Jonathan A. Chu August 28, 2017 Every armed humanitarian intervention since the end of the Cold War has been conducted under the auspices of an international organization (IO). Some authors explain this striking observation by pointing to the ability of IOs to mobilize mass support for war. Building on these domestic approaches, this article develops and tests a novel theory of how IOs influence public opinion. It argues that IOs can win citizen policy approval by sending cues about the social repercussions of military intervention, but such cues can only be effectively sent by IOs that represent the political identities held by citizens. Evidence from historical polls and original survey experiments regarding American opinion on humanitarian war validate the theory’s claims. The findings contribute to debates on the domestic channels of IO influence, the value of multilateralism in foreign policy, and the role of identity in world politics. Jonathan Chu is a Ph.D. candidate in the Department of Political Science at Stanford University ([email protected]). I thank Kirk Bansak, Matthew Baum, Songying Fang, Moritz Marbach, Mark Jacobsen, Risa Kitagawa, Steve Krasner, Melissa Lee, Philip Lipscy, Adam Liu, Julia Payson, Phil Potter, Jonathan Renshon, Piki Ish-Shalom, Ken Schultz, Paul Sniderman, Mike Tierney, Mike Tomz, and panel participants at Stanford, UC Merced, ISA, HEWG, MPSA, APSA, PSS, and EPSA for their feedback. I am also grateful for the support of Stanford’s Laboratory for the Study of American Values and the NSF Graduate Research Fellowship (Grant #DGE-114747). All views and errors remain my own. Stanford IRB Protocol #32333.
65
Embed
Jonathan A. Chu August 28, 2017 - Harvard Universitypeople.fas.harvard.edu/~jkertzer/HISC2017/schedule/papers/Chu.pdf · Jonathan A. Chu August 28, 2017 Every armed humanitarian intervention
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Jonathan A. Chu
August 28, 2017
Every armed humanitarian intervention since the end of the Cold War has been conducted
under the auspices of an international organization (IO). Some authors explain this striking
observation by pointing to the ability of IOs to mobilize mass support for war. Building on
these domestic approaches, this article develops and tests a novel theory of how IOs influence
public opinion. It argues that IOs can win citizen policy approval by sending cues about the
social repercussions of military intervention, but such cues can only be effectively sent by IOs
that represent the political identities held by citizens. Evidence from historical polls and
original survey experiments regarding American opinion on humanitarian war validate the
theory’s claims. The findings contribute to debates on the domestic channels of IO influence,
the value of multilateralism in foreign policy, and the role of identity in world politics.
Jonathan Chu is a Ph.D. candidate in the Department of Political Science at Stanford University
([email protected]). I thank Kirk Bansak, Matthew Baum, Songying Fang, Moritz Marbach, Mark
Jacobsen, Risa Kitagawa, Steve Krasner, Melissa Lee, Philip Lipscy, Adam Liu, Julia Payson, Phil Potter,
Jonathan Renshon, Piki Ish-Shalom, Ken Schultz, Paul Sniderman, Mike Tierney, Mike Tomz, and panel
participants at Stanford, UC Merced, ISA, HEWG, MPSA, APSA, PSS, and EPSA for their feedback. I am
also grateful for the support of Stanford’s Laboratory for the Study of American Values and the NSF Graduate
Research Fellowship (Grant #DGE-114747). All views and errors remain my own. Stanford IRB Protocol
#32333.
Humanitarian wars are almost always conducted under the auspices of an international
organization (IO).1 But why would countries, especially superpowers like the United States,
bother seeking the approval of an IO? After all, the process of obtaining institutional
authorization could limit a government’s policy autonomy, and in cases like Rwanda and Syria,
inaction in light of IO opposition could allow the brutal slaughter of innocent lives.
On this question, a flourishing research agenda finds that governments prize IOs for
their role in winning the policy approval of domestic audiences.2 When an IO recommends a
policy, citizens are more likely to support that policy, and political leaders care because popular
opinion often translates to votes during elections, perceptions of foreign policy competence,
and clout in other political arenas.3 The extant scholarship outlines two main reasons why IOs
can sway mass opinion on war. Legal approaches claim that IOs, depending on their status in
international law, can grant legal legitimacy on the use of force.4 In contrast, rationalist
approaches contend that IOs, depending on their membership’s preference distribution, can
signal information about the material consequences of a policy.5 By investigating the domestic
channels through which IOs affect international affairs, these authors make a fundamental
contribution to theories of international institutions, which have traditionally focused state-to-
state channels of IO influence.6
To advance the scholarship on the domestic politics of IOs, this paper develops and
tests a new explanation for how IOs affect mass opinion. Drawing from theories of social
identity and socialization,7 it argues that policy recommendations by IOs serve as cues about a
policy’s social repercussions and appropriateness. An IO’s policy endorsement answers
2011; Grieco et al. 2011; Tingley and Tomz 2012. 3 Hurwitz and Peffley 1987; Aldrich, Sullivan, and Borgida 1989; Gartner, Segura, and Barratt 2004; Gelpi
et al. 2009; Gelpi and Grieco 2015. On the effect of public opinion on foreign policy in democratic countries
more generally, see Risse-Kappen 1991; Sobel 2001; Holsti 2004; Aldrich et al. 2006; Weeks 2008; Stein
2014; Baum and Potter 2015; Milner and Tingley 2016. 4 Koh 1997; Tago and Ikeda 2015, 392. 5 Thompson 2009; Chapman 2011. 6 E.g. Keohane 1984; Abbott and Snidal 1998; Hurd 2008. 7 E.g. Johnson 2008; Abdelal et al. 2009.
question such as, does the policy (i.e. humanitarian intervention) represent pro-norm behavior?
Will peer countries participate to advance the policy? Will enacting this policy improve my
nation’s status and reputation? The theory further argues the ability of IOs to send these social
cues depends on the political identities they represent. Citizens are thus more susceptible to
IOs with which they identify.
This article examines this argument in the context of how the UN Security Council
(UNSC) and NATO affect U.S. public opinion on humanitarian war. The conventional wisdom
implies that the UNSC is crucial for garnering public support. After all, only the UNSC can
satisfy international legal requirements, and because it sparingly authorizes war, its policy
recommendations arguably indicate that military force is merited and will have good
consequences. The social identity view alternatively hypothesizes that the UNSC is not so
crucial. For citizens of liberal democracies like the U.S., NATO embodies the collective identity
shared by a broader democratic community,8 while the UNSC represents a more diffuse group of
countries. This democratic identity thus empowers NATO to send social cues about the
appropriateness of intervention, regardless of the UNSC, which in turn shapes mass opinion.
For evidence, this study analyzes historical polls since the end of the Cold War and a
series of original public opinion experiments that disentangles the effects of NATO and the
UNSC on American policy attitudes, along with the causal mechanisms that explain their
influence. The core of these studies are vignette-based surveys in which American citizens read
about a civil conflict abroad, and then express their support for U.S. military intervention to
save foreign lives. Based on experimental assignment, different survey takers receive different
information about whether the IOs endorse or renounce the use of force.
Validating the theory, the analysis demonstrate that NATO sways mass opinion
regardless of the UNSC’s policy position. These effects hold true even among liberal
internationalists, the politically sophisticated public, and those who are informed about
international law. To undertake the challenging task of assessing causal mechanisms, the
analysis employs three distinct empirical strategies to show that NATO’s influence over public
opinion is indeed due to social identity factors. Together, the evidence supports the argument
8 Deutsch et al. 1957; Risse-Kappen 1995.
that IOs can mobilize popular support for humanitarian intervention because of the collective
identities they represent.
