining perceptions and behavior, Benton, Kelley, and Liebling (1972) found that subjects who
observed a negotiator o↵er concessions were more likely to believe the negotiator was “fair” and the
outcomes were more satisfying than when the negotiator stayed firm. Furthermore, they found that
when the negotiator made a high demand and stuck to it, respondents believed the negotiator was
“unfair” and “bad,” although they were also more likely to say the negotiator was “strong.” These
studies support the theory that compromise can generate public support and is viewed di↵erently
than other forms of political inconsistency.
Although psychology and management studies find that leaders who compromise are viewed
as fair, it is important to ask whether we should expect that perceptions of fairness will result in
heightened support for leaders in international negotiations. A burgeoning literature shows that
individuals’ values and perceptions of fairness do play an important role in shaping public opinion
of foreign policy (Brewer et al., 2004; Gottfried and Trager, 2016; Kertzer et al., 2014; Rathbun
et al., 2016). Scholars have demonstrated that one of the most important set of values for predicting
attitudes toward foreign policy is self-transcendence values – those associated with fairness and caring
for others – which strongly predict cooperative internationalism (Rathbun et al., 2016). Examining
historical negotiations, Kertzer and Rathbun (2015) show that individuals’ concerns for fairness and
5
prosocial behavior played an important role in shaping bargaining behavior in the Anglo-French and
Franco-German negotiations after the First World War. These studies demonstrate that concerns
for fairness play a significant role in shaping foreign policy attitudes and suggest that those who
value fairness are likely to support leaders who negotiate international compromises.
Furthermore, the finding that publics respond favorably to compromise and concessions is consistent with theories of graduated reciprocation in tension-reduction models (Osgood, 1962), which
have also shown that backing away from demands makes agreement more likely in negotiations (Benton, Kelley, and Liebling, 1972). Concessions and compromise thus play a vital role in the resolution
of disputes and leaders who embrace these strategies are more likely to be viewed as fair and their
audiences are more likely to be satisfied with the negotiated outcome. Rather than jeopardizing the
leader’s reputation, backing down from demands and reaching a compromise can generate positive
approval from audiences and increased satisfaction with the bargaining outcome. This suggests that
compromise is viewed in a substantively di↵erent manner than other forms of political inconsistency,
and leaders who compromise should garner greater support than those who engage in other forms
of inconsistency.
Hypothesis 1: Ceteris paribus, leaders who back down from threats and reach compromises will have higher approval than those who back down and choose not to engage.
Proposal Power
There is a potential trade-o↵ that leaders face when backing down from threats or ultimatums and
reaching compromises. Leaders who make concessions are less likely to be viewed as strong, even
though they are more likely to be viewed as fair and good leaders (Benton, Kelley, and Liebling,
1972). Given that international relations is argued to be dictated by power, or perceptions of power,
leaders ought to take steps to preserve their reputations for strength. I argue that leaders can
exercise proposal power to enhance audience perceptions of their leadership strength and bolster
support for negotiated compromises.
Proposal power – the framing of a negotiated settlement such that the leader is viewed as
initiating the agreement, as opposed to accepting it – shapes audiences’ perceptions of leaders’
handling of negotiations in two ways. First, it enhances the public’s perception of the negotiator’s
leadership qualities. According to leadership theories, leadership requires initiative, coordination,
6
and direction (Bass, 1990). When leaders take initiative, audiences and followers tend to give them
higher ratings (Van Vugt, 2006). Experimental studies of leadership have found that those who
take initiative are more likely to be nominated as leaders in the future (Kremer and Mack, 2013),
suggesting that voters are more likely to support and re-elect politicians who take the initiative and
propose agreements. These theories of leadership suggest that when audiences view politicians taking
the initiative, as is the case when they exercise proposal power, the leader fulfills the audience’s
expectation for what a leader ought to do, which builds the perception of the politician as an
e↵ective and strong leader. Consistent with theories of leadership, proposal power allows politicians
to preserve their reputations and maintain support among domestic audiences.
Proposal power also influences public opinion by providing a reference cue to the domestic
audience about the negotiation. Given that the public is often poorly informed when it comes to
foreign policy (Holsti, 1996), audiences are likely to turn to their political leaders and respected elites
for cues that shape their foreign policy opinions (Berinsky, 2007; Saunders, 2015). In the context
of international negotiations, audiences look for in-group elites from whom to take cues – one of
the most salient elite cues may come from being in the national in-group. Since in-group favoritism
and out-group-anxiety play an important role in shaping foreign policy attitudes (Mansfield and
Mutz, 2009; Mutz and Kim, 2016), the domestic leader proposing an agreement provides a strong
cue that the leader of the national in-group endorses the agreement. Based on the leader’s proposal
power, the public is more likely to infer the agreement is in their interest and allay fears that
the agreement is being pushed by the foreign out-group. When evaluating negotiated settlements,
audiences will focus on whether a compromise is proposed by the leader of their country, since
international negotiations inevitably involve an “us” versus “them” divide, and be more likely to
support compromises proposed by their country’s leader.
Hypothesis 2: Leaders who propose compromises will have greater approval than those
who accept foreign leaders’ compromise proposals.
Partisanship
While psychology, leadership, and management studies give us reason to believe that compromise
strategies can generate broad support from the public, in the political realm we should expect
partisanship to play an important role in shaping how audiences respond to compromise. Similar
7
to proposal power, partisanship is a salient reference point for domestic audiences that allows them
to seek out cues from like-minded elites (Berinsky, 2007; Cohen, 2003; Nicholson, 2012). Much like
proposal power can provide a salient cue to audiences, partisanship provides an even more specific
cue about the negotiation. If the leader who initiates the proposal is a member of a constituent’s
party, then the compromise is proposed by a leader who is both a member of their party and their
country, making them the most likely audience to support the compromise settlement. Given these
overlapping reference groups for cues, audiences are more likely to support compromises when their
country’s leader proposes the settlement, and even more so when the proposer is a member of their
own party.
Hypothesis 3: Audiences will have greater approval for leaders of their own political
party who reach compromises, as opposed to those from opposing parties.
Partisanship identification also acts as a salient sorting mechanism in domestic politics. Recent
scholarship suggests that partisan identification reflects deeper individual traits that are likely to
a↵ect how people view the world and react to foreign policy (Jost et al., 2003, 2007; Jost, Nosek,
and Gosling, 2008). In the United States, liberals are more likely to identify as Democrats and conservatives are more likely to be Republicans. A growing literature on the psychological foundations
of political ideology finds that across countries conservatives are more resistant to change, oppose
ambiguity, and prefer definitive outcomes that provide closure (Jost et al., 2003, 2007; Jost, Federico,
and Napier, 2009; Zavala, Cislak, and Wesolowska, 2010). This suggests that conservatives, who are
more likely to resist change and oppose ambiguity, should be the most likely to disapprove of leaders
who change their minds or shift strategies. Additionally, conservatives are likely to view the world
as more threatening, and are thus more likely to prefer aggressive responses, such as military action (Jost, Federico, and Napier, 2009; Zavala, Cislak, and Wesolowska, 2010). Each of these traits
should lead Republicans to be more skeptical of compromises in international negotiations than
Democrats. In contrast, liberals are more likely to be accepting of uncertainty (Jost, Federico, and
Napier, 2009), and to be accommodationist and support cooperative internationalism (Holsti and
Rosenau, 1993, 1996). Sorting of liberals and conservatives based on these traits occurs in the U.S.,
where Democrats are found to be more dovish than Republicans (Herrmann, Tetlock, and Visser,
1999; Petrocik, 1980; Zavala, Cislak, and Wesolowska, 2010), which suggests that Democrats should
have more favorable views of compromise than Republicans.
8
Hypothesis 4: Democrats will have higher approval for compromise than Republicans.
Experimental Methods
To analyze how audiences respond to leaders’ strategies in international negotiations, I analyze a
series of four experiments. The first three build on previous experiments that use the classic “repel
an invader” scenario, while the fourth experiment employs an economic scenario, testing whether the
theories put forth have support across issue areas. Experimental studies are particularly well suited
to isolate changes in public approval in negotiation scenarios, given that they allow the researcher to
hold the context constant across a range of strategies and outcomes, ensuring internal validity of the
study. Recent experiments have found domestic public opinion costs to be present across a range
of contexts and countries, providing a robust literature on which I build (Davies and Johns, 2013;
Levendusky and Horowitz, 2012; Tomz, 2007; Trager and Vavreck, 2011).4 However, no experiment
has examined how the public reacts when a leader escalates a negotiation, fails to follow through on
a threat or promise, and then negotiates a settlement or compromise.
Using a similar framework as previous studies, I fielded a series of online survey experiments.
Experiment 1 was fielded in the spring of 2014 on a sample of 916 respondents. Experiment 2 was
fielded in spring of 2013 on a sample of 1761 respondents. Both studies recruited respondents using
Amazon’s Mechanical Turk and a discussion of the demographics of the samples and the limitations
and advantages of them is provided in the appendix, section 1. While Berinksy et al. (2012) and
others have shown the reliability of mTurk results in many areas, Experiment 3 fielded a truncated
version of the experiment on a nationally representative sample of 613 respondents administered by
Survey Sampling International (SSI) in the spring of 2016.5 The three experiments find consistent
results across four years of study and di↵erent sample populations, showing the results are not a
function of time or sample method.
The fourth experiment was fielded by SSI on a nationally representative sample of 543 Americans
in the spring of 2016. Experiment 4 was designed to test the theories of compromise and proposal
4
For further discussion on the apprpriateness of experiments for this project see appendix, section
1.
