Mechanisms of Morality: Sources of Support for Humanitarian Intervention

Transcription

1 Mechanisms of Morality: Sources of Support for Humanitarian Intervention Sarah Kreps Associate Professor of Government Cornell University Sarah Maxey PhD Candidate in Government Cornell University **DRAFT: PLEASE DO NOT CITE WITHOUT AUTHORS PERMISSION** Abstract As ongoing United States (U.S.) airstrikes and aid drops in northern Iraq and the 2011 operation in Libya illustrate, humanitarian interventions have become a common feature of U.S. foreign policy. Observational public opinion research indicates that humanitarian interventions receive high levels of support from the U.S. public, but why are respondents more likely to approve of humanitarian interventions? In particular, how important is the humanitarian aspect of these interventions relative to information about cost and strategic importance? We answer these questions with three forms of evidence about support for humanitarian intervention: public opinion data on all post-cold War interventions and a pair of survey experiments, which together show that individuals support humanitarian intervention not for instrumental reasons but out of a sense of moral obligation, defined most saliently in terms of individual-level foundations such as whether harm would be done to weak or vulnerable civilians in the absence of intervention. The analysis has important implications for theories about post-cold War intervention norms and the circumstances under which states use military force. 1

2 I. Introduction Humanitarian interventions deployments of military force across borders to protect civilians from violence (Finnemore 2003, 53) have become a prominent form of United States (U.S.) military action in the post-cold War period. Of the U.S.-led interventions in the last twenty-five years, about half have had primary objectives that were humanitarian in nature. A typical example is the 2011 intervention in Libya, carried out because, as Obama put it, Qadhafi had forfeited his responsibility to protect his own citizens and created a serious need for immediate humanitarian assistance and protection, with any further delay only putting more civilians at risk (Obama 2011). While the prospect of saving strangers has gained acceptance as a moral form of intervention (Wheeler 2000; Finnemore 2003), the practice raises fundamental questions about the domestic sources of state behavior. In particular, the paper focuses on the question of whether the theoretical sense of morality attributed to humanitarian intervention is shared by the United States (U.S.) domestic audience, whose support is crucial for initiating and sustaining the use of force. If so, what are the mechanisms of morality through which the U.S. public s support for humanitarian intervention operates? Alternatively, to the extent that the public supports humanitarian intervention, how much is that support driven by instrumental factors that the public associates with it, such as the reputational costs of inaction or the expectation of lower costs? On the one hand, there is little reason to expect humanitarian intervention to generate high levels of support, let alone that the populace would see intervention as both morally permitted and morally required (Wheeler 2000, 310). A number of studies have found that the populace is driven by dispassionate, hard-headed national interest rather than humanitarian objectives (Drezner 2008, 52) and is more willing to use military force when national interests 2

3 are at stake versus when they are not (Herrmann et al 1999, 558). On the other hand, studies of public opinion have generally shown that humanitarian interventions tend to elicit higher levels of public support (Eichenberg 2005; Jentleson and Britton 1998), adding credence to scholarship that expects public attitudes to reflect the internalization of humanitarian intervention norms (Finnemore 2003 and 1996; Welsh 2004; Wheeler 2000). While public opinion polls offer a useful starting point for mediating between these two perspectives, relying solely on observational data presents problems. Among them is that even if humanitarian interventions elicit higher levels of support, observational data and existing research cannot speak directly to why the public supports humanitarian intervention. We propose a more direct approach for evaluating support for humanitarian intervention. Using a pair of survey experiments, the paper first asks whether individuals are more likely to support humanitarian intervention. It then considers the basis on which support for humanitarian intervention turns, whether on normative factors such as the morality of intervention or on factors that individuals associate with humanitarian intervention, such as an expectation of lower costs or of strategic costs for inaction. A pretest shows that respondents are significantly more likely to support humanitarian interventions relative to the standard foreign policy restraint (FPR) scenario (Jentleson 1992; Jentleson and Britton 1998) in which a country invades a neighbor. Based on this finding, we then carry out a nationally representative survey experiment. We incorporate two cost and consequence moderators identified from the pretest burden sharing and reputational costs and assess the effect of morality mediators on support for humanitarian intervention (Baron and Kenny 1986). These mediators draw on moral foundations theory (Graham et al. 2013; Haidt & Graham 2007; Haidt & Joseph 2004; Kertzer et al 2014) and include five sets of mechanisms associated with individual and collective senses of 3

