Kettles of Hawks: Public Opinion on the Nuclear Taboo PDF
Document Details
Uploaded by KidFriendlyBananaTree
null
2022
Janina Dill, Scott D. Sagan, and Benjamin A. Valentino
Tags
Summary
This article examines public opinion on nuclear weapons and non-combatant immunity in the United States, United Kingdom, France, and Israel. The research finds that attitudes are not exceptional to the United States, with Israeli respondents expressing the most hawkish preferences, and British the least supportive. The conclusions suggest a consequentialist reasoning behind public support for nuclear weapons.
Full Transcript
Security Studies ISSN: (Print) (Online) Journal homepage: www.tandfonline.com/journals/fsst20 Kettles of Hawks: Public Opinion on the Nuclear Taboo and Noncombatant Immunity in the United States, United Kingdom, France, and Israel Janina Dill, Scott D. Sagan & Benjamin A. Valentino To cite this arti...
Security Studies ISSN: (Print) (Online) Journal homepage: www.tandfonline.com/journals/fsst20 Kettles of Hawks: Public Opinion on the Nuclear Taboo and Noncombatant Immunity in the United States, United Kingdom, France, and Israel Janina Dill, Scott D. Sagan & Benjamin A. Valentino To cite this article: Janina Dill, Scott D. Sagan & Benjamin A. Valentino (2022) Kettles of Hawks: Public Opinion on the Nuclear Taboo and Noncombatant Immunity in the United States, United Kingdom, France, and Israel, Security Studies, 31:1, 1-31, DOI: 10.1080/09636412.2022.2038663 To link to this article: https://doi.org/10.1080/09636412.2022.2038663 © 2022 The Author(s). Published with license by Taylor & Francis Group, LLC Published online: 28 Feb 2022. Submit your article to this journal Article views: 6549 View related articles View Crossmark data Citing articles: 9 View citing articles Full Terms & Conditions of access and use can be found at https://www.tandfonline.com/action/journalInformation?journalCode=fsst20 SECURITY STUDIES 2022, VOL. 31, NO. 1, 1–31 https://doi.org/10.1080/09636412.2022.2038663 Kettles of Hawks: Public Opinion on the Nuclear Taboo and Noncombatant Immunity in the United States, United Kingdom, France, and Israel Janina Dill , Scott D. Sagan, and Benjamin A. Valentino ABSTRACT Recent scholarship has established that a majority of Americans will support the use of nuclear weapons and violate the principle of noncombatant immunity when American lives are on the line. Some scholars contend, however, that these hawkish American attitudes are an outlier and that other Western democratic publics have more fully internalized the nuclear taboo, as well as the prohibition on deliberately killing civilians. To investigate cross-national attitudes on these important norms, we conducted a survey experiment of American, British, French, and Israeli citizens. We find that American attitudes are not exceptional. Rather, Israeli respondents display the most hawkish preferences; French and American citizens are roughly equally hawkish; and the British public is consistently the least supportive of nuclear use or targeting civilians. Categorical prohibitions—against nuclear use and targeting civilians—do little to shape public opinion in these four countries. Instead, public opinion in each state follows the same consequentialist logic: a majority or near majority of respondents are willing to support using nuclear weapons when they are more effective than conventional options, but support declines when collateral civilian deaths rise. Respondents’ preferences for compatriots over foreign civilians and respondents’ retributiveness help explain individual-level variation in attitudes. At the end of the Cold War, both scholars and policymakers were optimistic about the power of norms to regulate the use of force. One of the most prominent theories asserted that the use of nuclear weapons had become subject to a powerful taboo and that a broader norm of noncombatant Janina Dill is the John G. Winant Associate Professor of US Foreign Policy at Oxford University. Scott D. Sagan is the Caroline S. G. Munro Professor of Political Science and a senior fellow at the Center for International Security and Cooperation, the Freeman Spogli Institute, at Stanford University. Benjamin A. Valentino is a professor of government at Dartmouth College. Supplemental data for this article can be accessed at publisher’s website. This article was originally published with errors, which have now been corrected in the online version. Please see Correction (http://dx.doi.org/10.1080/09636412.2022.2050037) ß 2022 The Author(s). Published with license by Taylor & Francis Group, LLC