5
Additional details regarding the national sample and a comparison of core results across studies
are provided in the appendix, section 2. For examples of publications in leading political science
journals using SSI studies, see Berinsky, Margolis, and Sances (2014); Kertzer and Brutger (2016);
Malhotra, Margalit, and Mo (2013).
9
power in a new issue area – international economics. Since the public’s perception of the high politics
of national security may di↵er from how they view the low politics of international economics, it is
important to test whether the theories put forth explain perceptions of negotiation strategies and
foreign policy across issue areas. The results of the fourth experiment provide further support for
the theory of proposal power and the partisan nature of public opinion toward compromise. I will
return to Experiment 4 after a discussion of the three security focused experiments.
The first three experiments began with a script that read:
“The following questions are about U.S. relations with other countries around the world.
You will read about a situation our country has faced many times in the past and will
probably face again. Di↵erent leaders have handled the situation in di↵erent ways. We
will describe one approach U.S. leaders have taken, and ask whether you approve or
disapprove.”
Respondents next read about a crisis where a country sent its military to take over a territorial
region in a neighboring country. To avoid country-specific confounders and to maintain comparability
with previous studies, the scenarios deliberately avoided using any country names for the foreign
parties. After reading the background on the crisis, respondents were randomly assigned one of
the president’s strategies. While each of the first three experiments follow a similar approach,
Experiment 3 employs somewhat di↵erent language to address potential concerns with the first
experiments. I begin by introducing the text from Experiment 1 and then progress through a
discussion of the modifications made to the treatments in Experiments 2 and 3. In the baseline
condition, labeled Stay Out, the president does not engage in the crisis.
• Stay Out: “The U.S. president, who was a Democrat (or Republican), said the United States
would stay out of the conflict. The attacking country continued to invade and the conflict
ended with the attacking country taking control of 20 percent of the contested territory.”
In each of the remaining treatment conditions the U.S. escalated with a threat and then the
respondent was randomly assigned to the Compromise, Engage, or Not Engage condition:
Threat: “The U.S. president, who was a Democrat (or Republican), said that if the
attacking country continued to invade, the United States military would immediately
engage and attempt to push out the attacking country. The president sent troops to the
region.”
• Compromise Condition: “The attacking country continued to invade, but the president did
not immediately engage. The U.S. president (or leader of the attacking country) proposed
10
a settlement, which was agreed to by all parties, and the conflict ended with the attacking
country taking control of 20 percent of the contested territory.”6
• Engage Condition: “The attacking country continued to invade and the U.S. president ordered
the military to engage. The U.S. did not lose any troops in the conflict and the conflict ended
with the attacking country taking control of 20 percent of the contested territory.”
• Not Engage Condition: “The attacking country continued to invade. The U.S. president
ordered the military not to engage. The attacking country continued to invade and the conflict
ended with the attacking country taking control of 20 percent of the contested territory.”
For those respondents in the Compromise condition, there is an additional nested treatment
that specifies who proposed the negotiated settlement. After reading the president’s choice of action,
respondents were provided a summary of the events. Respondents were then asked whether they
“approve, disapprove, or neither approve nor disapprove of the way the U.S. president handled the
situation” and then how strongly they felt that way. These responses generated the dependent
variable, which is an approval score on a seven point scale ranging from -3 being strongly disapprove
to +3 being strongly approve.
A challenge of evaluating public approval for leaders’ strategies in real-world negotiations is
that audiences are likely to have preferences over both the strategy used by the leader and the
policy outcomes of the crisis (Chaudoin, 2014; Snyder and Borghard, 2011). While disaggregating
the audience’s preference for outcomes and strategies is challenging in observational studies, using
an experimental scenario allows the magnitude of the domestic cost to be measured while holding
policy outcomes constant. In order to isolate the cost across all strategies, as noted in the treatments
above, and for comparability to recent studies, I control for the outcomes across all conditions in
Experiment 1.7 This allows me to first test how the president’s strategy alone a↵ects the magnitude
of audience costs. I then analyze nested treatments in Experiment 2 to test whether approval for
compromise is conditioned by proposal power and whether additional concessions that lead to worse
policy outcomes influence approval.8
Experiment 3 is a modified version of the first experiments, which repeats the tests of proposal
power and partisanship on a nationally representative sample fielded by SSI. Experiment 3 also
6
In earlier versions, the Compromise treatment wording did not include “which was agreed to by
all parties,” and the results remained consistent. I also tested audience reactions when the conflict
ended with the attacking country taking control of 50 percent of the contested territory. The results
were not significantly di↵erent when the compromise gave the attacking country 50 percent of the
territory – a finding that is discussed in greater detail in the results section.
7
Casualties are held constant at zero. A further discussion of the importance of holding casualties
and territory constant at these levels is provided in the appendix, section 3.
8
The full text of the Experiment 2 treatments appears in the appendix, section 4.
11
modifies the language of the compromise treatment to address potential concerns regarding the
lexical equivalence of the compromise and not engage treatments in the first two experiments. In
the first two experiments the compromise treatment included language that the president “did not
immediately engage”. This language was chosen to emphasize the inconsistency of the president
failing to follow through on his threat to “immediately engage”, however it di↵ers in that the not
engage treatment says “the president ordered the military not to engage”. While the treatments
across the experiments all emphasize the same general strategic choices, Experiment 3 shows that
the results are consistent when the language is modified to “the president ordered the military not
to engage” in the compromise condition, which is identical to the corresponding language in the not
engage treatment of experiments 1 and 2. The exact wording for experiment 3 is as follows:
• Compromise Conditions: The attacking country continued to invade, but the U.S. president
ordered the military not to engage. The U.S. president (or leader of the attacking country)
proposed a settlement, which was agreed to by the parties, and the conflict ended with the
attacking country taking control of 20 percent of the contested territory.
Results: Security Experiments
The average approval scores, conditional on the party of the president, for Experiments 1 and 2
are reported in Figure 1.9 These results hold the outcome of the crisis constant, with the foreign
country taking over 20 percent of the territory in each scenario, isolating the e↵ect of the president’s
strategy. The first notable result is that both experiments replicate previous studies’ findings that
audience costs exist for leaders in the Not Engage condition, where the leader makes a threat, does
not follow through on it, and has no further involvement in the crisis. The Not Engage condition
consistently receives the lowest approval score regardless of whether the president is a Republican or
Democrat. These findings suggests that the experiments capture the baseline audience cost e↵ect as
in previous studies (Davies and Johns, 2013; Levendusky and Horowitz, 2012; Tomz, 2007; Trager
and Vavreck, 2011), which is a necessary starting point given that the following results focus on
mitigating public disapproval through compromise.
9
The average approval scores not conditioned on party of the president are shown in the appendix,
section 5.
12
Average Approval Score for Each Strategy
With a Republican as President
Republican President
Experiment 1
● Experiment 2
1
●
0
●
−1
●
●
−2
Average Approval Score
2
Figure 1:
Stay Out
Engage
Not Engage
●
●
0
1
2
Democrat President
Experiment 1
● Experiment 2
●
−1
●
−2
Average Approval Score
Compromise
Stay Out
Engage
Not Engage
Compromise
Figure 1 shows the average approval score and 90 percent confidence intervals for experiments 1 and 2 for each randomly assigned crisis strategy and party of the president.
Outcomes are held constant with no casualties and 20% territory gained by the attacking country. The approval score is measured on a scale from -3=Strongly Disapprove to
3=Strongly Approve.
The E↵ect of Compromise
Figure 1 also shows that regardless of the president’s partisanship, and even when holding outcomes
constant, audiences prefer leaders who back down from their threat and negotiate a compromise, as
opposed to those who simply choose not to engage after having made a threat, providing support for
13
hypothesis 1. The impact of shifting from the not engage condition to the compromise condition is
even more pronounced when analyzing the percent of respondents who approved of the President’s
handling of the situation. In Experiments 1 and 2, the percent of respondents approving of the
president is 20.6 (p<0.01) and 26.2 (p=0.02) percentage points higher when the president compromises than when she chooses not to engage.10 However, the general results obscure the e↵ect of
partisan audiences since they do not account for the interaction of the audience and the President’s
partisanship. To fully examine partisanship, Table 1 presents the results, conditional on both the
party of the audience and the president.11 For each combination of partisan audience and party
affiliation of the president, average public approval is higher in the Compromise condition than the
Not Engage condition.12 The di↵erence in means between these conditions is always positive and
is statistically significant for audiences of Democrats, while the results are positive but not significant for Republicans. Without taking into consideration the president’s proposal power, which will
be discussed in subsequent sections, Table 1 suggests that audiences, particularly Democrats, prefer
leaders who reach compromise settlements as opposed to not engaging after a threat, which supports
both hypotheses 1 and 4.13
Table 1: Experiment 1 - Di↵erences in Approval Between Not Engage and Compromise
Party of Audience
Republican
Democrat
President is Democrat
0.39
(0.38)
1.57***
(3.94)
President is Republican
1.07
(1.37)
1.34***
(3.15)
Statistical significance: ⇤ p<0.1; ⇤⇤ p<0.05; ⇤⇤⇤ p<0.01
Table 1 shows the di↵erence in average approval scores (Compromise N ot Engage) and t-statistics
among Democrats and Republicans based on the party of the president while holding territorial
and casualty outcomes constant at no casualties and 20% territory gained by the attacking country. The approval score is measured on a scale from -3=Strongly Disapprove to 3=Strongly Approve.
10
Additional results using a dichotomous measure of whether respondents approved or not of the
president’s handling of the situation are shown in the appendix, section 6.
11
The full results for Experiment 1 and Experiment 2 are reported in the appendix, section 7.
12
Results for Independents are reported in the appendix, section 8.