This article brings a fresh perspective to the question of how and why IOs affect mass
attitudes: the domestic power of IOs rests on social foundations, as opposed to legal or material
ones. By making this theoretical contribution, the argument also contributes to the study of
how domestic norms and identities affect international relations.9 Stephen Krasner recently
wrote that “There are only three big causal factors in the study of politics: material interests,
physical power and security, and ideas/identities/norms. We understand the first two pretty
well, but not the third.”10 The theory and evidence thus help meet this shortage.
Making empirical contributions as well, this study employs the first research design to
experimentally measure the independent effect of NATO, meeting the call of many researchers
to explore institutions beyond the UNSC.11 Studying NATO alongside the UNSC not only
sheds light on the two most important security institutions in U.S. foreign policy, but also
provides the empirical variation in IOs that is much-needed for theory testing. Such a study
design additionally helps to overcome inferential issues relating to causal inference and the
lack of consistent and theoretically-relevant data. The conclusion further discusses the paper’s
empirical and theoretical contributions.
This study explains American public opinion on humanitarian war. It joins a growing
scholarship that focuses on the external factors that explain mass attitudes.12 Such literature
demonstrates that policy cues from international institutions influence the opinions of domestic
audiences even after accounting for domestic political cues.13
9 Work on public opinion, norms, and war include Herrmann and Shannon 2001; Gelpi 2002; Press, Sagan,
and Valentino 2013. 10 Krasner 2016. 11 Thompson 2006, 30; Grieco et al. 2011, 571; Fang 2008, 314. 12 Others, in contrast, examine individual-dispositional (Kertzer et al. 2014) or domestic-level (Zaller 1992;
Berinsky 2009) sources of policy preferences. 13 Grieco et al. 2011; Hayes and Guardino 2013; Guardino and Hayes n.d.
On these topics, one prominent research agenda investigates the role of IOs as
“information providers” to citizens.14 These theories observe that policy makers hold a
disproportionate amount of knowledge in international affairs, and thus, mass publics face an
information deficit when evaluating foreign policy. Furthermore, citizens might not be able to
trust the recommendations of their leaders, who could be promoting a policy for parochial
reasons.15 Given this dilemma, people might turn to IOs to obtain a “second opinion” about a
policy’s merits.16 Explained from a different angle, elites who cannot credibly convince their
publics that a policy is worth undertaking may turn to an external source of policy validation.
This literature generally agrees that IOs can persuade citizens to support war, but
scholars debate why this is the case. Various theories provide different explanations for two
fundamental questions. First, what about an IO allows it to sway mass opinion? Second, when
an IO recommends a policy, what is it communicating to the cue-recipient? Legal and
rationalist approaches provide distinct answers to these two questions.
From the legal perspective, an IO is able to sway mass opinion due to its status under
international law. Under international law, the UNSC has broad authority to sanction the use
of force to address threats to international peace and security. According to this doctrine, only
the UNSC can legalize humanitarian war.17 Other IOs like NATO. Americans might be
influenced by legal considerations for several reasons.18 First, UNSC authorization might
signal to citizens that a humanitarian crisis poses a “threat to international peace and security,”
and inaction would allow the crisis to spread across national borders. Second, legal norms
diffuse and are internalized by citizens, especially those who live in a country with a legalistic
culture that respects the rule of law.19 Thus, people might consider an IO-sanctioned policy to
hold “legal legitimacy.”20
14 Fang 2008; Thompson 2009; Chapman 2011; Grieco et al. 2011; Tago and Ikeda 2015. 15 Fang 2008. 16 Grieco et al. 2011 17 This interpretation is supported by treaty law, the words of governments, and legal scholars (Frank 2002). 18 Of course, law does not work exclusively through publics. For example, law could shape policy via
centralized agencies (König and Mäder 2015). 19 Chong 1993; Koh 1997; Finnemore and Sikkink 1998; Goodman and Jinks 2013; Hafner-Burton,
LeVeck, and Victor 2016 20 Tago 2005, 589; Tago and Ikeda, 392.
Citizens could also care about the law for instrumental reasons. Guzman (2008)
outlines “three Rs of [international law] compliance”: reciprocity, reputation, and retaliation.
The U.S. is unlikely to face reciprocity for waging an illegal humanitarian war, but it could
face reputation loss and foreign retaliation. Tomz (2008) shows that American and British
citizens and elites are more supportive of economic sanctions against a law-violating country
than against an otherwise identical country that respects the law, and this “law effect” is
partially channeled through reputational concerns. On retaliation, Voeten (2005, 528) proposes
that “when governments and citizens look for an authority to legitimize the use of force…they
want political reassurance about the consequences,” which includes punishment by powerful
countries.21 Thus, Americans might believe that an unsanctioned intervention could damage
national reputation or provoke other countries to punish the U.S.
Turning to a rationalist perspective, other authors claim that an IO’s ability to influence
citizen attitudes depends on the preference distribution of its membership. A prominent variant
of this argument claims that the most persuasive cues are given by conservative or independent
IOs.22 In particular, a conservative IO is one whose pivotal member is ideologically distant
from the country seeking authorization. Based on this criterion, these theories explicitly
hypothesize that the UNSC should have a greater effect on American attitudes than NATO.23
Similar arguments are made about the role of IOs in convincing foreign publics, arguing that
neutral IOs (i.e. IOs that are comprised of a diverse set of countries) are also more persuasive
when it comes to reassuring skeptical citizens.24
These theories also provide insight into the content of an IO’s cue. Because citizens
are uncertain about a politician’s true motivations for proposing war, they worry that war will
lead to bad outcomes.25 When an IO recommends the use of force, it is letting citizens know
that the policy will have “desirable consequences,” and that a policy will “not be exceedingly
costly or overly aggressive.”26 On the flip side, an IO’s endorsement could also signal the
21 Voeten does not advance a “theory of international law,” but his argument fits with Guzman’s framework. 22 These authors draw from models of cheap talk and models on the role of legislative committees (e.g.
benefits of a policy. For example, a humanitarian intervention will actually save lives, and an
intervention into a civil war will actually generate stability.
The existing literature lends insight into the domestic politics of IO influence, but several
empirical anomalies reveal the need for new theory building and empirical investigation.
Calling into question the legal perspective, in 1999 the United States and NATO allies
conducted armed humanitarian intervention in Kosovo without the approval of the UNSC. But
rather than condemning the U.S., the international community—including some many lawyers
who should be the most-likely political actors to care about legality—declared the illegal
intervention to be nevertheless legitimate (Independent International Commission on Kosovo
2000). Furthermore, public support for Kosovo, as an analysis later in this paper will show,
was at levels similar to cases in which both the UNSC and NATO endorsed intervention (e.g.
Libya). In terms of the rationalist argument, several studies call into question both the idea that
IOs lead citizens rationally update their beliefs about the material costs of war,27 and the
premise that an IO’s conservativeness or membership diversity affects an IO’s ability to
persuade domestic audiences.28 Collectively, these anomalies imply that researchers have yet
to fully explain how IOs affect citizen attitudes.