This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives License (http://creativecommons.org/licenses/by-nc-nd/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited, and is not altered, transformed, or built upon in any way. 2 J. DILL ET AL. immunity had taken hold, at least among Western publics and elites.1 Recent public opinion research focusing on the United States, however, has cast doubt on the strength of these norms. Survey experiments have demonstrated that although the American public prefers using conventional to nuclear weapons and prefers not to intentionally kill foreign civilians, these preferences readily give way to the desire to keep compatriot soldiers safe or to maximize the effectiveness of US military operations.2 Some scholars have reacted to these findings by suggesting that the US public is an outlier in its willingness to use nuclear weapons and kill noncombatants. In fact, we know very little about attitudes toward nuclear weapons or noncombatant immunity in other nuclear-armed countries. “The work on nuclear nonuse remains American-centric,” Michal Smetana and Carmen Wunderlich note, adding that “moving beyond the U.S. experience is crucial for broader understanding of the strength and nature of the norm.”3 This article helps fill this important gap by comparing public attitudes toward nuclear use and noncombatant immunity in four nuclear-armed democracies: the United States, the United Kingdom (UK), France, and Israel. Do these publics differ in the extent to which they have internalized the nuclear taboo and the prohibition on intentional attacks against civilians? Understanding norms regarding the use of force also requires understanding the important distinction between categorical and consequentialist reasoning. This distinction demarcates a central axis of debate in moral philosophy: Are our moral obligations principled and unmoving, or are Neta C. Crawford, “Targeting Civilians and U.S. Strategic Bombing Norms: Plus ça change, plus c’est la m^eme chose?” in The American Way of Bombing: Changing Ethical and Legal Norms, from Flying Fortresses to Drones, ed. Matthew Evangelista and Henry Shue (Ithaca, NY: Cornell University Press, 2014), 64–86; Steven Pinker, The Better Angels of Our Nature: Why Violence Has Declined (New York: Viking, 2011); Thomas C. Schelling, “An Astonishing Sixty Years: The Legacy of Hiroshima” (Prize Lecture, Nobel Media, 8 December 2005), https://www.nobelprize.org/uploads/2018/06/schelling-lecture.pdf; Nina Tannenwald, The Nuclear Taboo: The United States and the Non-Use of Nuclear Weapons since 1945 (Cambridge: Cambridge University Press, 2007). 2 Lisa Langdon Koch and Matthew Wells, “Still Taboo? Citizens’ Attitudes toward the Use of Nuclear Weapons,” Journal of Global Security Studies 6, no. 3 (September 2021): 1–18; Daryl G. Press, Scott D. Sagan, and Benjamin A. Valentino, “Atomic Aversion: Experimental Evidence on Taboos, Traditions, and the Non-Use of Nuclear Weapons,” American Political Science Review 107, no. 1 (February 2013): 188–206; Brian C. Rathbun and Rachel Stein, “Greater Goods: Morality and Attitudes toward Use of Nuclear Weapons,” Journal of Conflict Resolution 64, no. 5 (May 2020): 787–816; Scott D. Sagan and Benjamin A. Valentino, “Revisiting Hiroshima in Iran: What Americans Really Think about Using Nuclear Weapons and Killing Noncombatants,” International Security 42, no. 1 (Summer 2017): 41–79; Paul Slovic et al., “Virtuous Violence from the War Room to Death Row,” Proceedings of the National Academy of Sciences 117, no. 34 (25 August 2020): 20474–82; Michal Smetana and Marek Vranka, “How Moral Foundations Shape Public Approval of Nuclear, Chemical, and Conventional Strikes: New Evidence from Experimental Surveys,” International Interactions 47, no. 2 (2021): 374–90. 3 Michal Smetana and Carmen Wunderlich, “Forum: Nonuse of Nuclear Weapons in World Politics: Toward the Third Generation of ‘Nuclear Taboo’ Research,” International Studies Review 23, no. 3 (September 2021): 1072–99. 