4 morality that theoretically undergird support for humanitarian intervention. In particular, we suggest that to the extent that they support humanitarian intervention, individuals could theoretically be responding to mechanisms that involve the protection of civilians (individuallevel mechanisms of morality) or to the maintenance of international society (community-level mechanisms). In practice, each case of humanitarian intervention could draw on a combination of moral mechanisms, for example, concerns both about the harm done to civilians and the preservation of international order. However, determining which mechanisms most directly influence public support for intervention provides a more nuanced understanding of how individuals form attitudes about the use of force, when humanitarian interventions are most likely to be popular, and how international norms such as the Responsibility to Protect (R2P) influence the likelihood that states will respond to humanitarian crises. Additionally, determining when the public is most likely to rally behind humanitarian causes has policy implications for how states can be mobilized to act consistently with the principles of R2P and prevent future mass atrocities. The rest of the article proceeds as follows. First, it provides a brief overview of existing scholarship on humanitarian intervention, in particular, the largely post-cold War practice of military force aimed at preventing or ending widespread and grave violations of the fundamental human rights of individuals other than its own citizens, (Holtzgrefe 2003, 18) and the degree to which these humanitarian interventions were sanctioned by the public compared to other types of interventions. Second, we develop three potential sources of individual attitudes about humanitarian intervention, including views about moral obligation, cost, and consequences. The next section presents three forms of evidence about support for humanitarian intervention. As a first evaluation of public attitudes toward humanitarian intervention, we 4

5 provide public opinion data on all post-cold War interventions. While this data offers evidence of higher levels of support for humanitarian intervention, it is only suggestive because it speaks neither to the independent effect of the humanitarian aspect of an intervention nor to the reasons why the public might support this type of intervention. Our main evidence takes the form of the two experimental tests discussed above, which together provide strong evidence for the argument that individuals support humanitarian intervention not for instrumental reasons but out of a sense of moral obligation, defined most saliently in terms of the individual-level foundation of whether harm would be done to weak or vulnerable civilians in the absence of intervention. The final section concludes with a discussion of how individuals form opinions about the use of force and, in turn, the circumstances under which states are more likely to pursue humanitarian intervention. II. Humanitarian Intervention Norms The end of the Cold War brought a number of changes to how states use military force abroad. First, as Voeten (2001, 846) observes, the number of United Nations Security Council (UNSC) vetoes plummeted as the great power paralysis that had characterized the Cold War gave way to a more cooperative tenor. Concurrently, the United Nations dramatically increased the number of Article VII resolutions it issued on threats to international peace and security from 22 during the Cold War to 145 in the eight years after and authorized 15 peacekeeping missions compared to one authorized during the Cold War. 1 Second, an ethos about the protection of civilians began driving how states used force abroad (Finnemore 1996; 2003). Whereas the pursuit of economic and security interests 1 Humanitarian interventions can be distinguished from peacekeeping operations by the absence of target 5

6 motivated the use of force in the Cold War, 2 humanitarian objectives came to offer a primary motivation for intervention. 3 The international community also increasingly galvanized behind the idea of saving strangers (Wheeler 2000), or intervening with military force when state leaders violate the rights of their populace. As Kofi Annan put it when accepting the Nobel Peace Prize in 2001, the sovereignty of states must no longer be used as a shield for gross violations of human rights (2001). This idea is enshrined in the Responsibility to Protect, which calls for international action when states are unable or unwilling to protect their citizens (Evans 2004, 87). This growing consensus that the protection of civilians legitimates the use of military force came to be associated with the norm of humanitarian intervention (Bellamy 2005, 33). As evidence for the norm of humanitarian intervention, Finnemore points not only to states increasing propensity to intervene for humanitarian purposes but also to leaders public justifications for intervention (Finnemore 1996, 158). The continued prevalence of humanitarian justifications is evident in Obama s recent accounts of U.S. airstrikes against the Islamic State of Iraq and the Levant (ISIL), which helped save the lives of thousands of innocent men, women, and children (Obama 2014), as well as in justifications for the 2011 Libyan intervention that focus on Qadhafi s forfeiting of his right to sovereignty (Obama 2011). Presidents claims that force is driven by humanitarian objectives suggest leaders either feel a personal sense of moral obligation or expect their domestic audiences to be more receptive to interventions couched in moral terms. 2 For example, while humanitarian concerns were raised in U.S. interventions in the Dominican Republic and Grenada, the protection of civilians was clearly subordinate to the strategic importance of the region, the threat of communism, and the protection of American lives. 3 A clear example of the primacy of humanitarian objectives is the U.S. intervention in Somalia. As Finnemore observes (1996, 156), economically, Somalia was insignificant to the United States, but the U.S. nonetheless led a multilateral intervention to assist in relief operations in December