This section proposes a new theory on the role of IOs in influencing mass foreign policy
preferences. Like the legal and rationalist perspectives, it begins with the premise that domestic
audiences look to international cues when forming their attitudes toward foreign policy, and
that governments can leverage foreign cues to mobilize public support. The main point of
departure, however, is to focus on the role of identity and social context.29 This implies that
people look to the judgments of IOs for cues about the social, rather than the legal or material,
implications of a particular policy. I develop this theoretical intuition through three claims: the
27 Tingley and Tomz 2012. 28 Fey, Jo, and Kenkel 2015; Hainmueller, Mummolo, and Xu 2017, 25-8; Chu 2017. 29 Johnston 2008; Abdelal et al. 2009
democratic community reflects a collective identity shared by its citizens, and policy
endorsements by this community are social cues, and IOs facilitate the sending of these cues.
At this point, the idea that like-minded cue givers are more persuasive might echo theories
of heuristics in political behavior.30 In the context of domestic elite cues, these theories argue that
people look to the policy positions of elites whom share their policy interests as a way to quickly
evaluate whether a policy fits their preferences. The social theory here builds on these insights, but
as the discussion below makes clear, it provides a distinct rationale for why like-minded endorsers
shape mass preferences.
In the context of U.S. humanitarian intervention, what exactly is the relevant social context?
At least since the Cold War, people from liberal democracies like the U.S. have come to
associate with one another. Indeed, a long tradition of research on political communities argues
that the U.S. is embedded in a global community of democracies, especially with regards to
NATO.31 Two features of this community are relevant to understanding the political influence
of IOs.
One facet of the democratic identity—and of any identity—is the norms that define or
constitute it.32 In general, those who are part of an identity-sharing community follow a “logic
of appropriateness” or share a set of beliefs about how they should behave.33 These beliefs
might be codified into formal rules or exist informally as commonly understood expectations
or a “community of practice.” For the liberal democratic community, governments and their
citizens adhere to norms of consultation and norms against coercive bargaining,34 though these
norms do not necessarily apply to interactions with governments outside the club of
democracies.35 Importantly, the norms of consultation highlight the need for collective
30 E.g. Calvert 1985; Lupia and McCubbins 1998 31 Deutsch et al. 1957; Risse-Kappen 1995; Adler and Barnett 1998; Hemmer and Katzenstein 2002 32 Finnemore 1996; Katzenstein 1996b; Adler 1997; Wendt 1999 33 March and Olsen 1998 34 Risse-Kappen 1995; Adler 2008, 204-6 35 Doyle 1986; Maoz and Russett 1993; Tomz and Weeks 2013
deliberation the club of democracies when conducting foreign policy,36 and this process is often
institutionalized within an IO.
Second, people sharing an identity are also defined by their “relational comparisons.”37
This aspect of identity helps to answer the question, what are the characteristics of others that
we do not hold or practice that define who we are? Just as the Cold War experience constructed
a sense of community among democracies, it also demarcated those who did not belong—
countries with authoritarian government, closed economies, and a lack of basic rights.
Together, these features delineate both an in- and out-group in international affairs.38
Even after the Cold War, the democratic identity remains, and it embodies a normative
institution that continues to shape politics. For example, Herrmann and Shannon (2001, 642)
show that in-group identity among advanced democracies explain people’s support for war,
and Risse-Kappen (1996) finds that the existence of a democratic community explains why
NATO persists even though structural conditions predicted its collapse. Reflecting on the
enduring nature of political identities, Peter Katzenstein observes that political actors “attribute
far deeper meanings to the historical battles that define collective identities than to the transient
conflicts of daily politics.”39
How does the theoretical discussion thus far shed light on the way in which the UNSC
and NATO shape American attitudes toward humanitarian intervention? It implies that
Americans considering whether or not to support humanitarian intervention will value, first
and foremost, endorsements from within their social group and thus of other liberal democratic
36 This literature already explores the historical formation of the democratic identity, so for space I focus on
the implications of this identity rather than its origins. 37 Abdelal et al. 2009, 20-4 38 This discussion is compatible with psychological approaches to group identification (e.g. Mercer 1995;
McDermott 2009). 39 Katzenstein 1996a, 3. Whether U.S. President Donald Trump’s rhetoric has fundamentally shifted
American identity in world politics (for some people) remains an open question. It is worth noting that after
Trump assailed NATO during his campaign, the U.S. congress reacted by bringing forth a bipartisan
resolution to affirm the U.S.’s commitment to NATO (Accessed 5/31/16 at:
http://foreignpolicy.com/2016/05/18/exclusive-in-rebuke-of-trump-house-resolution-defends-nato/). A
PEW study also found that “[w]hile Trump recently called into question the value of U.S. participation in
NATO, Americans overwhelmingly view NATO membership as beneficial for the United States…Large
majorities in both parties say NATO membership is good for the U.S.” (Accessed 6/14/16 at:
http://www.people-press.org/2016/05/05/public-uncertain-divided-over-americas-place-in-the-world/). In
any case, the theory does not clam that identities last forever, and the conclusion revisits the idea of identity
change.
countries. These in-group cues are most prevalently captured by the policy endorsements of
NATO, which is emblematic of the liberal democratic community.40 On the other hand, after
the core community is consulted, cues from out-group countries or IOs with more diffuse
representation like the UNSC do not substantially shape popular opinions. Once the democratic
community-via-NATO has taken a stance on intervention, Americans are unlikely to place
much weight on additional judgments from the UNSC.
The process of socialization sheds light on the mechanisms through which political
communities affect people’s policy preferences.41 In particular, I argue that policy judgments
by IOs send social messages, cuing those with a particular identity as to whether adopting a
particular policy is appropriate or will produce positive social (in contrast to material)
outcomes.42 What does it mean, though, to send a cue about a policy’s social appropriateness
and implications?
Drawing from socio-psychology and political science, Johnston (2001) outlines two
microprocesses that are a part of the socialization of individuals: persuasion and social
influence. Persuasion “involves changing minds, opinions, and attitudes about causality and
affect (identity) in the absence of overtly material or mental coercion,” which includes the
changing of attitudes via cues or endorsements (496-7). Meanwhile, social influence “refers to
a class of microprocesses that elicit [intersubjectively understood] pro-norm behavior through
the distribution of social rewards and punishments,” where the rewards and punishment affect
one’s status or image rather than material outcomes (499-501).43 As a result of persuasion and
social influence, community members are convinced or feel pressured to adopt a certain policy,
participate in group activities, and be seen as engaging in pro-norm behavior.
40 NATO, of course, does not perfectly represent the liberal democratic community. It largely represents (and
certainly more so than the UNSC) the club of democracies. 41 Johnston (2001, 494) defines socialization as “the process by which social interaction leads novices to
endorse ‘expected ways of thinking, feeling, and acting.’” While Johnston focuses on the socialization of
“novices,” I treat socialization as occurring continuously even after members have adopted a shared identity. 42 Existing research finds that IOs are central to facilitating socialization (e.g. Finnemore 1993; Li 2010). 43 Johnston (2008, 80) notes that persuasion involves changing another’s beliefs about what is right, while
social influence is a form of peer pressure that changes behavior without changing internally held beliefs.
But this paper does not attempt to identify the particular proportion of each process that constitutes the ability
of IOs to shape public opinion.
This discussion implies that when NATO recommends a particular foreign policy (e.g.
humanitarian intervention), it engages in persuasion and social influence for those holding the
liberal democratic identity. More concretely, cues from NATO answer questions such as, how
will engaging in this policy affect my country’s status or reputation?44 Will others in my group
collectively act to advance this policy? Does pursuing this policy conform to pro-norm
behavior for countries like mine? In the case of humanitarian war, pro-norm behavior includes
championing liberal values in international affairs, which encompasses stopping human rights
violations. These factors, in turn, elicit support for intervention.