13
The experiment also finds support for the portrayal of Republicans as more hawkish, as is shown
in the appendix, section 7. On average, republicans are more supportive of engagement, relative to
staying out, than Democrats.
14
To understand why audiences tend to favor compromise, I examine the most prominent explanation for why audiences choose to punish leaders’ inconsistency – reputational concerns (Brutger
and Kertzer, 2015; Fearon, 1994; Guisinger and Smith, 2002; Weeks, 2008; Tomz, 2007). To measure
concerns about reputation, each respondent was asked “On a scale of 1-5, how much damage do
you think there would be to the President’s reputation as a result of the President’s handling of
the situation?” When leaders chose to reach a compromise, after backing down from a threat, as
opposed to not engaging, they received a significant improvement in their perceived reputation (0.65,
p < 0.01). These results are confirmed using nonparametric mediation analysis (Imai et al., 2010)
to test how perceptions of reputation a↵ect approval of the president’s handling of the situation.
When examining the increase in approval from the Not Engage to the Compromise condition, the
mediating e↵ect of reputation is responsible for 61 percent of the increase in approval (p < 0.01).14
These findings demonstrate that leaders who are concerned about their reputations among domestic
audiences, can preserve or enhance their reputations by reaching diplomatic compromises.
Sensitivity to Outcomes
While audiences favor leaders who negotiate compromise settlements over leaders who back down and
don’t engage, the next question is whether these results are robust to variations in crisis outcomes.
Snyder and Borghard argue that the substantive policy outcomes of a crisis play a significant role in
determining public approval, which could undermine the mitigating e↵ect of compromise on audience
costs if leaders are punished for making additional concessions (Snyder and Borghard, 2011). To
evaluate this concern, Experiment 2 tested the sensitivity of audiences’ approval to randomly varied
crisis outcomes. Experiment 2 follows the same structure as the first experiment, but the amount
of territory gained by the attacking country was varied.15 I also randomly assigned the number of
casualties within the Engage condition, with the expectation that additional casualties will decrease
approval for engagement (Gartner, 2008; Gartner and Segura, 1998; Gelpi, Feaver, and Reifler, 2005).
Interestingly, I find that audiences do not punish leaders for worse outcomes in the Compromise
condition. The outcomes of the negotiated settlement led to the foreign country gaining a baseline
of 20 percent of the contested territory or an additional 30 percent, which emphasizes the leader’s
14
The mediation results are shown in the appendix, section 9.
A discussion of Experiment 2 and the full treatment conditions are provided in the appendix,
section 4.
15
15
concessions when compromising. Among Democrats the shift in average approval when the attacking
foreign country gained an additional 30 percent of the territory was essentially zero at 0.11 (p=0.72)
and among Republicans it was -0.45 (p=0.34).
In contrast to the Compromise condition, audiences of Democrats are sensitive to outcomes in
the Engage condition. Table 2 shows the e↵ects of changes in territorial outcomes and casualties
on public approval in the Engage condition. The outcomes of the military engagement led to the
foreign country gaining either none or 20 percent of the contested territory and the U.S. experienced
either no casualties or 1,000 casualties in the engagement. In the event the foreign country did
not gain any territory, Democrats’ approval score dropped by 1.03 (p=0.01) when there were 1,000
casualties in comparison to no casualties. The e↵ect of the foreign country gaining 20 percent of the
territory (with no casualties) is similar, with a decline in approval score among Democrats of 1.13
(p<0.01). Republicans on the other hand did not show a significant change in approval in response
to crisis outcomes and instead consistently support military engagement. These results support the
characterization of Republicans as consistently hawkish, whereas Democrats’ approval of military
strategies is highly dependent on how favorable an outcome is achieved.
Table 2: Experiment 2 - Di↵erences in Approval Based on Outcomes
Party of Audience
Republican
Democrat
Casualties
-0.35
(-0.51)
-1.03***
(-2.63)
Territory
-0.35
(-0.51)
-1.13***
(-3.02)
Casualties and Territory
0.40
(0.77)
-1.84***
(-4.74)
Statistical significance: ⇤ p<0.1; ⇤⇤ p<0.05; ⇤⇤⇤ p<0.01
Table 2 shows the di↵erence in average approval scores for the Engage condition based on the crisis
outcomes, showing the e↵ects of 1,000 casualties, the hostile country gaining 30% more territory,
or both, along with the t-statistics among Democrats and Republicans. The approval score is
measured on a scale from -3=Strongly Disapprove to 3=Strongly Approve.
16
Proposal Power
A final component of compromises that is likely to impact audiences’ approval is the perception of
the leader’s role in reaching a settlement, specifically whether they exercise proposal power. To test
the e↵ect of proposal power, Experiment 2 randomly assigned whether a compromise was proposed
by the president or the foreign leader.
The power of being the proposer generally plays in the favor of the president, but partisanship
plays a significant role in conditioning the audience’s response. To measure the role of partisanship,
I analyze proposal power across di↵erent combinations of the president’s party and the audience’s
party. I use a “Party Match” variable that indicates whether the audience member and president
are from the same party. As is shown in Figure 2, Democrats consistently have favorable opinions
of presidents who are Democrats and agree to compromises, regardless of whether the president
proposes or accepts the compromise. The only situation where Democrats look unfavorably on
compromise is when the president is a Republican and the compromise is initiated by the foreign
leader. In contrast, and consistent with hypothesis 4, Republican audiences are more selective when
supporting compromise. Republicans strongly support a Republican president when she proposes
the compromise, but not when she accepts a compromise initiated by the foreign country. Consistent
with hypotheses 2 and 3, these findings support the idea that audiences take cues from their leaders
in international negotiations, resulting in greater support for the leader when she is a member of their
party and when the domestic leader initiates the compromise. The results also confirm a long-held
idea that “only Nixon could go to China,” meaning that Republican presidents, who are more likely
to be viewed as hawks, actually have an easier time making peace (Cowen and Sutter, 1998; Schultz,
2005; Trager and Vavreck, 2011). A Republican president who proposes a compromise receives strong
support from audiences of both parties, with support at 57.4 percent among Democrats and 59.1
percent among Republicans. In contrast, a Democrat who proposes a compromise receives strong
support from her own party (58.5 percent), but only 25.0 percent support from Republicans, which
highlights the partisan nature of compromise.16
The results from the nationally representative sample in Experiment 3 add further support to the
influence of partisanship and proposal power on public opinion toward international compromises.17
16
Additional proposal power and partisanship results based a dichotomous measure of whether
respondents approve are displayed in the appendix, section 6.
17
The full results for experiment 3 are displayed in the appendix, section 2.
17
Figure 2: Average Approval Score for Compromise
Based on Party-Match and Proposal Power
1.5
Audience of:
Democrats
Republicans
1.0
0.5
●
●
0.0
●
−1.0
−0.5
●
−1.5
Average Approval Score
●
Party Match
President Proposes
Party Match
Foreign Leader
Proposes
No Party Match
President Proposes
No Party Match
Foreign Leader
Proposes
Figure 3 shows the average approval score and 90 percent confidence intervals for audiences of Democrats and Republicans based on whether the respondent and president
are from the same party (party-match) interacted with whether the U.S. president or
the foreign leader proposed the the negotiated settlement that was then agreed to by all
countries, while holding crisis outcomes constant. The approval score is measured on a
scale from -3=Srongly Disapprove to 3=Strongly Approve.
Among both Democrats and Republicans, leaders always received the highest approval score from
members of their own party when they exercised proposal power and initiated the compromise.
Additionally, the results demonstrate that di↵erences in language across the experiments did not
alter the underlying treatment e↵ects. Consistent with the results in Experiment 2, Experiment 3
found that the average approval score among Democrats was always positive when a compromise
was reached by the Democratic leader or the compromise was proposed by the domestic leader.
Across comparable conditions in experiments 2 and 3, the di↵erence in average approval scores was
essentially zero (0.005, p=0.98), highlighting that the nationally representative and mTurk samples generated remarkably consistent results. Republicans have a somewhat lower average baseline
approval in Experiment 3, but the comparable treatment e↵ect of switching from a Democrat to Republican president proposing the compromise, remains positive and significant (0.90, p=0.04). The
robustness of the results across samples, experimental wording, and time demonstrate that proposal
18
power and partisanship play critical roles in shaping public support for international negotiations
and compromise.
Economic Negotiations
To examine whether partisanship and proposal power a↵ect perceptions of compromise across issue
areas, Experiment 4 tested attitudes toward compromise in an international investment dispute.
The experiment was fielded in the spring of 2016 by SSI on a nationally representative sample of 543
respondents. The experiment presented respondents with a hypothetical investment dispute where
a foreign firm sued the US using investor-state dispute settlement (ISDS) arbitration. Rather than
duplicating all of the conditions from the earlier studies, Experiment 4 limits its focus to proposal
power, randomizing whether the U.S. leader or the foreign leader proposed a compromise. The
experiment also di↵ers in that it specifies President Obama as the U.S. leader, which means the
party of the president is not varied. However, this has the added advantage of allowing us to see
how the audience’s partisanship interacts with a known president and the e↵ect of that interaction
on public approval.
The respondents read about an investment dispute where “TransCorp”, a hypothetical company
from a neighboring country, sued the U.S. and the Obama administration for $15 billion dollars.18
Similar to the security experiments, the situation involved the U.S. escalating the dispute with a
clear promise to fight the challenge, specifying that the Obama administration would “fight the
challenge until the arbitration panel made its decision.” After reading the president’s commitment,
either the Obama administration or TransCorp “proposed a settlement granting TransCorp twenty
percent of the value of the suit” and the settlement was accepted by the other party. After reading
the scenario, respondents were asked the same approval question as in the earlier experiments.