7 III. Morality and International Relations Support for humanitarian intervention driven by the public s sense of moral obligation would not be surprising given the frequency with which scholars invoke morality in studies of foreign policy. Tannenwald (2005, 11), for example, characterizes her discussion of the nuclear taboo as a moral interpretation of nuclear weapons and argues, the taboo possesses an important moral component, for which power and interest explanations cannot fully account. She asserts that this moral component is driven by an intuition that killing civilians and causing excessive mass destruction is wrong, but does not directly test the role of morality or its components (Tannenwald 2005, 11). Similarly, in their examination of the democratic peace, Tomz and Weeks (2013, ) reference democratic citizens feeling morally reluctant or a moral aversion to democracies attacking other democracies, but do not expand on how they or the respondents think about morality. In studies of intervention more specifically, morality is presented as an important mediator of the effects of UNSC approval (Tingley and Tomz 2014) and as a deciding factor in the tension between the norms of humanitarian intervention and state sovereignty (Wheeler 2000, 3; Finnemore 2003). Despite the frequent invocation of morality as a mediator for the relationship between public opinion and the use of force, international relations scholars have rarely elaborated on the sentiments that constitute a sense of moral obligation. How individuals think about morality matters for understanding when it is most likely to influence public opinion, whether subsets of the population will be comparatively more receptive to different moral claims, and how individuals approach the sometimes competing humanitarian and military components that make up humanitarian intervention. Developing moral principles in the context of foreign policy, Kertzer et al. (2014) and Liberman (2006) offer important departures from the typical scholarly silence on the attributes of 7

8 morality in international politics. Liberman (2006, 687) uses death penalty support as a proxy for retributiveness or humanitarianism to explain why foreign villains and good-versus-evil framing heighten public support for war and provides evidence that the aspects of morality influence public opinion in different directions. Kertzer et al. (2014) take a step further by using survey questions that directly probe different categories of morality. Although Liberman (2006) and Kertzer et al. (2014) take an important step towards understanding morality, the authors rely on correlations of moral foundations with foreign policy attitudes. For example, Kertzer et al. (2014) ask, when you decide whether something is right or wrong, to what extent are the following considerations relevant to your thinking? Responses cover each of five moral foundations harm/care, fairness/reciprocity, in-group/loyalty, authority/respect, and purity/sanctity and are then correlated with foreign policies such as individuals attitudes about the Kyoto Protocol, the decision to intervene in Iraq, and general foreign policy attitudes such as whether it is appropriate to take steps to stop expansionist powers. Moreover, while Liberman (2006) and Kertzer et al. (2014) associate hawkishness with retributivism and militant internationalism, respectively, they do not provide clear analytical expectations for humanitarian intervention. Humanitarian interventions combine militarism with humanitarianism components that studies of morality often analyze separately. Thus, existing analyses cannot speak to which aspects of morality are most likely to affect support for an important subset of military force, humanitarian intervention. We build on Liberman (2006) and Kertzer el al. s (2014) approach to evaluate whether and which moral foundations provide the basis of support for humanitarian intervention. To conceptualize the idea of morality that may underlie support for humanitarian intervention, we 8

9 use moral foundations theory from social psychology, which disaggregates morality into individual and community-based attributes (Graham et al. 2013). According to Haidt (2007, 998), individuals sense of intuitive ethics is driven by five different foundations that can be divided into two categories. The first category includes individualizing foundations that focus on whether individuals are being harmed or treated with compassion, as well as on fairness and reciprocity. A second category addresses community welfare writ large and is made up of binding foundations that can be subdivided into respect for social order and obedience, ingroup cohesion, and standards of social decency. On the one hand, consistent with individual-level mediators, attempts to mobilize public support are often based on the assumption that providing evidence of harm done to civilians by easily identifiable perpetrators increases the popularity of interventions (Keck and Sikkink 1998). This assumption is consistent with moral foundation theory s concept of attentiveness to harm/care of individuals (Haidt 2007). In the context of intervention, the observable implication is for individuals to act on the basis of whether harm was done to weak or vulnerable civilians (Kertzer et al 2014; Appendix). The public could also manifest concerns based on individuallevel factors through attentiveness to fairness/reciprocity. The observable implication associated with this factor is support for intervention that is driven by the concern that violators of human rights are held accountable for their actions. 4 Alternatively, if the public is motivated by a more general need to promote one s own values and uphold the standards of the international community, the community-based attributes of morality are more likely to be persuasive. In terms of respect/authority, the plausible rationale related to support for the use of force would include whether the invading country s actions caused disorder within international society. Obligations to preserve cohesion against out- 4 These indicators are adapted from Kertzer et al. (2014). 9

10 group interference in the context of intervention could be associated with being true to one s country s values and identity. Beyond these two community-level values, the public could also be concerned with the values of purity/sanctity (Haidt 1997; Kertzer et al 2014), which would be manifested in the context of intervention as whether an invading country violated standards for how civilized states should behave. Table 1 summarizes the morality mechanisms, including their associated characteristics, and the way we operationalize them in the context of support for humanitarian intervention. Table 1. Morality Mechanisms 5 Foundations Individualizing (Cooperative Internationalism) Community/ Binding (Militant Internationalism) Moral Values Harm/ Care Fairness/ Reciprocity Authority/ Respect Ingroup/ Loyalty Associated Traits Concern for the suffering of others, compassion, caring for others and protecting them Individuals should be treated with equal concern and respect, inequality is unfair and unjust Maintenance of social hierarchies to assure social order, obedience, respect, role fulfillment Obligations to group to preserve cohesion against outgroups Indicator Whether or not harm would be done to weak or vulnerable civilians. Whether or not the perpetrators would be able to get away with violating human rights. Whether or not the invading country s actions caused disorder within international society. Whether or not a failure to respond would betray my country s values and identity. Purity/ Sanctity Admonitions to maintain bodily and spiritual cleanliness Whether or not the invading country s actions violate my expectations for how a civilized country should act. Differentiating between morality mechanisms allows us to determine which types of humanitarian interventions are most likely to generate public support and speak to why support for interventions fluctuates over time. If individual-level mechanisms are most salient, public support should be highest in cases with visible evidence of harm done to civilians or easily 5 Values and language for the first three columns are drawn from Kertzer et al. (2014, 829). 10