The ability of IOs to influence domestic audiences is rooted in the collective identities and
community of countries they represent. What political purpose, then, does an IO serve beyond
their member states? Firstly, IOs are fora that help disparate community members coordinate
and centralize their decision making,45 which implies that individual governments may fail to
reach a group decision without a formal organization. Additionally, IOs may increase the
accessibility of a foreign cue. Americans might be more likely to hear the policy platform of a
country via an IO than that of the same country announcing its policy position on its own
outside of an IO. From this point of view, IOs do not add anything substantive to a groups’
policy message. Instead, the IO’s primary role is to make sure a community of countries
successfully comes together to make the policy recommendation in the first place and that the
recommendation is accessible to domestic audiences.
IOs, however, do more than facilitate coordination. This paper argues that social cues
reveal the implications of a foreign policy on the likeliness of collective action, reputation and
status, and beliefs about pro-norm behavior: IOs can amplify each of these social mechanisms.
An IO facilitate monitoring and enforcement, which could increase the likeliness of collective
action to fulfil a policy.46 Enforcement in the case of NATO takes the form of social
sanctioning, implying that IOs raise the reputational stakes of not going along with a group
44 Because this study does not attempt disentangle the instrumental and non-instrumental reasons for valuing
status (Johnston 2008, 82-84; Renshon 2016), it uses the more colloquial definition of “reputation” to
encompass all forms of status. 45 Keohane 1984; Abbott and Snidal 1998. 46 Martin 1992.
activity like humanitarian intervention.47 Lastly, when individuals make a policy endorsement,
it may not be clear whether they are doing so because of their identity or because of other
unrelated interests. If an IO symbolizes a particular identity (such as in the way NATO
symbolizes a community of liberal democratic countries), then recommending a policy via an
IO could clarify the social meaning of that recommendation.
This discussion implies that members of the democratic community can sway the
opinion of the American public, but they are more influential when they voice their views via
NATO. Contrast this with the alternative theories. The international law argument highlights
the legal authority held by the IO itself, rather than the IO’s members. Thus, a cue from a group
of countries will have little influence if it is not sent through an IO. Meanwhile, information
theories argue that the strength of an informational cue is a function of the distribution of
preferences held by an IO’s component countries, which implies that the informational content
of a cue from a group of a countries versus the same group of countries acting through an IO
is about the same.
This study evaluates three principal claims derived from the community identity theory: (1)
NATO’s policy position sways mass support for humanitarian intervention, independently of
the UNSC’s stance, (2) NATO affects public opinion via causal mechanisms relating to the
social repercussions and appropriateness of intervention, and (3) liberal democratic countries
affect American public opinion, but their influence is magnified when sent via NATO.
The main evidence comes from four survey experiments, but before turning to the experimental
design, it is useful to assess these claims in light of observational data. While relying on
they motivate the experimental design and buttress the external validity of the larger study.48
Table 1 allows us to assess a few descriptive patterns in the historical relationship
47 Johnston 2008 48 Driscoll and Maliniak (2016) gives an example of validating experiments with observational data.
between the UNSC, NATO, NATO’s countries, and support for humanitarian war.49 It includes
cases in which the U.S. considered humanitarian intervention and opinion polls were available.
Column 1 names the case. Columns 2 through 4 gives the policy position of the NATO’s
countries, NATO and the UNSC. NATO’s policy position on intervention is generally
equivalent to policy position of its member countries. But when NATO did not consider the
case of intervention (i.e. NATO = N/A), the country positions is coded using public statements
and actions of NATO’s member states. Column 5 then gives the percentage of respondents that
supported intervention (unreported are those who opposed war or responded “don’t know” or
“no opinion”). The right most column gives the average support for intervention at different
degrees of international policy endorsement. The data are from the Roper Center database.
Detailed information on the polls along with coding notes are in Appendix C.
The record reveals three broad classes of humanitarian interventions. Interventions
with little international support (e.g. Syria), ones in which NATO and the club of democracies
support intervention but the broader international community vis-à-vis UNSC does not
(Kosovo), and interventions with widespread international backing (Bosnia). Public opinion in
these three groups suggest that Americans prefer war with international backing over wars
without it. Only 28 to 31 percent of Americans supported intervention without foreign
approval, while a majority of Americans supported interventions with some degree of IO
approval. Interestingly, Americans do not reveal a preference for interventions with both
UNSC and NATO approval over interventions with only NATO approval.50 These patterns are
consistent with the identity theory’s prediction that once a cue from the in-group is received
(i.e. NATO), the additional cue from other countries has little effect.
49 The polls specifically ask for people’s support for military intervention for humanitarian reasons. They do
not include commonly asked but less relevant questions like, “Do you agree with how President Clinton is
handling the situation in Somalia?” 50 Libya might be classified as a “NATO Only” intervention because the UNSC’s resolution arguably did not
cover the NATO airstrikes against Qaddafi.
Table 1: Past National Polls on American Support for Intervention
Cases Support Intervention? % Supporting War
NATO Countries NATO UNSC By Case Average
Rwanda 1994
Syria 2013
No (minus France)
No
N/A
No
No
No
28
33 31
Kosovo 1999
(Libya 2011)
Yes
(Yes)
Yes
(Yes)
No
(No)
53
(56)
53
(55)
Somalia 1992
Haiti 1994
Bosnia 1994
Libya 2011
Yes (Strong)
Yes (Weak)
Yes
Yes
N/A
N/A
Yes
Yes
Yes
Yes
Yes
Yes
74
34
57
56
55
Note: This table summarizes data from historical surveys conducted during episodes of potential
humanitarian intervention. NATO is coded N/A in when the case was considered “out of area” at the
time. The Rwanda poll was taken prior to France’s Operation Turquoise, which received UNSC
approval. The Libya case might be classified under “NATO Only” because the UNSC resolution
arguably did not cover the NATO airstrikes.
These polls provide initial evidence for the identity theory, but they nevertheless face
major shortcomings and speak to the need for an alternative research design. First, the historical
data cannot account for the problem of selection bias. It is likely that U.S. decisions to seek IO
approval are a function of domestic politics, which could frustrate inference about the effect of
IOs on public opinion. Second, there are many factors that explain both the policy positions of
IOs and public opinion, and these “omitted variables” confound the observed relationship
between IOs and public opinion. For example, both IOs and citizens could form their opinions
based on the merits of intervention rather than the policy positions of each other. Third, many
of the cases have idiosyncratic features. Somalia, for example, was the U.S.’s first
contemporary case of humanitarian intervention just after the popular First Gulf War. These
factors, rather than the fact that there was virtually unanimous international consensus on
intervention, might explain the high levels of public enthusiasm for intervention in Somalia.
Finally, historical polls generally lack information useful for analyzing causal mechanisms,
which is essential for evaluating and building theory.
To overcome the problems arising from solely analyzing historical polls, this study deploys a
novel experimental design. Conducting an experiment that randomly and independently
assigns the policy cues of IOs overcomes the problem of selection and causal inference. And
designing original survey questions creates an opportunity to measure key variables regarding
the causal mechanisms that explain how IOs persuade citizens.