1 KETTLES OF HAWKS 3 they subject to tradeoffs?4 Many scholars have posited that the prohibitions on the use of nuclear weapons and the direct targeting of civilians are categorical, expected to apply regardless of circumstances. Yet consequentialist reasoning can also apply to nuclear weapons and the question of civilian targeting. Consequentialist norms identify certain outcomes as morally bad but explicitly ask us to balance those negative consequences against the benefits of our actions. Evidence of cost–benefit calculations is often taken as evidence for the weakness of norms in international relations (IR) scholarship, but traditional just war theory and international law recognize both categorical and consequentialist norms as applicable to warfare.5 Directly targeting civilians is categorically prohibited by the principle of noncombatant immunity. The principle of proportionality, in contrast, asks decision makers to weigh collateral civilian deaths against the military advantage of destroying a target. If some publics have internalized categorical prohibitions, whereas Americans follow a consequentialist logic, we would expect significant differences in the ways citizens of other countries evaluate the use of force. Of course, even if a consequentialist logic prevails across all nuclear-armed democracies, each public might weigh the particular costs and benefits of using force in different ways. In this article, we explore two potential factors that might influence citizens’ consequentialist calculations in different countries. The most immediate cost of using force tends to be loss of life among foreign civilians; the desired benefit is often to keep civilians at home safe. We, therefore, examine whether differences in what we call “compatriot partiality”—the extent to which citizens value the lives of compatriot civilians over foreign civilians’ lives—explain individual attitudes toward nuclear use and targeting civilians. We also investigate whether variation in retributiveness—the desire to punish others for perceived wrongdoing—might explain why some people are more hawkish than others. To examine these questions, we employ an original cross-national survey experiment designed around a series of mock news stories about a Libyabased terrorist organization that is planning a chemical weapons attack on the capital city of each country in which we surveyed. By manipulating selected aspects of each story, we can isolate the influence of the categorical and consequentialist logics. Larry Alexander and Michael Moore, “Deontological Ethics,” Stanford Encyclopedia of Philosophy, last modified 30 October 2020, https://plato.stanford.edu/entries/ethics-deontological/; Samuel Scheffler, The Rejection of Consequentialism: A Philosophical Investigation of the Considerations Underlying Rival Moral Conceptions (Oxford: Oxford University Press, 1994). 5 Janina Dill, Legitimate Targets? Social Construction, International Law and US Bombing (Cambridge: Cambridge University Press, 2015); Scott D. Sagan and Benjamin A. Valentino, “Not Just a War Theory: American Public Opinion on Ethics in Combat,” International Studies Quarterly 62, no. 3 (September 2018): 548–61; Michael Walzer, Just and Unjust Wars: A Moral Argument with Historical Illustrations, 5th ed. (New York: Basic Books, 2015). 4 4 J. DILL ET AL. We report four key findings. First, US attitudes toward nuclear use and noncombatant immunity are not exceptional. In fact, we find little support for the widespread internalization of a nuclear taboo or a categorical prohibition against targeting civilians in any country. Rather, the attitudes of citizens of all four countries follow a consequentialist logic. Majorities in the United States, France, and Israel, and 48 percent in the UK, support using nuclear weapons when they are described as more effective than conventional weapons at eliminating a significant terrorist threat, but that support declines substantially in every country when nuclear weapons provide no military advantage or when they are estimated to kill much larger numbers of foreign civilians. When subjects were informed that the strike would deliberately target civilian populations, support for the use of force declined significantly only in the UK, and majorities in all four countries supported this strike if carried out using conventional weapons. Second, when we asked subjects to make normative assessments of using nuclear weapons, majorities or near majorities indicated that the use of nuclear weapons was ethically wrong but supported it anyway. Crucially, ethical beliefs about nuclear use likewise largely followed a consequentialist logic in all four countries, with respondents rating a more effective nuclear strike as less unethical. Third, we demonstrate that differences in compatriot partiality and retributive beliefs help explain individual-level variation in attitudes toward nuclear use and killing civilians. Retributive and highly partial consequentialists are more hawkish than less vengeful consequentialists or those who weigh human life more equitably. Fourth, we find a consistent pattern in the relative “hawkishness” of citizens of each