11 identifiable perpetrators who can be held accountable for human rights abuses. Public concern with civilian harm is in line with studies of the CNN effect which suggest that the graphic and emotive media reporting of suffering people, as characterized the humanitarian crises in northern Iraq and Somalia, drums up domestic demands for action (Robinson 2011, 3). Individual-level explanations are also consistent with increased support that stems from individuals desire to hold perpetrators accountable for human rights abuses, as in the reaction to the Kony 2012 campaign which focused on bringing a warlord accused of recruiting child soldiers to justice. If community-level mechanisms are most salient, however, the public will be most receptive to interventions that help enforce international law or maintain international society. This sub-set of mechanisms draws on individuals perceptions of how international society should be ordered, their expectations for how civilized states behave, and where they place their own state s national identity within the international order. The community-level explanations expect interventions to generate the most public support when intervention responds to an international call for action, breaks international law or conventions as in the case of Syria s use of chemical weapons in 2013 or taps into the state s self-declared values and national identity. The potential strength of community-level mechanisms is reflected in efforts, such as the Responsibility to Protect R2P, to formalize universal standards for human rights practices and expectations for international action in the face of gross human rights violations. By creating an international consensus on standards for behavior and action, R2P aims to prevent any future failure of international will in the face of mass atrocities, as was seen in the Rwanda case (ICISS 2001, 1). Such efforts will help to generate the domestic political will necessary for 11

12 action only if public support is driven by mechanisms that emphasize community-level standards of decency. IV. Alternative Mechanisms of Public Support Signals of Cost Instead of, or perhaps in addition to a sense of moral obligation, support for humanitarian interventions could also be driven by attributes that the public associates with these missions, for example that they are less costly or risky operations. Humanitarian interventions have tended to be characterized in one or more of the following ways. First, they have invariably had some form of multilateral support both in terms of authorization and the implementation of operations. The only post-cold War humanitarian intervention that did not have UNSC authorization was the 1999 intervention in Kosovo, which received authorization from the North Atlantic Treaty Organization (NATO). Second, humanitarian interventions have also received material support from allies that limits the depth of the United States commitment. One example is the 1994 Haiti intervention, which was UNSC-authorized and initially led by the United States. After the U.S. achieved its objectives in early 1995, it handed off the peacekeeping responsibilities to a robust multinational force. The result was that rather than having an openended commitment, the United States could quickly channel its resources elsewhere after the initial intervention (Kreps 2007). Existing literature suggests that interventions carried out multilaterally are associated with higher levels of public support in part because international approval serves a legitimating (Thompson 2006) or signaling function in terms of the importance of intervention (Chapman 2009). These findings that the public is more likely to approve of multilateral intervention (see also Grieco et al 2011) are consistent with other studies that demonstrate that higher burdens of any kind decrease support. For example, some scholars have found that high casualties dampen 12

13 support for intervention (Mueller 1973; Gartner 2008), so connections between humanitarian interventions and low risks of casualties could explain increased approval. 6 Similarly, humanitarian interventions may also be linked to lower financial costs, which are found to drive public support (Geys 2010). Taken together, the cost signaling mechanisms multilateral approval, burden-sharing, low expectations of casualties, and low expectations of financial costs imply that support for humanitarian interventions may be driven by characteristics the public automatically associates with such operations rather than any genuinely humanitarian impulse. Strategic Signals Finally, the public may support humanitarian interventions because they understand that inaction can lead to a number of negative strategic consequences. Scholars have consistently argued that foreign states internal security influences the U.S. national interest in two main ways: by creating instability and by threatening the U.S. reputation for resolve. The internal instability created by humanitarian crises is associated with safe havens for terrorists, regional spillover violence, and access to economic resources. Piazza (2008, 469) provides empirical support for the link between internal instability and terrorism with his finding that, states plagued by chronic state failures are statistically more likely to host terrorist groups that commit transnational attacks, have their nationals commit transnational attacks, and are more likely to be targeted by transnational terrorist themselves. If the U.S. public has internalized this connection between weak states and terrorist threats, heightened support for humanitarian interventions could be driven by concerns about the national security implications of inaction. A second, related consideration that precedes the post-9/11 focus on terrorism is the concern that inaction in a humanitarian crisis may produce broader regional and international 6 For conditions on this finding, see Gelpi, Feaver, and Reifler (2009). 13