Specifically, this study conducted a series of experiments inserted into four public
opinion polls fielded in the United States. Surveys #1 through #3 help test the claims about the
effect of IOs and the causal mechanisms through which policy cues influence Americans, while
Survey #4 evaluates the effect of cues sent by the democratic community with and without an
NATO. I describe the first three surveys here and revisit Survey #4, which has a slightly
different design, in a later section. The surveys were administered online from January 2015
to April 2016. Survey #1 was conducted via YouGov to a nationally representative sample of
Americans, while Surveys #2 and #3 were given to diverse but convenience samples via
Amazon Mechanical Turk (MTurk).51 Appendix A gives the respondent attributes.
In the surveys, respondents read about a humanitarian crisis in which “Military groups
fighting in [country] have killed thousands of civilians, including women and children, and
have left tens of thousands homeless and starving.” In Survey #1, the target “[country]” was
randomly selected from the following five potential target countries: Azerbaijan, Burma,
(Myanmar), Chad, Colombia, and Yemen. In Surveys #2 and #3, the country was either
Azerbaijan or Chad. These countries have a history of civil conflict, which helps to deepen the
study’s realism and political relevance. The five countries also represent different regions so
that the result could inform a range of cases. In the statistical analyses, data across these
different potential country scenarios are pooled together, but results conditional on the target
country are given in Appendix D.52
Next, respondents read about the UNSC and NATO’s stance on humanitarian war. In
Survey #1, respondents were randomly assigned to read one of three potential scenarios that
mirror the historical record presented in Table 1: both IOs oppose intervention (None), only
NATO supports intervention (NATO Only), and both IOs endorse intervention (Both). In
Survey #2, respondents read all three scenarios plus a fourth scenario in which only the UNSC
but not NATO supports intervention (UNSC Only). The fourth scenario is unprecedented
historically, but this “off-equilibrium” case is useful for evaluating theory, which is discussed
fully in the analysis. In Survey #3, respondents were assigned to read either the None or NATO
Only scenarios, and several additional experiments were embedded into this survey to better
understand the how people process information about international law and material outcomes.
To increase realism and external validity, respondents also received information about the key
countries belonging to each IO.53 But note that Survey #4 will disentangle the effect of naming
countries versus naming the IOs.
Finally, respondents in each survey expressed their attitudes toward the use of force by
answering the following: “In this situation, do you support or oppose the US sending its
51 Research shows that MTurk samples produce findings that are comparable to more representative samples
(Berinsky, Huber, and Lenz 2012). 52 The country is independently assigned and thus does not confound estimates of the effect of IOs.
Randomizing the country, however, adds variance to the dependent variable, which actually makes detecting
statistically significant effects less likely. 53 News articles on the UNSC and NATO generally mention key countries from the IOs. For example, when
the UNSC authorizes the use of force, news often report the voting record of the P5 in the article lede.
military to help civilians in [country]?” There were six replies: support or oppose a little, a
moderate amount, or a great deal. People’s answers to this question formed the study’s main
dependent variable, Support for Intervention. Appendix B contains the full questionnaire.
Note that the forthcoming analysis compares levels of the dependent variable across
the treatment groups, often using the None condition as a baseline group. The experimental
design did not implement a group in which respondents receive no information about either
IO. Respondents in such a “no information” condition are free to assume, either consciously or
subconsciously, whatever they wish about the main explanatory variable, IO Support. This is
highly problematic. IOs have approved of virtually all contemporary humanitarian
interventions, and both NATO and the UNSC recommended the recent intervention in Libya.
Given these facts, many respondents are likely to assume a degree of IO endorsement, either
because people use analogies to reason through hypothetical scenarios or because people
subconsciously associate the concept of humanitarian intervention with NATO and the UNSC.
This could generate respondent non-compliance, driven by respondents who, in a sense, “treat”
themselves with the IO endorsement.54 Thus, the surveys explicitly informed respondents the
positions of each IO to avoid this design flaw.55
Data from Survey #1 allow us to estimate the main effect of IOs on mass opinion. For clarity,
I analyze the dependent variable, Support, in its binary form (i.e. either support or oppose the
use of force), but the results are robust to using the full 6-point scale. Table 2 shows that moving
from a war that NATO and the UNSC oppose (None) to one that is endorsed by NATO but not
the UNSC (NATO Only) heightens Support by 26.4 percentage points. The 95 percent
confidence interval indicate that this increase is statistically distinguishable from zero at the
54 Chu 2016 55 In contrast, see Johns and Davies (2014). This study employs a rich survey of British opinion on war, but
for its experiment on IOs, informs respondents that either NATO or the UN supports war. However, hearing
“UN supports war” does not imply that NATO opposes wars, and vice versa. Thus, the fact that the UN and
NATO have recently supported policies in tandem could explain why the study finds similar levels of support
between these experimental conditions.
0.05 level. By contrast, the additional impact of the UNSC’s endorsement (moving from NATO
Only to Both) is both substantively small (4.3 points) and statistically insignificant.
These results largely mirror the historical data presented in Table 1, speaking to the
external validity of the findings. Further bolstering the validity of these findings, a joint
endorsement by NATO and the UNSC raises support for war by about 30 points, which
replicates the results from a previously published nationally representative survey conducted
via telephone.56 Finally, these results are robust to limiting the sample of respondents to groups
that are particularly influential in American foreign policy: the sophisticated public and liberal
internationalists (see Appendix D).57 This suggests that the effect of IOs is not driven by policy
ignorance or people who do not engage with international affairs.
56 Grieco et al. 2011, table 3 57Wittkopf (1990, Ch. 5) finds that elites are more internationalist than the general population. Milner and
Tingley (2016) find that liberal internationalists form a coalition that drives U.S. foreign policy.
Table 2: NATO Increases Support for Intervention but
a Security Council Endorsement Adds Little More
Treatment
Group
% supporting
intervention
(Support)
Difference in %
supporting
intervention
None
N=354
48.3
[43.0, 53.6]
26.4
[19.4, 33.4]
4.3
[-2.2, 10.8]
NATO Only
N=332
74.7
[69.6, 79.3]
Both
N=314
79.0
[74.1, 83.4]
Note: This table gives the percent of respondents supporting
intervention for each treatment group, and the difference in
support between groups. “None” indicates an endorsement from
neither NATO nor the UNSC. “NATO” is an endorsement from
only NATO. “Both” is an endorsement from both IOs. 95%
confidence intervals are reported in brackets. N=1,000. Data are
from Survey #1.
Next, Survey #2 allows us to rule out an alternative interpretation of the main results,
that Survey #1’s results reveal a first-versus-second cue giver effect rather than a NATO-
versus-UNSC effect. Put another way, there could be diminishing returns to policy cues, and
so any second cue (not just the UNSC’s) will have little influence on public opinion. In Survey
#2, respondents are assigned to one of four experimental groups: the original three scenarios
(None, NATO Only, and Both), and a fourth scenario in which only the UNSC endorses
intervention (UNSC Only). Figure 1 presents the results and reveals two key patterns. First, the
original three treatment groups replicate the results from Survey #1. The effect of moving from
None to NATO is large, while the additional endorsement from the UNSC is insubstantial.
Second, and in contrast, the UNSC alone only slightly raises enthusiasm for war (from None
to UNSC), but NATO’s additional recommendation still substantially boosts intervention
support (from UNSC to Both). Finally, the data allow us to calculate the effect of each IO
averaging across the other IO’s policy positions. The effect of NATO is 15.7 points and the
effect of UNSC is 6 points, and the difference is statistically significant.