The results, which are displayed in Figure 3, show that when the U.S. leader proposes the
settlement their approval score goes up by 0.4 (p<0.01). The e↵ect of proposal power based on
the partisanship of the audience shows that Democrats approval score increases by 0.65 (p=0.04)
whereas Republicans do not have a significant change in approval when President Obama proposes,
with a shift in approval score of just 0.27 (p=0.51). To highlight the magnitude of these e↵ects,
18
The full text of the scenario is provided in the appendix, section 10. The scenario and monetary
value at stake were based on an actual dispute between a canadian firm and the U.S. (King and
Mauldin, 2016).
19
it is helpful to consider the change in the percent of Democrats and Republicans who support the
settlement based on proposal power. Only 32 percent of Democrats support the settlement when
the foreign leader proposes it, but support jumps to 63 percent when the U.S. leader proposes the
agreement. In contrast, 24 percent of Republicans support the agreement when the foreign leader
proposes, and 27 percent support when the U.S. leader proposes.19 These results add additional
support to the findings that Republicans do not support international compromises proposed by
Democrats and they have a lower baseline level of support for compromise as well.
Figure 3: Average Approval Score for Compromise
Based on Party-Match and Proposal Power
1.0
0.5
●
−1.0
−0.5
0.0
●
−1.5
Average Approval Score
1.5
Audience of:
Democrats
Republicans
●
US Proposes
Foreign Leader Proposes
Figure 3 shows the average approval score and 90 percent confidence intervals for audiences of Democrats and Republicans based on whether the U.S. leader or foreign leader
proposed the compromise. The approval score is measured on a scale from -3=Srongly
Disapprove to 3=Strongly Approve.
One criticism that could be raised against the scenarios of the experiments, is that each includes
an ultimatum threat or promise prior to the leader reaching a compromise. This sequence of negotiation events was chosen to be explicit that the leader was reaching a compromise, where they
clearly backed down from their earlier bargaining position. However, since political negotiations are
inherently strategic, it is worth asking whether the leader would have been better o↵ never making
a threat in the first place. To test this proposition, Experiment 4 included an additional condition
where President Obama never issued a threat or promise, but the same settlement was reached.
19
The di↵erence in di↵erence of proposal power among partisan groups for percent supporting is
statistically significant (p=0.02)
20
In this final scenario respondents read “The Obama administration proposed a settlement granting
TransCorp twenty percent of the value of the suit, and TransCorp accepted the settlement.” This
condition was identical to the earlier condition where Obama proposed the settlement, with the
exception that this final condition does not include the threat to fight the dispute to the end. The
expectation among scholars who assume compromise is punished would be that approval would be
lower for the leader who made a threat and then backed down from it to reach a settlement, as
opposed to a leader who reached the same settlement without the threat. However, the results show
that there is no di↵erence in approval score between the two conditions (0.04, p=0.39) among the
full sample. Among Democrats the approval score is actually slightly higher for the leader who backs
down and then reaches a compromise (0.24, p<0.01)20 and among republicans there is no di↵erence
(0.04, p=0.63).
The results of the investment experiment provide the first analysis of how public opinion toward
compromise and proposal power translate across issue areas. While the investment experiment
did not include every potential treatment condition to test each hypothesis, the results supported
hypotheses two, three, and four, showing that proposal power and partisanship have a significant
e↵ect on support for negotiated outcomes and compromise. The evidence presents a strong case
that opinions toward compromise, and the partisan foundations that influence them, apply to both
the high politics of security issues and the low politics of international economics.
Conclusion
The evidence put forth in this paper builds a case for a reformulation of how we think about public
opinion and commitments in international bargaining. Three critical points emerge for understanding
the role of domestic audiences in international bargaining. First, is that leaders can influence the
level of public approval when negotiating a compromise by exercising proposal power. When leaders
negotiate settlements and are perceived as the proposer they are able to protect their reputations,
which allows them to mitigate or eliminate public disapproval. This introduces a new component to
the bargaining dynamic that allows leaders to frame the bargaining outcome for their home audience
and provides greater flexibility to the leader to negotiate concessions.
20
This is consistent with studies that found observers had greater satisfaction with negotioted
outcomes when the negotiator made concessions (Benton, Kelley, and Liebling, 1972).
21
The second takeaway is that the composition of a leader’s audience is of vital importance when
evaluating whether publics constrain leaders behavior and can create costly signals on the international stage. Whereas previous studies found that audience costs were non-partisan (Levendusky
and Horowitz, 2012; Trager and Vavreck, 2011), I find that partisanship significantly a↵ects approval
for compromise when both the audience and the leader’s partisanship are taken into account.21 Republican presidents have greater leeway to negotiate compromise settlements and receive strong
support from audiences of Democrats and Republicans when they propose compromises. In contrast, presidents who are Democrats are more constrained and only receive support from their own
party when they propose settlements. The role of partisanship in shaping public approval of foreign
policy strategies adds further support to a growing literature which argues that increasingly politics
does not stop at the water’s edge (Ahn, 2011; Cavari, 2013; Trubowitz and Mellow, 2005).
Lastly, when we include the possibility of compromise, audiences rarely lock in leaders in international negotiations and domestic approval concerns are, at best, inconsistent signals of leaders’
resolve. Rather than assuming publics disapprove of compromise, as Morgenthau (1948) argues,
these results fundamentally change how we think domestic audiences respond to international diplomacy. In the Syrian case, audience costs theory would predict that the president’s public threat
would generate domestic costs and that the decision not to follow through on the threat would
result in the president being punished. While some elites criticized the president and argued that
the failure to follow through with military action hurt the president’s credibility (Sink, 2014), the
general public was supportive of the compromise.22 Public approval for the president’s handling
of foreign policy showed no decline after the president failed to follow through on his threat. According to CBS and NBC/Wall Street Journal Polls, the president’s foreign policy approval ratings
were 43 and 46 percent respectively in the months before Syria used chemical weapons and stayed
relatively consistent at 45 and 46 percent following Syria’s use of chemical weapons (Polling Report,
2014). When specifically asked about President Obama’s handling of foreign policy toward Syria,
56 percent of Republicans and 80 percent of Democrats supported the decision not to engage in
airstrikes and instead pursue a compromise strategy (Pew Research, 2013). This case illustrates the
21
Although compromise appears to be a partisan issue, Figure 1 shows that both Republicans
and Democrats impose audience costs for the traditional audience cost treatment e↵ect, which is
consistent with earlier findings.
22
While elite cues play an important role in shaping public opinion (Berinsky, 2007; Saunders,
2015), in this case the general public did not follow the cues of the more hawkish critics, such as
Panetta (Sink, 2014).
22
importance of evaluating how public opinion responds to negotiations that go beyond the traditional
crisis bargaining model, where the leader either attacks or backs down, and shows how the potential
for compromise can prevent leaders from tying their hands in international negotiations.
While President Obama issued a prominent public threat in the Syrian crisis, he was able to
accept a compromise settlement and avoid being punished by the general public. Given that the
president was generally perceived as accepting a foreign leader’s proposal, this example is a least
likely case for compromise to successfully mitigate audience costs, yet audiences across the political
spectrum supported the President’s decision not to follow through on his threat. However, partisanship still played a key role in shaping public opinion, evidenced by the 24 percentage point higher
approval among Democrats than Republicans for President Obama’s choice to pursue a compromise. This example highlights the role of compromise in allowing leaders to maintain public support
and the importance of reconsidering how publics react to international negotiations. Rather than
constraining leaders diplomatic options and jeopardizing international peace and security (Morgenthau, 1948), domestic public opinion often supports leaders who make concessions, which limits the
signaling value of public threats and suggests that leaders have significant flexibility to negotiate
international compromise settlements.
23
References
Ahn, Taehyung. 2011. “Politics at the Water’s Edge: The Presidency, Congress, and US Policy
toward North Korea.” Pacific Focus 26 (December): 336-359.
Bass, Bernard. 1990. Bass and Stogdill’s Handbook of Leadership: Theory, Research, and Managerial
Applications, 3rd Edition. New York: Free Press.
Benton, Alan, Harold Kelley, and Barry Liebling. 1972. “E↵ects of Extremity of O↵ers and Concession Rate on the Outcomes of Bargaining.” Journal of Personality and Social Psychology 24 (1):
73-83.
Berinksy, Adam, Gregory Huber, Gabriel Lenz, and Edited by R. Michael Alvarez. 2012. “Evaluating
Online Labor Markerts for Experimental Research: Amazon’s Mechanical Turk.” Political Analysis
20 (3): 351-368.
Berinsky, Adam J. 2007. “Assuming the Costs of War: Events, Elites, and American Public Support
for Military Conflict.” The Journal of Politics 69 (4): 975-997.
Berinsky, Adam J, Michele F Margolis, and Michael W Sances. 2014. “Separating the Shirkers from
the Workers? Making Sure Respondents Pay Attention on Self-Administered Surveys.” American
Journal of Political Science 58 (3): 739–753.
Bernard, Gwertzman. 2007. “Samore: N. Korean Deal a ‘Wise Compromise for Bush.” Council on
Foreign Relations. Interview February 13.
Borger, Julian, and Dan Roberts. 2013. “Syria crisis: Obama Welcomes Russia’s Chemical Weapons
Proposal.” The Guardian. September 9.
Brewer, Paul R, Kimberly Gross, Sean Aday, and Lars Willnat. 2004. “International trust and public
opinion about world a↵airs.” American Journal of Political Science 48 (1): 93–109.