14 instability (Fearon and Laitin 2004, 13). Through the spread of violence and refugees, humanitarian crises may generate instability in regions of strategic importance with consequences for U.S. national security (Adamson 2006). 7 Therefore, the public may support humanitarian interventions not for their moral desirability, but out of a self-interested understanding that global instability threatens domestic security. This mechanism suggests that while the rhetoric surrounding interventions changed following the Cold War, the public still responds primarily to national security concerns. 8 In addition to the negative consequences of instability, public support may result from concerns that inaction could harm the United States reputation for resolve and ability to achieve foreign policy goals in the future. Dafoe et al. (2014, 377) explain that states have strategic incentives to acquire certain reputations and status, since it can confer the belief on the part of other states that the holder of the strong reputation will stand firm in the future, helping it win contests of expectations. While some empirical studies suggest that reputation is not a major factor in crisis decisions (Huth 1997; Press 2005), Walter (2009, 19) finds that governments care a lot about their reputation and it is quite rational for them to do so. Dafoe et al (2014, 381) conclude that leaders, policy elites, and national populations are often concerned, even obsessed, with their status and reputation. Leaders and publics worry about their own as well as their state s credibility. Therefore, if the public perceives inaction in the face of humanitarian crisis as a threat to the United States credibility, heightened support for humanitarian intervention could stem from a fear of reputational costs. 7 For example, as Holzgrefe (2003, 42) notes, the Security Council determined that the Somali civil war represented a threat to international peace and security. It authorized a peacekeeping force to create food lines in Somalia that would address the concerns about starvation but also deal with the underlying causes of migration creating disruptions in neighboring countries. 8 Individuals may also support intervention out of concern for the economic consequences of inaction, particularly when the conflict threatens access to oil or other resources. Rotberg (2003) demonstrates the relationship between instability and states ability to generate and export natural resources. 14

15 III. Evidence of Public Support for Humanitarian Intervention Beyond the increasing prevalence of humanitarian intervention and international policies such as R2P, observational public opinion data is commonly cited as quantitative evidence of the contemporary salience of humanitarian norms. Eichenberg (2005) and Jentleson and Britton (1998) suggest that interventions with humanitarianism as the principle policy objective of intervention receive higher levels of support than questions that framed intervention in terms of internal political change. These authors draw on the average levels of support across the potential cases of post-cold War cases intervention. 9 While a valuable starting point, the existing comparisons typically sidestep cases of nonintervention, which would potentially offer valuable insights in terms of showing the relationship between public opinion and the use of force. Moreover, the existing studies tend to include a number of questions that seek to gauge the intensity of support for intervention and sensitivity to costs, for example by asking whether individuals would support intervention even in the event of a certain number of casualties. With these concerns in mind, we compiled data not only on the most recent cases of intervention, but also interventions that were proposed but not undertaken. We also excluded questions intended to address scope conditions for support rather than support for the purpose of intervention, which is the focus of our research question. Case Motivating Concern Type of Force Average Public Opinion Prior to Intervention (Favor %) Average Public Opinion During Intervention (Favor%) Gulf War (1991) Security Ground, Air 69% 79% Northern Watch ( ) Humanitarian Air % Somalia (1992- Humanitarian Ground, Air 72% 70% 9 Cases are considered as potential interventions if: 1) they warranted a national presidential address drawing attention to the underlying crisis, 2) the threat of force was explicitly placed on the table or the president vowed to consider all means necessary, and 3) public opinion polls were conducted that asked respondents if they favor or opposed the use of force. 10 Because the Northern Watch operation immediately followed the end of the Gulf War, we have thus far not come across direct poll questions that address support for this operation prior to its initiation. 15

16 1993) North Korea ( ) Security N/A 51% No Intervention 47 % 11 Rwanda (1994) Humanitarian N/A 37% 12 No Intervention Haiti (1994) Security Naval Embargo, Ground 69% 13 Bosnia ( ) Humanitarian Air 51% 26% 38% 50% 14 39% 15 Kosovo (1999) Humanitarian Air 51% 57% Afghanistan (2001-present) 85% (10/07/ /09/2001) Security Ground, Air 75% 50% (12/10/2001- present) Iraq ( ) Security Ground, Air 75% 50% Sudan: Darfur ( ) Humanitarian N/A 49% No Intervention Libya (2011) Humanitarian Air 51% 55% Syria (2013) Humanitarian Air (proposed) 39% No Intervention Syria/ISIL (2014) Security Air 59% 60% 16 Table 2. Post-Cold War instances of proposed and actual intervention. Source: Roper Center ipoll database. Drawing on questions from the Roper Center s ipoll database that directly asked respondents whether they favored or opposed a military action that was either proposed or carried out, Table 2 reports average public opinion prior to and during the intervention. In general, the levels of support registered were higher than recorded by Eichenberg s measure (2005, 157), with the exception of Haiti, where levels of public support for intervention are nine percent lower. That levels are somewhat higher is not surprising given that the previous study 11 Combined support for use of force to stop the killing and to provide aid. 12 Support for the use of force to stop the killing. 13 Support for the use of force to provide aid. 14 Air support for NATO operations. 15 Direct strikes and contribution of forces to SFOR. 16 This measure is taken from a single Gallup poll released on September 23, 2014 that asked if respondents approved or disapproved of U.S. military action. The measure will be updated as more data becomes available. 16