Figure 1: NATO Has a Greater Effect on Public Attitudes
than the UN Security Council
Note: This figure shows the proportion of respondents supporting
intervention for four experimental group. “None” indicates an
endorsement from neither NATO nor the UNSC. “NATO” is an
endorsement from only NATO. “UNSC” is an endorsement from only the
UNSC. “Both” is an endorsement from both IOs. 95% confidence
intervals are given. N=408. Data are from Survey #2
Analysis reported in Appendix D shows that NATO’s influence on public opinion
remains even under conditions designed to confirm the legal and information theories’
hypotheses. Beginning with the legal perspective, respondents’ knowledge about international
law should presumably suppress the effect of NATO while amplifying the effect of the UNSC,
but this is not the case (Figure A3). A follow-up experiment further shows that NATO’s
influence remains even after respondents are explicitly informed that the UNSC is necessary
for intervention to be legal (Figure A4). Lastly, the information theories argue that the more
conservative an IO, the greater effect that IO should have on public opinion, where
conservativeness is measured using the IO’s pivotal member’s ideological distance from the
cue-receiver. However, Table A5 show that the effect of NATO and UNSC are not conditional
on respondents’ perceptions about each IO’s pivotal member.
The investigation thus far shows that NATO affects mass support for humanitarian intervention
regardless of the UNSC’s position. It now evaluates additional implications of the identity
theory by assessing the causal mechanisms that explain how NATO affects public opinion. The
identity theory argues that, because American identify with NATO’s member-countries,
NATO raises support for intervention by revealing the social implications of deploying military
force, which include national status and reputation, collective action, and perceptions of norm
conformity. These mechanisms contrast with factors relating to international law, information
transmission, and other material or burden sharing variables.58 To be clear, these mechanisms
are not mutually exclusive, but they are useful to distinguish in the analysis.
I employ three empirical strategies to evaluate these mechanisms. First, I examine an
assumption about the subgroups among which the mechanisms should be most relevant:
Americans who identify with NATO’s members. Second, I conduct mediation analysis to
directly estimate the relevance of various mechanisms.59 Third, I indirectly assess the
importance of the social mechanisms by ruling out alternative causal pathways.
58 But note that Recchia (2015) argues that the burden sharing mechanism is primarily about convincing the
military elite to support intervention. 59 Imai et al. 2011
The identity theory implies that NATO’s effect is largest among Americans who mostly closely
identify with NATO’s members and thus care most about the social cues that NATO delivers.
Using data from Survey #2, I test this hypothesis by analyzing NATO’s average endorsement
effect conditional on country affinity, which I use as a measure of shared identity. To measure
affinity, respondents reported their opinions about whether the following four countries “share
a friendly, neutral, or hostile relationship with the U.S.”: Canada, Germany, France, and the
United Kingdom.60 I split the sample in half based on people’s average affinity of these four
countries, creating a dummy variable Affinity, which takes the value 0 among those who feel
relatively distant and a value of 1 among those who feel relatively close to these countries.
To analyze whether Affinity moderates NATO’s endorsement effect, I use a probit
model to regress Support on NATO, Affinity, and their interaction, along with a set of
demographic and political control variables. Figure 2 reports the results (control variable
estimates are in Appendix E). Starting at the top, it shows that NATO’s effect on Support
among the regression’s baseline category, people who express relatively low affinity toward
NATO’s countries, is about 5.4 percentage points (i.e. a 0.054 change in the probability of
supporting intervention), though this effect is not statistically significant. A NATO cue absence
of affinity has a trivial effect. The interaction term, however, indicates that the effect of NATO
is significantly larger among people who express relatively high affinity (19 points larger).
60 I also asked about China and Russia. Together, this encompasses two countries that are only in NATO,
only in the UNSC, and in both IOs. Data from all six countries are used in supplementary analysis in
Appendix D.
Figure 2: NATO’s Effect is Strongest among Americans
who Identify with NATO’s Members
Note: This figure reports the marginal predicted probabilities from
probit coefficients. The dependent variable takes a value of 0 if the
respondent opposes intervention, and a value of 1 if the respondent
supports intervention. The baseline category is the condition in which
NATO opposes intervention and respondents express low affinity toward
NATO’s members. Control variables (gender, age, education, party ID,
and ideology) are reported in Appendix E. 95% confidence intervals are
given. N=408. Data are from Survey #2
.054
-.0009
.19
NATO
Affinity
NATO*Affinity
-.4 -.3 -.2 -.1 0 .1 .2Change in Probability of Supporting Intervention
Next, I analyze six causal mechanisms that might explain NATO’s endorsement effect.
Reputation measures perceptions about whether intervention would improve or damage the
U.S.’s reputation.61 Collective Action indicates respondents’ beliefs about whether other
countries will contribute to military action. Humanitarian Motives captures beliefs about
whether intervention conforms to pro-norm behavior: that is, whether the U.S. is actually using
force for humanitarian reasons. Prevent Contagion indicates expectations about whether not
intervening would lead a spread of conflict. Retaliation Unlikely records beliefs about whether
other countries would punish the U.S. for using military force. Casualties Unlikely measures
expectations about American war casualties.
Recall that the identity theory implies that all three social considerations, Collective
Action, Reputation/Status, and Humanitarian Motives (pro-norm behavior), are significant
mediators. It also argues that fears about foreign retaliation should not vary due to democratic
norms against coercive diplomacy. The other mechanisms relating to Contagion, Retaliation,
and Casualties are not predicted by the theory but are included in the analysis for a point of
comparison.
With these hypotheses in mind, the causal mediation analysis proceeds in three steps:
estimate the effect of NATO Only on each mediator; estimate the effect of each mediator on
the dependent variable, Support for Intervention; and finally using information from the first
two steps, estimate the average mediation effects.
Figure 3 displays the results from steps one and two. For each mediator, the left panel
shows the effect of moving from None to NATO Only. The estimated probit coefficients are
reported in terms of predicted probabilities. When NATO recommends intervention, people
are more likely to believe that intervention benefits U.S. Reputation, leads to Collective Action,
and is for Humanitarian reasons. Next, NATO’s policy judgements make people believe that
war can Prevent Contagion, and that Retaliation and Casualties are Unlikely; however, the
effect of NATO on Retaliations Unlikely and Casualties Unlikely is small and statistically
insignificant. The right panel illustrates the effect of each mediator on Support. The mediators,
unlike the IO cues treatment, are not experimentally varied, so I include control variables (not
61 These six variables are measured using data from Survey #1 and further described in Appendix E.
reported in the figure) to guard against confounding from omitted variables.62 All of the
mechanisms are positively associated with popular support for the use of force, but to different
degrees. Reputation has the largest effect, while concerns about Retaliation have the smallest
effect.
Figure 3: Potential Explanations for how NATO Shapes Public Opinion
Note: This figure gives the predicted marginal effects from probit coefficient estimates. The left panel
plots the marginal NATO effect on six mediators, and right panel plots the effect of those mediators on
people’s intervention support, controlling for potential confounds. Estimates from the control variables
are not displayed. Each of the 12 estimates is obtained from a separate regression. 95% confidence
intervals are given. Data are from Survey #1.
62 The control variables are gender, age, education, income, race, religion, ideology, party identification,
voter registration, internationalism, beliefs about human rights, and news interest.