Brutger, Ryan, and Joshua Kertzer. 2015. “Exploring the Microfoundations of Reputation Costs.” Working Paper. http://people.fas.harvard.edu/~jkertzer/Research_files/
Website_Microfoundations%20of%20Rep.pdf.
24
Cavari, Amnon. 2013. “Religious Beliefs, Elite Polarization, and Public Opinion on Foreign Policy:
The Partisan Gap in American Public Opinion Toward Israel.” International Journal of Public
Opinion Research 25 (Spring): 1-22.
Chapman, Terrence L, and Dan Reiter. 2004. “The United Nations Security Council and the rallyround the flag e↵ect.” Journal of Conflict Resolution 48 (6): 886–909.
Chapman, Terrence L, Johannes Urpelainen, and Scott Wolford. 2013. “International bargaining,
endogenous domestic constraints, and democratic accountability.” Journal of Theoretical Politics
25 (2): 260–283.
Chaudoin, Stephen. 2014. “Promises or Policies? An Experimental Analysis of International Agreements and Audience Reactions.” International Organization 68 (1): 235-256.
Cohen, Geo↵rey. 2003. “Party Over Policy: The Dominating Impact of Group Influence on Political
Beliefs.” Journal of Personality and Social Psychology 85 (5): 808-822.
Cohen, Je↵rey E. 1995. “Presidential rhetoric and the public agenda.” American Journal of Political
Science: 87–107.
Cowen, Tyler, and Daniel Sutter. 1998. “Why Only Nixon Could Go to China.” Public Choice 97 (4):
605-615.
Davies, Graeme, and Robert Johns. 2013. “Audience Costs among the British Public: The Impact of
Escalation, Crisis, Type, and Prime Ministerial Rhetoric.” International Studies Quarterly 57 (4):
725-737.
Entous, Adam. 2013. “U.S. Talks Touch on Syria, Ramps Up Attack Planning.” The Wall Street
Journal. August 25.
Entous, Adam, and Carol Lee. 2012. “Obama Warns Syria on Chemical Weapons, President Threatens Military Response Against Any Use of the Banned Arms.” The Wall Street Journal. August
20.
Fearon, James. 1994. “Domestic Political Audiences and the Escalation of International Disputes.”
American Political Science Review 884 (September): 577 - 592.
25
Fingleton, John, and Michael Raith. 2001. “Open covenants, privately arrived at.” CEPR Discussion
Paper. http://papers.ssrn.com/sol3/papers.cfm?abstract_id=264362.
Gartner, Scott Sigmund. 2008. “The Multiple E↵ects of Casualties on Public Support for War: An
Experimental Approach.” American Political Science Review 102 (2): 95–106.
Gartner, Scott Sigmund, and Gary M. Segura. 1998. “War, Casualties, and Public Opinion.” The
Journal of Conflict Resolution 42 (3): 278-300.
Gearan, Anne, Karen DeYoung, and Will Englund. 2013. “Syria Says it ‘Welcomes’ Russia Proposal
on Chemical Weapons.” The Washington Post. September 9.
Gelpi, Christopher, and Joseph M. Grieco. 2015. “Competency Costs in Foreign A↵airs: Presidential
Performance in International Conflicts and Domestic Legislative Success, 19532001.” American
Journal of Political Science 59 (2): 440–456.
Gelpi, Christopher, Peter D. Feaver, and Jason Reifler. 2005. “Success Matters: Casualty Sensitivity
and the War in Iraq.” International Security 30 (3): 7-46.
Gottfried, Matthew S, and Robert F Trager. 2016. “A preference for war: how fairness and rhetoric
influence leadership incentives in crises.” International Studies Quarterly.
Guisinger, Alexandra, and Alastair Smith. 2002. “Honest Threats: The Interaction of Reputation
and Political Institutions in International Crises.” Journal of Conflict Resolution 46 (2): 175-200.
Haynes, Kyle. 2012. “Lame Ducks and Coercive Diplomacy: Do Executive Term Limits Reduce the
E↵ectiveness of Democratic Threats?” Journal of Conflict Resolution 56 (5): 771-798.
Herrmann, Richard, Philip Tetlock, and Penny Visser. 1999. “Mass Public Decisions to Go to War:
A Cognitive-Interactionist Framework.” American Political Science Review 93 (3): 553-573.
Holsti, Ole R. 1996. Public Opinion and American Foreign Policy. Cambridge University Press.
Holsti, Ole R., and James N. Rosenau. 1993. “The Structure of Foreign Policy Beliefs Among
American Opinion Leaders - After the Cold War.” Millennium: Journal of International Studies
22 (2): 235-278.
26
Holsti, Ole R., and James N. Rosenau. 1996. “Liberals, Populists, Libertarians, and Conservatives:
The Link between Domestic and International A↵airs.” International Political Science Review
17 (1): 29-54.
Hu↵meier, Joachim, and Phillip Alexander Freund. 2014. “Being Tough or Being Nice? A MetaAnalysis on the Impact of Hard- and Softline Strategies in Distributive Negotiations.” Journal of
Management 40 (3): 72-99.
Hummel, Patrick. 2010. “Flip-flopping from primaries to general elections.” Journal of Public Economics 94 (1112): 1020 - 1027.
Huth, Paul K. 1988. Extended deterrence and the prevention of war. Yale University Press.
Imai, Kosuke, Luke Keele, Dustin Tingley, and Teppei Yamamoto. 2010. “Causal Mediation Analysis
Using R.” In Advances in Social Science Research Using R, ed. H. D. Vinod. New York: SpringerVerlag.
Jentleson, Bruce W, and Rebecca L Britton. 1998. “Still Pretty Prudent Post-Cold War American
Public Opinion on the Use of Military Force.” Journal of Conflict Resolution 42 (4): 395–417.
Jones,
cal
Je↵rey.
Cli↵.”
2012.
Gallup.
“Americans
December
Widely
4,
Prefer
Compromise
to
Fis-
http://www.gallup.com/poll/159065/
americans-widely-prefer-compromise-fiscal-cliff.aspx.
Jost, John T., Brian A. Nosek, and Samuel D. Gosling. 2008. “Ideology: Its Resurgence in Social,
Personality, and Political Psychology.” 3 (2): 126-136.
Jost, John T., Christopher M. Federico, and Jaime L. Napier. 2009. “Political Ideology: Its Structure,
Functions, and Elective Affinities.” Annual Review of Psychology 60: 307-337.
Jost, John T., Jack Glaser, Arie W. Kruglanski, and Frank Sulloway. 2003. “Political Conservatism
as Motivated Social Cognition.” Psychological Bulletin 129 (3): 339-375.
Jost, John T., Jaime L. Napier, Hulda Thorisdottir, Samuel D. Gosling, Tibor P. Palfai, and Brian
Ostafin. 2007. “Are Needs to Manage Uncertainty and Threat Associated With Political Conservatism or Ideological Extremity?” Personality and Social Psychology Bulletin 33 (7): 989-1007.
27
Kertzer, Joshua D, and Brian C Rathbun. 2015. “Fair is Fair: social Preferences and reciprocity in
international Politics.” World Politics 67 (04): 613–655.
Kertzer, Joshua D, Kathleen E Powers, Brian C Rathbun, and Ravi Iyer. 2014. “Moral support:
How moral values shape foreign policy attitudes.” The Journal of Politics 76 (03): 825–840.
Kertzer, Joshua D, and Ryan Brutger. 2016. “Decomposing audience costs: bringing the audience
back into audience cost theory.” American Journal of Political Science 60 (1): 234–249.
King, Carolyn, and William Mauldin. 2016. “TransCanada Starts Legal Actions Over Keystone XL
Pipeline Denial.” Wall Street Journal. Online, January 6.
Kremer, John, and David Mack. 2013. “Pre-emptive Game Behaviour and the Emergence of Leadership.” British Journal of Social Psychology 22 (1): 19-26.
Kurizaki, Shuhei. 2007. “Efficient Secrecy: Public versus Private Threats in Crisis Diplomacy.”
American Political Science Review 101 (3): pp. 543-558.
Kydd, Andrew. 1997. “Game theory and the spiral model.” World Politics 49 (03): 371–400.
Landler, Mark. 2012. “Obama Threatens Force Against Syria.” The New York Times. August 20.
Leng, Russell J. 1993. “Reciprocating Influence Strategies in Interstate Crisis Bargaining.” Journal
of Conflict Resolution 37 (1): 3–41.
Levendusky, Matthew S., and Michael C. Horowitz. 2012. “When Backing Down is the Right Decision: Partisanship, New Information, and Audience Costs.” Journal of Politics 74 (2): 323-338.
Levy, Jack S., Michael K. McKoy, Paul Poast, and Geo↵rey P.R. Wallace. 2015. “Backing Out or
Backing In? Commitment and Consistency in Audience Costs Theory.” American Journal of
Political Science Online.
Malhotra, Neil, Yotam Margalit, and Cecilia Hyunjung Mo. 2013. “Economic Explanations for Opposition to Immigration: Distinguishing Between Prevalence and Conditional Impact.” American
Journal of Political Science 57 (2): 391–410.
Mansfield, Edward D, and Diana C Mutz. 2009. “Support for free trade: Self-interest, sociotropic
politics, and out-group anxiety.” International Organization 63 (03): 425–457.
28
McManus, Roseanne. 2014. “Fighting Words: The E↵ectiveness of Statements of Resolve in International Conflict.” Journal of Peace Research 51 (6): 726-740.
Mo, Jongryn. 1995. “Domestic institutions and international bargaining: The role of agent veto in
two-level games.” American Political Science Review 89 (04): 914–924.
Morgenthau, Hans J. 1948. Politics among Nations - The Struggle For Power and Peace. 1 ed. New
York: Alfred A. Knopf, Inc.