17 includes many questions intending to gauge scope conditions by evaluating support under hypothetical and often costly conditions. As the table shows, there is wide variation in the level of public support for all interventions and humanitarian interventions do not obviously receive more support than others. Indeed, interventions with a more explicit security focus, such as the Gulf War and Iraq War, were associated with higher support than humanitarian interventions other than that in Somalia. Even cases that might point more definitively to security circumstances deriving higher levels of support are indeterminate based on observational data. For example, when President Obama presented potential U.S. intervention as a humanitarian operation in response to chemical weapons attacks in 2013, only 39 percent of the public, on average, supported the intervention. By contrast, more than a majority of respondents support the ongoing 2014 intervention in response the security threat posed by ISIL. 17 We cannot infer, however, that support increased because the main motive changed to a security concern or rather whether support for intervention motivated by the chemical weapons attack would have increased as Obama spent more time setting the agenda for intervention premised on a humanitarian argument (Cohen 1995, 88). Beyond being somewhat indeterminate in terms of whether the public is more likely to back interventions with a humanitarian focus versus security (foreign policy restraint) focus, observational data has at least two additional limitations. One is that it is subject to endogeneity in terms of the relationship between humanitarian intervention and public opinion. Leaders may engage in humanitarian intervention or raise the prospect only once they have gleaned some public interest. In this case, public support would motivate leaders to propose humanitarian intervention, making it difficult to infer whether humanitarian intervention contributes to public 17 Pew polls from October 22, 2014 report that 57 percent of respondents approve of the U.S. military campaign against Islamic militants in Iraq and Syria. 17

18 support rather than the other way around. Another limitation is that observational data cannot explain why the public supports humanitarian intervention. It is plausible that individuals support of intervention is motivated by both a sense of moral obligation and by the costs and consequences associated with humanitarian action. IV. Experimental Design An experimental approach helps address both of these methodological shortcomings cited above. First, by randomly assigning individuals to a control or treatment scenario, we are able to draw more accurate inferences about the independent effect of the treatment humanitarian intervention on support for the use of force. Second, using mediation analysis, we can examine how humanitarian intervention affects support, allowing us to explicitly test mechanisms that observational data cannot mediate. 18 We carry out two separate experiments, first a pretest that evaluates the basic contours of whether morality or more instrumental arguments about cost or consequences are stronger sources of support for humanitarian intervention, and the second a more thorough investigation of the mechanisms of morality we developed earlier. We study public attitudes for both theoretical and policy reasons. Scholars citing evidence of internalized norms have pointed to attitudes held by the populace. Risse and Sikkink (1999, 5), for example, suggest that international ideas and norms can affect domestic change from the bottom up, as changes in public opinion put pressure on leaders to act in normconsistent ways. This norm-specific finding follows the more general observation about democratic constraints, which is that leaders tend to be attentive to public preferences insofar as they may pay electoral costs for acting in ways that are at odds with public attitudes. Thus, these 18 The design is broadly based on Tomz and Weeks (2013) and Tingley and Tomz (2014). 18

19 attitudes shape the incentives leaders face in using force in particular ways and thereby affect the range of likely foreign policy behaviors (see, inter alia, Tomz and Weeks 2013). Pretest: Assessing Support for Humanitarian Intervention To provide an initial follow-up of the observational data on support for humanitarian intervention, and to examine the basic motivations underlying this support, we first carried out a pretest. We recruited 330 American adults via the internet in August 2014 and randomly assigned participants to the treatment and control conditions. 19 In brief, we designed a baseline scenario that presented respondents with information about the circumstances underlying a potential use of force, specifically that a leader abroad has deployed troops to a neighboring country. The neighbor is unable to defend itself. The U.S. president has called for the U.S. military to intervene to expel the invaders. 20 The treatment group was presented with the same scenario but introduced the prospect of humanitarian motivations for intervention. The treatment condition indicated that the U.S. president has called for the U.S. military to intervene because civilians, many of whom are women and children, are systematically being massacred, drawing on language President Obama used in 2013 to generate support for intervention in Syria (Garrett and Martin 2013). Individuals assigned to the control and humanitarian treatment were then asked whether they supported intervention. To investigate the mechanisms underlying support for humanitarian intervention, we asked a number of follow-up questions that corresponded with the three sets of explanations 19 We used Amazon s Mechanical Turk and while not nationally representative indeed, typically younger and more liberal than the population at large MTurk has been found to produce roughly comparable treatment effects as in nationally representative samples (Berinsky et al. 2012). The complete pretest instrument is available in the Appendix. 20 This baseline draws on Jentleson (1992) and Jentleson and Britton s (1998) description of foreign policy restraint (FPR) objectives. The authors find that FPR operations, along with humanitarian interventions, generate the highest levels of public support. Eichenberg (2005) confirms this finding. Thus, we compare scenarios in which there is least likely to be a difference in support for humanitarian and baseline interventions. 19