Combining these two steps provides the basis for estimate each mechanism’s mediation
effect. Specifically, the analysis executes the procedures outlined in Imai et al. (2011) using
statistical software programmed by Hicks and Tingley (2011).63 Table 3 gives the average
mediation effect for each mechanism. Recall that NATO raises Support by 26.4 percentage
points. The analysis shows that 6.4 of those points are mediated by concerns about Reputation.
In other words, some of NATO’s effect on public opinion is channeled through people’s beliefs
about how intervention affects U.S. reputation or status. The results also demonstrate that
concerns about Collective Action, Humanitarian (Motives), and Contagion are significant
causal pathways, but beliefs about Retaliation and Casualties are not.
Consistent with the identity theory, the analysis shows that people’s social concerns
about Reputation, Collective Action, and Humanitarian Motives are important causal links. The
policy position of NATO sends a social cue to Americans. Unpredicted by the theory, concerns
about Contagion also explain the effect of NATO. Finally, concerns about Retaliation and
Casualties do not mediate the effect of NATO on attitudes.64 The null finding for Retaliation
finding is predicted by the identity theory. In sum, the mediation analysis rules out the
relevance of the casualties and retaliation mechanisms, and shows that social appropriateness
mechanisms largely explain how NATO sways public opinion.65
63 Appendix E gives a technical overview of this analysis. 64 The Casualties finding might seem counterintuitive. But to clarify, it does not contradict the conventional
wisdom that casualties affect public opinion (e.g. Mueller 1973; Gartner 2008). Figure 3 shows that
Casualties do indeed affect Support, but NATO endorsements do not change beliefs about Casualties. 65 Note that the decreasing mediation effects from Reputation to Casualties is unlikely to be a result of
question ordering. These questions appeared in the following order: Reputation, Casualties, Retaliation,
Humanitarian, Collective Action, and Contagion. They are rearranged in this paper to be consistent with the
theoretical grouping.
Table 3: Social Mechanisms Explain
NATO’s Endorsement Effect
Mediator
Average
mediation
effect
Reputation/Status 6.4*
Collective Action 3.9*
Humanitarian 1.5*
Prevent Contagion 3.2*
Retaliation Unlikely 0.4
Casualties Unlikely 1.0
Note: This table shows the mediation effects
of six variables that might explain how NATO
raises support for intervention. Asterisks
highlight estimates that are significant at the
0.05 level. Data are from Survey #1.
The final analysis of causal mechanisms demonstrates that NATO’s endorsement effect
remains even after accounting for material or burden sharing factors, such as the number of
lives that would be saved, the financial and human costs of military action, and the mode of
intervention. To rule out these alternative factors, Survey #3 included the original experiment
in which NATO supports or opposes military action (NATO), while additionally randomizing
whether respondents received or did not receive information about the material cost and
benefits of military action (Information).66 If NATO boosts support for war via providing
information about the material consequences of military action, then directly providing that
information to respondents should dampen the effect of NATO. In other words, the interaction
of NATO and Information should have a negative effect on people’s support for war.
The analysis applies a probit model to regress Support on NATO, Information, and their
interaction (NATO*Information). Figure 4 displays the results of this analysis in terms of
predicted marginal probabilities. The estimate of NATO shows that, among those who are not
given information about material cost and benefits, a NATO endorsement increases Support
by 29 percentage points (or 0.29 increase in probability of supporting war). The negative
coefficent of NATO*Information indicates that informing respondents about material cost and
benefits reduces the NATO effect, but only by 4.4 points. This effect is also statistically
insignificant. Put another way, 85 percent of NATO’s effect on Support (i.e. (29-4.4) /29)
remains even among those who are informed about the number of lives at stake, financial costs
to the U.S., and mode of intervention. Thus, 85% of NATO’s effect must be due to other
mechanisms such as those relating to social considerations.
66 Those who received information read the following: “Military action would save the lives of about [80
thousand OR 620 thousand] civilians. The operation would cost the U.S. government about [$850 million
OR $4.1 billion], but the U.S. would avoid risking casualties by not sending ground troops.”
Figure 4: NATO’s Endorsement Effect is Not About Material Factors
Note: This figure reports the effect of NATO and information about the material
consequences of intervention on people’s support for war. The estimates are from a probit
regression and given in terms of predicted marginal probabilities. The outcome takes a
value of 0 if the respondent opposes intervention, and a value of 1 if the respondent
supports intervention. The baseline category is the condition in which NATO opposes
intervention and respondents receive no information about material consequences.
N=1,141. Data are from Survey #4.
.29
.036
-.044
NATO
Information
NATO*Information
-.2 0 .2 .4Change in Probability of Supporting Intervention
The community identity theory argues that the ability of IOs to shape public opinion comes
from their member countries, but those countries are more persuasive when operating through
an IO. Survey #4 helps to test this argument. As in the previous surveys, some respondents
were experimentally assigned to read that “The North Atlantic Treaty Organization (NATO)
[opposes/supports] taking military action to help these civilians. NATO members include the
U.S., Canada, and several European countries.” But to distinguish the effect of NATO from
the effect of its named countries, other respondents read that NATO supports/opposes military
action without mention of any particular country names, while others read that “U.S., Canada,
and several European countries” support/oppose military action without mention of NATO.
This creates six experimental groups (Both Countries and IOs, IOs Only, or Countries Only
either support or oppose military action). Because Surveys #1 through #3 establish that
NATO’s endorsement effect is independent of the UNSC, this version did not mention and
vary the policy position of the UNSC, which helps to increase statistical power for the NATO
experiment. As before, the dependent variable is people’s support for armed intervention.
Appendix A and B provide the sample attributes and survey questionnaire.
Table 4 presents difference in proportion of war support for each pair of experimental
treatments (e.g. difference in support when Countries Only supports intervention from when
Countries Only opposes intervention). As the identity theory predicts, the countries of the
democratic community alone can sway American attitudes (Countries increases support for
intervention by 15.5 points), but those same countries acting via NATO have an even greater
influence (Both has a 32.8 point effect). The effect of IO Only is included for analytical
completion, but is actually difficult to interpret. Even though it does not mention NATO’s
member countries, respondents can easily infer that several democratic countries support
intervention simply from hearing “NATO.” What this data does help to clarify, though, is
whether the effect of the policy cue relies primarily on naming countries. The results show they
do not: IO Only has a 24.8 point effect on intervention support.
Table 4: The Role of IOs versus their Countries
The endorsement effect of… Change in War Support in
Percentage Points [95% C.I.]
Countries Only 15.5 [3.4, 27.6]
IO Only 24.8 [12.3, 37.2]
Both Countries and IO 32.8 [20.0, 45.6]
Note: This table shows the effect of policy endorsements on people’s
support for intervention. N=598. Data are from Survey #4.
A vast majority of humanitarian wars are conducted with the backing of an IO. To explain this
vexing empirical regularity, scholars argue that governments value the blessing of IOs for their
ability to mobilize public support. This conventional wisdom explains the link between IOs
and public opinion by focusing on the ability of IOs to fulfill legal requirements or relay
information about a policy’s merits. This paper demonstrates that these approaches overlook
the social and ideational role of IOs. In the case of U.S. humanitarian intervention, liberal
democratic countries share a collective identity that allows NATO, independently of the
UNSC, to communicate the social appropriateness of the use of force and thus convince
Americans to support war.
This article presents evidence from historical polls and a unique set of experimental
studies that disentangles the UNSC and NATO’s effects on public opinion, and identifies the
causal pathways through which IOs influence domestic audiences. The data demonstrate that
policy recommendations by NATO rally American citizens for war, and the mechanisms that
explain this effect are people’s social affinities and considerations about reputation, collective
action, and pro-norm behavior. The analyses also present a variety of empirical tests to account
for alternative explanations.