Mutz, Diana, and Eunji Kim. 2016. “The Impact of Ingroup Favoritism on Trade Preferences.” International Organization, Forthcoming. http://papers.ssrn.com/sol3/papers.cfm?abstract_
id=2818632.
Nicholson, Stephen P. 2012. “Polarizing Cues.” American Journal of Political Science 56 (1): 52-66.
Osgood, Charles E. 1962. “An Alternative to War or Surrender.”.
Petrocik, John. 1980. Some Contextual Sources of Voting Behavior: The Changeable American
Voter. Sage chapter in T he Electorate Reconsidered, edited by John Pierce and John Sullivan,
pp. 257–277.
Pew Research. 2013. “Public Backs Diplomatic Approach in Syria, But Distrusts Syria and Russia.”
Pew Research Center for the People and The Press. September 16.
Polling Report. 2014. “President Obama and the Obama Administration.” PollingReport.com. http:
//www.pollingreport.com/obama_ad.htm.
Poole, Keith T. 2007. “Changing Minds? Not in Congress!” Public Choice 131 (3-4).
Potter, Philip, and Matthew Baum. 2010. “Democratic Peace, Domestic Audience Costs, and Political Communication.” Political Communication 27 (4): 453-470.
Potter, Philip, and Matthew Baum. 2014. “Looking for Audience Costs in all the Wrong Places:
Electoral Institutions, Media Access, and Democratic Constraint.” Journal of Politics 76 (1):
167-181.
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.” International
Studies Quarterly 60 (1): 124–137.
29
Saunders, Elizabeth N. 2015. “War and the Inner Circle: Democratic Elites and the Politics of Using
Force.” Security Studies Forthcoming. http://home.gwu.edu/~esaunder/warinnercircle.pdf.
Schultz, Kenneth. 2005. “The Politics of Risking Peace: Do Hawks or Doves Deliver the Olive
Branch?” International Organization 59 (1): 1-38.
Sink, Justin. 2014. “Panetta: Obama’s ‘Red Line’ on Syria Damaged US Credibility.” The Hill.
October 7.
Slantchev, Branislav L. 2006. “Politicians, the Media, and Domestic Audience Costs.” International
Studies Quarterly 50 (2): 445-477.
Smith, Alastair. 1998. “International Crises and Domestic Politics.” American Political Science
Review 92 (3): 623-638.
Snyder, Jack, and Erica Borghard. 2011. “The Cost of Empty Threats: A Penny Not a Pound.”
American Political Science Review 105 (3): 437-456.
Stasavage, David. 2004. “Open-door or closed-door? Transparency in domestic and international
bargaining.” International Organization 58 (04): 667–703.
Tarar, Ahmer, and Bahar Leventouglu. 2009. “Public Commitment in Crisis Bargaining.” International Studies Quarterly 53 (3): 817-839.
Tarar, Ahmer, and Bahar Leventouglu. 2013. “Limited Audience Costs in International Crises.”
Journal of Conflict Resolution 57 (6): 1065-1089.
Tomz, Michael. 2007. “Domestic Audience Costs in International Relations: An Experimental Approach.” International Organization 61 (Autumn): 821-840.
Tomz, Mike, and Robert Van Houweling. 2012. “Political Repositioning - Detailed Synopsis.”. http:
//web.stanford.edu/~tomz/working/TomzVanHouweling_PoliticalRepositioning.pdf.
Trager, Robert, and Lynn Vavreck. 2011. “The Political Costs of Crisis Bargaining: Presidential
Rhetoric and the Role of Party.” American Journal of Political Science 55 (June): 526-545.
Trubowitz, Peter, and Nicole Mellow. 2005. ““Going Bipartisan”: Politics by Other Means.” Political
Science Quarterly 120 (Fall): 433-453.
30
Van Vugt, Mark. 2006. “Evolutionary Origins of Leadership and Followership.” Personality and
Social Psychology Review 10 (4): 354-371.
Weeks, Jessica. 2008. “Autocratic Audience Costs: Regime Type and Signaling Resolve.” International Organization 62 (January): 35-64.
Yarhi-Milo, Keren. 2013. “Tying Hands Behind Closed Doors: The Logic and Practice of Secret
Reassurance.” Security Studies 22 (3): 405-435.
Zavala, Agnieszka Golec De, Aleksandra Cislak, and Elzbieta Wesolowska. 2010. “Political Conservatism, Need for Cognitive Closure, and Intergroup Hostility.” Political Psychology 31 (4):
521-541.
31
Domestic Politics of International Compromise
Supplementary Appendix
Ryan Brutger⇤
Contents
1 Sample Demographics and Methodology
2
2 Comparing Results of Experiments 2 and 3
5
3 The Impact of Holding Outcomes Constant
6
4 Experiment 2 - Treatments
8
5 Average Approval Not Conditioned on Party of the President.
10
6 Results Based on Percent Approving
11
7 Full Results for Experiments 1 and 2
13
8 E↵ects Among Independents
15
9 Mediation Analysis - Reputation and Support for Compromise.
16
10 Experiment 4 - Treatments
17
⇤ Assistant Professor, University of Pennsylvania, Department of Political Science, Email: [email protected].
Web: http://web.sas.upenn.edu/brutger/.
1
1
Sample Demographics and Methodology
Choosing Experiments:
In addition to the discussion in the main text, it is worth noting that an additional advantage of
using survey experiments to test the e↵ects of compromise and proposal power, is that survey experiments represent the most likely case to observe audience costs, making them a harder case for the
ameliorating e↵ects of compromise and proposal power. In traditional audience cost experiments the
leader’s threat and subsequent inconsistency are temporally proximate and thus audiences should
be most likely to punish leaders who back down, whereas in real-world settings audiences may have
more difficulty connecting a past threat to the leader’s current choice of action. Furthermore, respondents are not presented with competing messages about the efficacy or value of the engagement,1
which means respondents are more likely to focus on the leader’s threat and subsequent inconsistency. For each of these reasons, experiments represent the most likely situation for leaders to be
punished, making them an ideal manner in which to test whether compromise and proposal power
can overcome the negative e↵ect of leaders’ inconsistency.
Sample Sizes and Recruitment Methodology:
Experiment 1: Fielded using Amazon Mechanical Turk in 2014 on a sample of 916 respondents.
Experiment 2: Fielded using Amazon Mechanical Turk in 2013 on a sample of 1761 respondents.
Experiment 3: Fielded on a nationally representative sample of 613 respondents using Survey Sampling International.
Experiment 4: Fielded on a nationally representative sample of 543 respondents using Survey Sampling International.
SSI uses an opt-in recruitment method, after which they randomly select panel participants for
survey invitations, using population targets rather than quotas to produce nationally representative
samples of respondents.
1 See Levendusky and Horowitz (2012) for an exception, where respondents are presented with new information
justifying the change in policy.
2
Limits and Advantages of mTurk samples:
Samples from mTurk are increasingly widely used in both political science and elsewhere in the
social sciences because of the extent to which they can replicate classic experiments on more diverse
samples than those traditionally employed in political psychology research, as recent review pieces
by Berinksy et al. (2012), Buhrmester, Kwang, and Gosling (2011), Horton, Rand, and Zeckhauser
(2011), Paolacci, Chandler, and Iperiotis (2010), and Rand (2012) make clear. As is common in
online surveys, the mTurk samples are more male, educated and liberal than the American population
as a whole – 56.2% of respondents were male, 58.9% have at least a four-year college degree, and
51.3% identify themselves as being liberals in Experiment 2. However, the samples are particularly
useful, since I focus my analysis on specific partisan subgroups, and thus draw conclusions contingent
on the relevant subgroup. Since the political skew of the sample is a primary concern in many
political science applications of mTurk surveys, I explicitly focus on within-party treatment e↵ects,
as opposed to focusing on national averages. For this reason, I do not employ survey weights to create
a synthetically representative sample. Furthermore, across comparable conditions in experiments 2
and 3, the di↵erence in average approval scores was essentially zero (0.005, p=0.98), highlighting
that the nationally representative and mTurk samples generated remarkably consistent results.
3
Sample Populations
Table 1: Sample Populations
Experiment 1
Experiment 2
Experiment 3
Experiment 4
25.4
63.1
10.3
1.2
19.1
63.6
15.7
1.6
11.6
36.7
33.8
20.8
9.5
32.9
38.9
21.3
IN COM E
Percent $0-$50,000
Percent $50,000-$100,000
Percent $100,000-$150,000
Percent $150,000-$200,000
Percent $200,000+
60.0
30.7
5.8
2.2
1.3
56.1
34.8
5.6
2.2
1.5
46.8
33.5
11.7
4.8
3.3
43.9
33.1
12.0
8.3
2.7
EDU CAT ION
Less than High School
High School / GED
Some College
College Degree
Masters Degree
PhD / JD / MD
0.5
12.7
31.1
47.5
5.7
2.5
0.7
10.9
29.4
48.4
8.0
2.5
2.6
27.5
11.4
39.8
13.6
5.2
3.1
25.5
13.1
38.6
15.0
4.6
AGE
Percent
Percent
Percent
Percent
Age
Age
Age
Age
18-24
25-44
45-64
65+
Not all percentages add to 100 due to rounding
.
4
2
Comparing Results of Experiments 2 and 3
Figure 1: Experiment 3 - Average Approval Score for Each Strategy
1.5
Audience of:
Democrats
Republicans
1.0
0.5
●
●
−1.0
−0.5
0.0
●
−1.5
Average Approval Score
●
Party Match
President Proposes
Party Match
Foreign Leader
Proposes
No Party Match
President Proposes
No Party Match
Foreign Leader
Proposes
Figure 1 shows the average approval score and 90 percent confidence intervals for audiences of Democrats
and Republicans based on whether the US leader or foreign leader proposed the compromise. The
approval score is measured on a scale from -3=Srongly Disapprove to 3=Strongly Approve.