20 morality, costs, and strategic interest outlined above. First, we gauged morality in part by asking about whether the United States had a moral obligation to intervene (Finnemore 2003). Second, to assess whether support for humanitarian intervention is driven by what the prospect of intervention signals about the likely strategy and cost of the intervention, we asked a battery of questions about anticipated burden-sharing, cost of the operation, and military strategy. Third, we queried individuals about the potential consequences if the United States did not intervene, ranging from creating a breeding ground for terrorists to generating reputational costs. In this initial assessment, we investigated differences in the levels of support between the baseline and the humanitarian treatment. In the control condition, just over a majority (52.6%) of the respondents supported intervention whereas in the treatment, 80% favored intervention, more than a 27% increase in support, which was statistically significant at the 1% level (pretest results can be found in the Appendix). These levels of support are in the same range as recent responses from the Chicago Council on Global Affairs survey (2014) which found that 71% of respondents supported using U.S. troops to stop a government from committing genocide and killing large numbers of its own people. The lower levels for our baseline condition about removing a state from a country that it has invaded similarly approximate the levels of support for using troops if North Korea invaded South Korea (47%). Thus, the magnitude of support reported in the treatment condition is not out of sync with levels of public opinion in the early stages of actual interventions. The pretest also investigated the mechanisms behind higher levels of individual support in the humanitarian intervention scenario. This analysis leverages recent contributions in mediation analysis, in particular Imai et al. (2010 and 2011), who provide accounts of how mediation analysis can correct for the influence of intervening variables that standard regression 20

21 or probit models may neglect. We implemented this analysis using the mediate package in Stata, which is designed to estimate the role of causal mechanisms that transmit the effect of a treatment variable on an outcome (Hicks and Tingley 2011, 1). Using this approach, we can disaggregate the effect of each mediator. 21 By far, the most important factor accounting for the difference in support between the humanitarian treatment and baseline scenario was the perception that the United States had a moral obligation to intervene, accounting for 38% of the increase in support. The impact on individuals understanding of the costs and strategic consequences was far more diminished, with the perception that intervention would generate help in terms of burden-sharing and concerns about reputational costs affecting support for intervention being the only two factors that even registered as potentially relevant (though the latter did not reach statistical significance even if it was substantively significant). V. Main Experimental Test: Mechanisms of Morality Pretest results show that respondents are significantly more likely to support interventions with humanitarian objectives. They also indicate that a sense of moral obligation accounts for the largest portion of the increased support for humanitarian intervention. These findings provide general evidence for normative accounts of humanitarian norms but raise the question left underdeveloped in previous research, which is how individuals think about morality in a humanitarian context. To unpack the mechanisms underlying the invocation of morality, we 21 While mediating among these potential mechanisms presents its own methodological challenges, alternatives such as coding open-ended follow-up questions did not present a viable alternative since individuals would likely find it challenging to identify what aspects of morality motivate their responses. For a critique of mediation analysis, see Bullock et al

22 carried out a nationally representative survey experiment to distinguish among the mechanisms that underlie morality-based public support for intervention, incorporating both moderators and mediators. As the diagram below suggests, moderator variables could affect the strength of the relationship between the type of intervention and level of support (Baron and Kenny 1986, 1174) and mediators address the mechanisms through which that relationship operates. Figure 1. Role of moderators and mediators in support for the use of force. As in the pretest, participants were randomly assigned to either a baseline or treatment scenario. In this experiment, however, we also included two main moderator variables, contextual factors that change the impact of a humanitarian crisis on support for the use of force (Baron and Kenny 1986, 1174). The moderators build on the two factors other than morality that appeared to affect perceptions of humanitarian intervention in the pretest: the likelihood that other countries would help share the burden and the sense of how inaction would affect the U.S. s reputation for resolve and ability to achieve future foreign policy goals. In addition to the 22