Together, the contributions of this paper are both theoretical and empirical. The study
provides new insights on the domestic politics of IO influence, and it does so by developing
the first theory that connects social identities, mass opinion, and international institutions. The
findings also provide a new answer to the age old question of why do international institutions
appear legitimate. Existing theories explain institutional legitimacy in legal and rationalist
terms, while this paper finds that legitimacy has social foundations. Turning to the empirical
contributions, this study conducts and analyzes the first experiment to evaluate the independent
effect of an IO other than the UNSC. By examining NATO, the evidence sheds light on one of
the most important institutions in American foreign policy, and for the sake of theory testing,
it provides a point of comparison to the Security Council. The evidence also takes a multi-
faceted approach to assess the causal mechanisms that link IOs to public opinion. Lastly, by
showing that the experimental results are borne out in the historical data, the evidence can
address concerns about both internal and external validity.
The findings also sheds light on the nature of forum shopping in international relations.
In an insightful paper, Voeten (2001) observes that the UNSC since 1990 has rarely vetoed
resolutions regarding the use of force despite the existence of members like Russia and China
whose interests are generally against humanitarian wars. If the UNSC is so necessary, then
why do those Security Council members not simply exercise their veto? Through game
theoretic models, Voeten argues that the existence of a credible outside option leads Security
Council members to make policy concessions and refrain from exercising a veto. In the case
of humanitarian intervention, the most salient outside option is NATO. Why might NATO be
a credible outside option? And is there evidence for such an outside option? Addressing these
unanswered questions, this paper shows that NATO is a credible outside option to the UNSC
because it can fulfill the function of mobilizing domestic support for intervention, filling in a
long-neglected empirical gap in the literature on forum shopping in international security.67
This project has two natural extensions: assessing non-humanitarian wars and non-U.S.
public opinion. Regarding non-humanitarian wars, national security concerns could moderate
the effect of IOs. After all, the identity theory does not claim that social considerations override
all else. If security matters, then one hypothesis could be, the more the justifications for war
concern core national security interests, the smaller the effect of IOs. For example, in wars of
self-defense IOs will probably have little effect on enthusiasm for military action.68 Second,
the identity theory could generalize beyond the American context. Given the U.S.’s superpower
status and culture of exceptionalism among some Americans, the social effect of IOs would
presumably be even stronger when it comes to the democratic citizens of smaller countries.
Additionally, citizens of autocracies like Iran or China are unlikely to exhibit the same patterns
in their opinions when it comes to the endorsement of NATO. It would be useful for future
research test the theory in other countries to see if this is the case.
In terms of U.S. foreign policy, policymakers seeking to raise support for humanitarian
intervention should obtain multilateral support, though the form of multilateralism should first
and foremost contain the U.S.’s fellow liberal democracies. Once that support is gained,
obtaining UNSC authorization might be important for some people, but not the broader
67 Also see Lipscy (2017) for applications beyond the security realm. 68 However, Herrmann and Shannon (2001, 641) show that normative factors affect American elite support
for armed intervention even when material interests are threatened.
American public. The normative implications of these findings are for a different study, but
this paper’s findings imply that one way to influence global constraints on U.S. foreign
policymaking is via socialization. While Americans have held a stable liberal democratic
identity for over six decades, that identity is not timeless. Fostering a cosmopolitan American
identity might increase the sway that foreign political actors have on U.S. public opinion, while
promoting an identity of exceptionalism could have the opposite effect.
Abbott, Keneth W., and Duncan Snidal. 1998. “Why States Act through Formal International
Organizations.” Journal of Conflict Resolution 42 (3): 3–32.
Abdelal, Rawi, Yoshiko M. Herrera, Alastair Iain Johnston, and Rose McDermott. 2009.
Measuring Identity: A Guide for Social Scientists. New York, NY: Cambridge University
Press.
Adler, Emanuel. 1997. “Seizing the Middle Ground: Constructivism in World Politics.” European
Journal of International Relations 3 (3): 319-63.
Adler, Emanuel. 2008. “The Spread of Security Communities: Communities of Practice, Self-
Restraint, and NATO’s Post-Cold War Transformation.” European Journal of Political
Science 14 (2): 195-230.
Adler, Emanuel and Michael Barnett (eds.). 1998. Security Communities. New York, NY:
Cambridge University Press.
Aldrich, John H., John L. Sullivan, and Eugene Borgida. 1989. “Foreign Affairs and Issue Voting.”
American Political Science Review 83 (1): 123–141.
Aldrich, John H., Christopher Gelpi, Peter Feaver, Jason Reifler, and Kristin Thompson Sharp.
2006. “Foreign Policy and the Electoral Connection.” Annual Review of Political Science
9 (1):477-502.
Baum, Matthew A. and Philip B.K. Potter. 2015. War and Democratic Constraint: How the Public
Influences Foreign Policy. Princeton, NJ: Princeton University Press.
Berinsky, Adam J. 2009. In Time of War: Understanding American Public Opinion from World
War II to Iraq. Chicago, IL: University of Chicago Press.
Berinsky, Adam, Gregory Huber, and Gabriel Lenz. 2012. "Using Mechanical Turk as a Subject
Recruitment Tool for Experimental Research." Political Analysis 20: 351–368.
Brooks, Deborah J. and Benjamin A. Valentino. 2011. “A War of One’s Own: Understanding the
Gender Gap in Support for War.” Public Opinion Quarterly 75 (2): 27-86.
Calvert, Randall L. 1985. “The Value of Biased Information: A Rational Choice Model of Political
Version 2. Princeton Survey Research Associates International [producer]. Storrs,
CT:Roper Center for Public Opinion Research, RoperExpress [distributor],
accessed Feb-22-2016.
3. Kosovo 1999
a. Notes: The UNSC would not authorize military intervention (due to Russia’s veto
threat), but NATO’s North Atlantic Council authorized war instead.
b. Q: Next we have a few questions about the situation in the Kosovo region of
Yugoslavia -- the area that has been in the news recently. As you may know, last
week, the military alliance of Western countries called NATO launched air and
missile attacks against Serbian military targets in Yugoslavia. Do you favor or
oppose the United States being a part of that military action?
c. Ref: CNN/USA Today. Gallup/CNN/USA Today # 9903019: Kosovo, Mar, 1999
[dataset]. USAIPOCNUS1999-9903019, Version . Gallup Organization
[producer]. Storrs, CT:Roper Center for Public Opinion Research,
RoperExpress [distributor], accessed Feb-22-2016.
4. Libya 2011
a. Notes: The UNSC passed a Chapter VII resolution to protect civilians, but many
including Russia argue that the resolution did not cover NATO’s airstrikes that in
effect aided rebels in overthrowing Qaddafi.
b. Q: A coalition of international forces has been enforcing a no-fly zone over Libya,
and attacking Libyan forces when they're believed to be threatening civilians. Do
you support or oppose the participation of U.S. military aircraft in this effort? c. Ref: ABC News / The Washington Post. ABC News/Washington Post Poll: June
Monthly--Federal Budget Deficit/War on Terrorism/2012 Presidential Election,
Jun, 2011 [dataset]. USABCWASH2011-1124, Version 2. Langer Research
Associates/TNS Intersearch [producer]. Storrs, CT:Roper Center for Public