Comparing the results displayed in Figure 1 of the appendix to those in Figure 2 of the main
paper illustrates that the e↵ects are remarkably consistent across the mTurk and SSI samples. In
each study, proposal power and partisanship lead to the highest approval scores for leaders who
initiate the compromise among respondents of their own political party. These results highlight
that di↵erences in wording, and potential concerns about the wording in Experiments 1 and 2, do
not alter the fundamental results of the paper. Due to a randomization error in Experiment 3, no
treatment condition with a Republican president and the foreign leader proposing the compromise
was presented. However, there were still six relevant treatment groups, as shown in Figure 1, that
provide a useful comparison of results across samples and treatment wordings, which demonstrate
that the main results are consistent across all three security experiments.
5
3
The Impact of Holding Outcomes Constant
The Impact of Holding Casualties Constant:
In order to isolate the treatment e↵ect of the leader’s strategy, it is necessary that casualties are
held constant at zero across all conditions, otherwise respondents’ approval could shift because of
di↵erent outcomes as opposed to strategies. In the control and treatment conditions that do not
result in military engagement (Stay Out, Not Engage, and Compromise) casualties are obviously at
zero, and thus no mention of casualties is included. However, in the Engage condition respondents
may assume higher costs resulting from war,2 even if outcomes are not explicitly varied (for a
related point, see Sher and McKenzie (2006) and Tomz and Weeks (2013) on “information leakage”).
To avoid conflating respondent’s expectations about the human cost of war with approval for the
leader’s strategy, I specify that casualties are zero in the engage condition. This inclusion means
that the treatment conditions are not lexically equivalent, but instead they are logically equivalent.
This technique was also used by Kertzer and Brutger (2016) who tested the e↵ect of making the
conditions lexically equivalent by omitting any mention of casualties in the Engage condition. Kertzer
and Brutger found that audiences do indeed impose a higher cost when casualties are not mentioned,
which means I am choosing a more conservative measure here, since any bias introduced by specifying
casualties at zero would make the Engage treatment relatively more appealing to audiences (meaning
that Compromise would be relatively less appealing).
The Impact of Holding Territory Constant:
Territorial outcomes are held constant with the attacking country gaining 20 percent of the territory
so that there is common support across all treatment conditions and the support is realistic given
the strategy of the president. Since a compromise means that neither party completely backs down
from their position, the territorial outcome cannot be set to 0 percent for the attacking country.
This does raise a potential concern, since any military engagement where the other party receives
anything could be viewed as a loss. This means that some respondents may interpret the common
support as a loss in the engage condition, which would put downward pressure on approval in the
engage treatment condition.
Three points are worth noting in response to this concern. First, although the territorial outcome
2 This would be expected, given that the crisis bargaining model explicitly includes a cost of war that is separate
from the division of the contested prize.
6
could put downward pressure on approval since the attacking country was allowed to gain some
territory, this concern should apply to all treatment conditions. If a respondent views the 20 percent
territorial outcome as a loss for the U.S., it should be viewed as a loss in all conditions. If the
experiment had di↵erent territorial outcomes more favorable to the U.S. in the engage condition
than the others, then it would be conflating approval for the outcome itself with the strategy used
to achieve it. That said, the question of whether the salience of the perceived loss is heightened
when the U.S. military is engaged, which may change expectations of the public, is one that deserves
further analysis, but remains beyond the scope of this project.
Second, given that the scenario holds casualties to zero, any potential downward bias due to the
territorial outcome will be partially o↵set by the optimistic outcome of having a military engagement
with zero casualties. As is discussed above, the scenario represents a conservative test in this manner,
since zero casualties in the engage condition puts upward pressure on the approval score in the Engage
condition, making the comparison of the Compromise treatment to the Engage condition an even
harder test for this study than would likely occur in a real conflict where casualties would be a
realistic possibility.
Lastly, the concern raised only a↵ects comparisons between the Engage and Compromise treatments, but has no bearing on the tests of proposal power and partisanship. Even if the two proceeding
points do not alleviate the readers concerns, the vast majority of results in the paper do not rely
on comparisons with the Engage condition. The only section of the paper analyzing the Engage
condition is the Sensitivity to Outcomes Section, which means all other results are una↵ected and
thus those results cannot be biased by the Engage treatment.
7
4
Experiment 2 - Treatments
Experiment 2 followed a similar format at Experiment 1, but includes additional treatments to test
how sensitive audiences are to crisis outcomes. Each of the treatment conditions varies:
1) The strategy of the president
2) The partisanship of the president
3) The outcomes of the crisis
Within the Compromise condition, there is an additional treatment that varies whether the U.S.
president or the leader of the foreign country proposed the compromise. The crisis outcome treatments, which vary both the territory gained and casualties, are designed with two goals in mind.
The first is to have common support across all conditions, meaning that there is a common policy
outcome across all treatments. This allows the e↵ect of the other treatments to be examined while
holding outcomes constant. This study follows Kertzer and Brutger (2016) and uses 20 percent of
the territory gained by the attacking country and no casualties as the common outcome. The second
goal of the treatment is to have the variation in outcomes follow realistic conditions, meaning that
shifts away from the common policy outcomes reflect that the U.S. receives better outcomes when
they use force than when they don’t. Additionally, the casualty treatments are limited to the Engage
condition, since it is not realistic to have military casualties when the U.S. does not engage. For a
further discussion on casualties, please see appendix, section 3.
• Stay Out: “The U.S. president, who was a [Democrat or Republican], said that the United
States would stay out of the conflict. The attacking country continued to invade and took
control of [20 or 50] percent of the contested territory. ”
Consistent with Experiment 1, in each of the remaining treatment conditions the U.S. escalated with
a threat and then the respondents were randomly assigned to one of the following conditions.
Threat: “The U.S. president, who was a [Democrat or Republican], said that if the
attacking country continued to invade, the United States military would immediately
engage and attempt to push out the attacking country. The president sent troops to the
region.”
• Compromise: “The attacking country continued to invade, but the president did not immediately engage. The [U.S. president or leader of the attacking country] proposed a settlement,
which was agreed to by all parties, where the attacking country stopped its invasion and took
control of [20 or 50] percent of the contested territory.”
8
• Engage: “The attacking country continued to invade and the U.S. president ordered the military to engage. [The U.S. did not loose any troops or The U.S. lost 1,000 troops] in the conflict
and the conflict ended with the attacking country [not taking control of any of the contested
territory or taking control of 20 percent of the contested territory].”
• Not Engage: “The attacking country continued to invade. The U.S. president ordered the
military not to engage. The attacking country continued to invade and took control of [20 or
50] percent of the contested territory.”
Experiment 2 also di↵ers from Experiments 1 and 3, since Experiments 1 and 3 did not include
the language “where the attacking country stopped its invasion” in the compromise condition. The
language in Experiment 2 adds emphasis to the fact that the settlement stopped the attack, which
could be argued to bias results in favor of the compromise condition. To test whether this di↵erence
in language biased the results, the comparable results are shown in the main paper in figure 1.
Figure 1 shows that average approval in the compromise conditions is not higher in Experiment 2
than Experiment 1, alleviating concerns that approval for compromise is driven by this di↵erence in
language. In aggregate, the results across all four experiments show that the compromise treatment
e↵ects are not conditioned by changes in the issue area or modest changes to the wording of the
scenario.
9
Average Approval Not Conditioned on Party of the President.
Average Approval Score for Each Strategy
2
Figure 2:
1
Experiment 1
Experiment 2
●
0
●
●
●
−1
Average Approval Score
●
−2
5
Stay Out
Engage
Not Engage
Compromise
Figure 2 shows the average approval score and 90 percent confidence intervals for experiments 1 and
2 for each randomly assigned crisis strategy. Outcomes are held constant with no casualties and 20%
territory gained by the attacking country. The approval score is measured on a scale from -3=Strongly
Disapprove to 3=Strongly Approve.
10
6
Results Based on Percent Approving
Figures 3 and 4 replicate the top and bottom panels of Figure 1 from the main paper, but show the
percent of respondents who approve of the president’s handling of the crisis instead of the seven point
approval score. The results are generally consistent regardless of whether the dependent variable is
a dichotomous measure of approval or the full approval score used in the main analysis.
80
Figure 3: Percent Approving for Each Strategy
60
●
50
●
40
●
30
Percent Approving
70
Republican President
Experiment 1
● Experiment 2
20
●
Stay Out
Engage
Not Engage
Compromise
Figure 3 shows the percent of respondents who approve and 90 percent confidence intervals for audiences
of Democrats and Republicans based when a Republican is president. Approval is measured as those
who lean toward approving, somewhat approve, or strongly approve.
80
Figure 4: Percent Approving for Each Strategy
●
50
60
●
40
●
●
20
30
Percent Approving
70
Democrat President
Experiment 1
● Experiment 2
Stay Out
Engage
Not Engage
Compromise
Figure 4 shows the percent of respondents who approve and 90 percent confidence intervals for audiences
of Democrats and Republicans based when a Democrat is president. Approval is measured as those who
lean toward approving, somewhat approve, or strongly approve.
11
Figure 5 replicates Figure 2 from the main paper, but shows the percent of respondents who
approve of the president’s handling of the crisis instead of the seven point approval score. The results
are generally consistent regardless of whether the dependent variable is a dichotomous measure of
approval or the full approval score used in the main analysis.