23 two main baseline scenarios involving an invading country and a humanitarian crisis, the multilateralism moderator indicates that other countries would provide substantial resources to cover most of the cost of the mission compared to the alternative in which the United States covers most of the cost of the mission. The reputation moderator indicates that whether or not the U.S. intervenes will [not] affect the U.S. s reputation as a strong actor and/or make it harder for the U.S. to achieve future foreign policy goals. To better understand the respondents moral obligation, we included the potential mediators that correspond to the morality mechanisms discussed earlier, organized around the five pillars of harm/care, fairness/reciprocity, authority/respect, ingroup/loyalty, and purity/sanctity. These mechanisms were probed through randomized follow-up questions asked of each respondent, to which we added potential responses about whether individuals were responding to protect the security of their country, another option about the protection of other countries, and an option that none of those considerations came to mind. Beyond these mechanisms, we also asked individuals issue-specific questions about relative hawkishness, measured by whether people agree with the view that the use of military force usually makes problems worse (Horowitz and Levendusky 2011; Tomz and Weeks 2013) to control for whether propensities to use force account for observed responses. In addition, the survey included standard demographic questions about political identification, gender, education, and familiarity with the news. Additionally, we considered whether people who are naturally more altruistic with their own time or financial resources were more inclined to support intervention by asking whether individuals had volunteered time or donated money to any non-profit organization or charity in the last month. The full survey instrument and experimental conditions are provided in the appendix. 23

24 VI. Results In this section, we present the results of our main experiment, which examined support for humanitarian intervention and compared support relative to baseline foreign policy restraint intervention using both moderators and mediators. First, we present an overview of the findings, summarized in figure 2, which shows the changes in support for intervention scenarios relative to the baseline scenario in which a country invades a neighboring country that is unable to defend its borders (the foreign policy restraint scenario). In this baseline scenario, about 63% of the respondents favored the use of force by the United States to expel the invading country. As the figure shows, almost all of the treatment conditions exhibited statistically significant differences relative to the baseline scenario, other than the low multilateral help and reputational costs manipulations for the baseline scenario. In all cases the humanitarian intervention scenario was statistically higher than for the baseline scenario, with the baseline humanitarian intervention condition (without moderators) over 10% higher, significant at the 5% level FPR No Help FPR ML Help FPR No Rep Costs FPR Rep Costs HI HI No Help HI ML Help HI No Rep Costs HI Rep Costs Figure 2. Point estimates represent change in support for war compared to the baseline treatment. Vertical lines represent 95% confidence intervals. All tests are two-tailed. 24

25 In terms of support for humanitarian intervention relative to the baseline foreign policy restraint scenario, different types of individuals appear similarly moved by the prospect of humanitarian intervention. Indeed, as table 3 suggests, standard demographic factors such as gender, party identification, education, and age do not play a role in support for humanitarian intervention. Taking into account the additional covariates such as altruism and hawkishness, we do find that those who disagree with the view that military force makes things worse were more likely to support the use of force while those who had volunteered their time or money in the last month were no more likely to support humanitarian intervention, which suggests that individuals support for intervention is not driven by an underlying sense of morality but something specific to the humanitarian scenario itself. Table 3. Humanitarian Intervention and Support for War (1) Humanitarian Intervention Treatment.42 (.17) ** Party Identification (2=Rep, 1=Ind, 0=Dem) -.17 (.11) Gender (1=female) -.02 (.17) Altruism.14 (.18) Hawk.92 (.18) *** Age -.04 (.06) Education.01 (.9) News interest.35 (.21) Income -.00 (.00) Constant -.14 (.41) N 268 NB: Robust standard errors in parentheses. All significance tests are two-tailed; *p < 0.1, **p<0.05 *** p < 0.01 Evidence about Moderators 25

26 Next, we examine the role of moderators, looking at the impact of the multilateral assistance and reputational cost variables on support within a type of intervention. Consistent with our pretest findings, the prospect of multilateral help for an intervention had a larger effect than that of concerns for reputational costs, effects that held for both types of intervention. As table 4 shows, the effect of multilateral assistance is greater in the case of the baseline scenario of foreign policy restraint than it is for humanitarian intervention. The relatively larger effect may have to do with ceiling effects, however, since a larger percentage of respondents favored humanitarian intervention even in the context of having little multilateral support (73%) such that the prospect of multilateral assistance had reached an upper bound of individuals who would support intervention in any context. In both types of interventions, however, the prospect of multilateralism did impact support for intervention as expected in its hypothesized role as a moderator. That it increased support for intervention offers a challenge to previous research showing that the public is mainly concerned with the imprimatur of United Nations authorization (Chapman and Reiter 2004; Grieco et al 2011) and more consistent with recent research pointing to a public interest in the burden-sharing elements associated with multilateralism (Brooks and Brooks 2015). On the other hand, reputational costs did not have any meaningful effect on support for either type of intervention, with no statistical difference in a context where the reputational costs and impact on future foreign policy decisions was high compared to one where they were low. Consistent with Mercer s research on reputation (1996), individuals may find situational factors more persuasive in terms of whether the state should act, or may not buy the premise that current inaction would affect future foreign policy actions, consistent with Press s (2005) findings that past actions have little to do with future credibility and outcomes. 26