80
Figure 5: Percent Approving for Each Strategy
Audience of:
Democrats
Republicans
60
●
●
50
●
40
●
20
30
Percent Approving
70
●
Party Match
President Proposes
Party Match
Foreign Leader
Proposes
No Party Match
President Proposes
No Party Match
Foreign Leader
Proposes
Figure 5 shows the percent of respondents who approve and 90 percent confidence intervals for audiences
of Democrats and Republicans based on whether the respondent and the president are from the same
party (party-match) interacted with whether the U.S. president or the foreign leader proposed the
negotiated settlement. Approval is measured as those who lean toward approving, somewhat approve,
or strongly approve.
12
7
Full Results for Experiments 1 and 2
Table 2: Experiment 1: Di↵erences in Approval Score for Strategies and Party
Di↵erence in Means
Audience of Democrats &
President who is Democrat
Stay Out
Engage
Not Engage
-0.42
(-0.75)
-2.17***
(-4.94)
-1.75***
(-3.11)
-0.60
(-1.50)
-0.18
(0.33)
1.57***
(3.94)
-1.33***
(-2.82)
-1.71***
(-3.65)
-0.38
(-0.73)
-0.37
(-1.03)
0.96**
(2.23)
1.34***
(3.15)
0.33
(0.28)
-2.00**
(-2.31)
-2.33*
(-1.91)
-0.93
(-1.22)
-1.26
(-1.10)
1.07
(1.37)
0.61
(0.50)
-1.75
(-1.68)
-2.36
(-1.66)
-1.36*
(-1.85)
-1.97
(-1.63)
0.39
(0.38)
Engage
Not Engage
Audience of Democrats &
President who is Republican
Stay Out
Engage
Not Engage
Audience of Republicans &
President who is Republican
Stay Out
Engage
Not Engage
Audience of Republicans &
President who is Democrat
Stay Out
Engage
Not Engage
Compromise
⇤ p<0.1; ⇤⇤ p<0.05; ⇤⇤⇤ p<0.01
Note:
Table 2 shows the di↵erence in average approval scores (Column Score Row Score)
and t-statistics among Democrats and Republicans based on strategies and the party of the
president while holding territorial and casualty outcomes constant at no casualties and 20%
territory gained by the attacking country. The approval score is measured on a scale from
-3=Srongly Disapprove to 3=Strongly Approve.
13
Table 3: Experiment 2: Di↵erences in Approval Score for Strategies and Party
Di↵erence in Means
Audience of Democrats &
President who is Democrat
Stay Out
Engage
Not Engage
-0.08
(-0.24)
-1.27***
(-4.44)
-1.35***
(-3.97)
-0.45
(-1.53)
0.53
(1.53)
0.82***
(2.78)
0.03
(0.10)
-1.67***
(-6.22)
-1.70***
(-5.00)
-0.90***
(-3.15)
-0.93***
(-2.64)
0.76***
(2.61)
0.97
(1.55)
-1.81***
(-3.51)
-2.78***
(-4.61)
-0.31
(-0.62)
-1.27**
(-2.16)
1.50***
(3.16)
0.70
(1.15)
-1.36**
(–2.45)
-2.07***
(-2.37)
-0.92*
(-1.76)
-1.62***
(-2.79)
1.50***
(3.16)
Engage
Not Engage
Audience of Democrats &
President who is Republican
Stay Out
Engage
Not Engage
Audience of Republicans &
President who is Republican
Stay Out
Engage
Not Engage
Audience of Republicans &
President who is Democrat
Stay Out
Engage
Not Engage
Compromise
⇤ p<0.1; ⇤⇤ p<0.05; ⇤⇤⇤ p<0.01
Note:
Table 3 shows the di↵erence in average approval scores (Column Score Row Score)
and t-statistics among Democrats and Republicans based on strategies and the party of the
president while holding territorial and casualty outcomes constant at no casualties and 20%
territory gained by the attacking country. The approval score is measured on a scale from
-3=Srongly Disapprove to 3=Strongly Approve.
14
8
E↵ects Among Independents
Table 4: Experiment 1: Di↵erences in Approval Score for Strategies and Party
Di↵erence in Means
Audience of Independents &
President who is Democrat
Stay Out
Engage
Not Engage
-1.12*
(-1.96)
-1.63***
(-3.28)
-0.51
(-0.85)
-0.85**
(-2.05)
0.28
(0.52)
0.79*
(1.75)
-1.11*
(-1.95)
-2.75***
(-6.53)
-1.64***
(-3.00)
-1.49***
(-3.71)
-0.38
(-0.72)
1.26**
(3.38)
Engage
Not Engage
Audience of Independents &
President who is Republican
Stay Out
Engage
Not Engage
Compromise
⇤ p<0.1; ⇤⇤ p<0.05; ⇤⇤⇤ p<0.01
Note:
Table 4 shows the di↵erence in average approval scores (Column Score Row Score)
and t-statistics among Independents based on strategies and the party of the
president while holding territorial and casualty outcomes constant at no casualties and 20%
territory gained by the attacking country. The approval score is measured on a scale from
-3=Srongly Disapprove to 3=Strongly Approve.
15
9
Mediation Analysis - Reputation and Support for Compromise.
Figure 6: Average Causal Mediation Analysis:
The E↵ect of Repuation on Support for Compromise
ACME
ADE
Total
Effect
0.0
0.2
0.4
0.6
0.8
1.0
1.2
Figure 4 shows the average causal mediation e↵ect (ACME) of reputation on audience approval, the
average direct e↵ect (ADE) of the Compromise treatment (relative to Not Engage), the total e↵ect
of both the mediation and treatment e↵ects, and 95 percent confidence intervals, calcualted using the
mediation package in R (Tingley et al., 2014). Outcomes are held constant with no casualties and 20%
territory gained by the attacking country. The approval score is measured on a scale from -3=Strongly
Disapprove to 3=Strongly Approve, and the Reputation score is measured from 1 to 5.
16
10
Experiment 4 - Treatments
Experiment 4 presented respondents with an investment dispute, which was presented in the format
of a brief news report. The dependent variable is the same as in the first three experiments, where
respondents were asked whether they approved, disapproved, or neither, and then how strongly they
felt that way. The text of the experiment is as follows:
Company Starts Legal Actions Over Investment Denial
TransCorp., a hypothetical company based in a neighboring country, on Wednesday said
it was pursuing legal actions against the United States and the Obama administration
in response to its refusal to issue a border-crossing permit for the company’s project.
TransCorp said in a statement that it would initiate an international arbitration case
against the U.S. under an international agreement. Through a process known as investorstate dispute settlement (ISDS), companies and investors from one country can challenge
the acts of a foreign government and receive compensation if they can show they weren’t
treated in accordance with international law.
TransCorp said it would attempt to recover more than $15 billion in costs and damages
that the company said it has su↵ered as a result of the U.S. administration’s breach of
its international obligations.
The final paragraph of the report was randomly assigned, showing either the “U.S. proposes”,
“foreign proposes”, or “U.S. proposes, without threat” treatments, which are shown below.
• U.S. Proposes: The Obama administration originally responded by stating it would fight
the challenge until the arbitration panel made its decision. The Obama administration has
since proposed a settlement granting TransCorp twenty percent of the value of the suit, and
TransCorp accepted the settlement.
• Foreign Proposes: The Obama administration originally responded by stating it would fight
the challenge until the arbitration panel made its decision. TransCorp has since proposed
a settlement granting TransCorp twenty percent of the value of the suit, and the Obama
administration accepted the settlement.
17
• U.S. Proposes, Without Threat: The Obama administration proposed a settlement granting
TransCorp twenty percent of the value of the suit, and TransCorp accepted the settlement.
References
Berinksy, Adam, Gregory Huber, Gabriel Lenz, and Edited by R. Michael Alvarez. 2012. “Evaluating
Online Labor Markerts for Experimental Research: Amazon’s Mechanical Turk.” Political Analysis
20 (3): 351-368.
Buhrmester, Michael, Tracy Kwang, and Samuel D. Gosling. 2011. “Amazon’s Mechanical Turk: A
New Source of Inexpensive, Yet High-Quality, Data?” Perspectives on Psychological Science 6 (1):
3-5.
Horton, John J., David G. Rand, and Richard J. Zeckhauser. 2011. “The online laboratory: conducting experiments in a real labor market.” Experimental Economics 14 (3): 399-425.
Kertzer, Joshua D, and Ryan Brutger. 2016. “Decomposing audience costs: bringing the audience
back into audience cost theory.” American Journal of Political Science 60 (1): 234–249.
Levendusky, Matthew S., and Michael C. Horowitz. 2012. “When Backing Down is the Right Decision: Partisanship, New Information, and Audience Costs.” Journal of Politics 74 (2): 323-338.
Paolacci, Gabriele, Jesse Chandler, and Panagiotis G. Iperiotis. 2010. “Running experiments on
Amazon Mechanical Turk.” Judgment and Decision Making 5 (5): 411-419.
Rand, David G. 2012. “The promise of Mechanical Turk: How online labor markets can help theorists
run behavioral experiments.” Journal of Theoretical Biology 299: 172-179.
Sher, Shlomi, and Craig R.M. McKenzie. 2006. “Information leakage from logically equivalent
frames.” Cognition 101: 467-494.
Tingley, Dustin, Teppei Yamamoto, Kentaro Hirose, Luke Keele, and Kosuke Imai. 2014. “mediation:
R Package for Causal Mediation Analysis.” Journal of Statistical Software 59 (5): 1–38.
Tomz, Michael R., and Jessica L. P. Weeks. 2013. “Public Opinion and the Democratic Peace.”
American Political Science Review 107 (4): 849-865.
18