country.6 Israeli respondents display the most hawkish preferences in almost every scenario, while the British public is consistently the least willing to support using nuclear weapons or targeting civilians. French and American citizens are roughly equally hawkish. These findings represent a significant advance in our current understanding of public attitudes toward nuclear weapons and noncombatant immunity. The American public is not a conspicuous outlier among nuclear-armed democracies. None of the four publics have internalized either a nuclear taboo or a categorical prohibition on targeting civilians. Though most citizens across all four countries recognize the use of nuclear weapons as ethically wrong, they tend to support it if presented with significantly better consequences. Differences in the fundamental logic of public opinion across countries, therefore, do not explain cross-national differences in hawkishness. Highly partial and retributive consequentialists are more hawkish than consequentialists who think about human life in more equitable terms. The collective noun for hawks, swirling together in 6 We use the word “hawkish” to refer to subjects who are more willing to use nuclear weapons and to kill foreign civilians. KETTLES OF HAWKS 5 hot air to gain altitude, is a kettle. Hawks exist in all four of our nuclear democracies; some countries simply have many more hawks in their kettle. In the first section of this article we review previous efforts to examine cross-national differences in attitudes toward nuclear weapons and noncombatant immunity. Second, we articulate hypotheses about the individual-level sources of preferences on these issues and hypotheses about cross-national differences. We outline the research design in the third section. Fourth, we present the results. We close with a discussion of our findings’ implications. Explaining Cross-National Differences Although recent scholarship has raised doubts about the strength of the nuclear taboo and the noncombatant immunity norm in the United States, there has been little empirical study of public preferences in other nuclear-armed countries. Nevertheless, several scholars have suggested that American views on these subjects may be exceptional. Nina Tannenwald, for example, argues that “judging by public opinion polls and government policies, most Western democracies, with the possible exception of France, have been even more antinuclear than the United States.”7 William Walker maintains that the “nuclear taboo has always seemed present in the UK,” and “it would be most surprising if the British public shared the absence of aversion to the use of nuclear weapons found in a recent study of U.S. public opinion.”8 Avner Cohen also maintains that Israelis hold “a certain taboo”: “The notion that under virtually no conceivable circumstance can it [a nuclear weapon] ever be used.”9 Scholars have offered at least three reasons to expect that the US public is less likely to have internalized a nuclear taboo than publics in other Western democracies. First, the United States remains the only state to have used nuclear weapons in war. As Robert Jay Lifton and Greg Mitchell argue, Americans’ motivated biases to defend their government’s decisions might lead to “nuclear denial” of the use of nuclear weapons’ human consequences among Americans.10 Second, American public support for the atomic attacks on Japan remains significantly higher than in other countries. Our survey, for example, found that 46% of respondents in the United States approve of President Harry S. Truman’s decision, compared 7 Tannenwald, Nuclear Taboo, 376–77. William Walker, “Managing, Reconciling, and Manipulating the Deterrence and Disarmament Norms: The Case of the United Kingdom,” Contemporary Security Policy 39, no. 3 (2018): 423. 9 Tom Kutsch, “Nuclear Weapons Are Israel’s ‘Worst-Kept Secret,’ Says Israeli Historian,” interview with Avner Cohen, Al Jazeera, 18 October 2013, http://america.aljazeera.com/articles/2013/10/18/nuclear-weaponsareisraelsworstkeptsecretsaysisraelihistorian.html. 10 Robert Jay Lifton and Greg Mitchell, Hiroshima in America: A Half Century of Denial (New York: Avon Books, 1995). 8 6 J. DILL ET AL. to 27% in Israel, 25% in the UK, and 14% in France. Third, the US public may be particularly vengeful. The United States is the only country in our sample to retain and use the death penalty. Rachel M. Stein has shown that American citizens who support the death penalty are more “retributive” and more supportive of the use of nuclear weapons than nonretributive individuals.11 Likewise, Peter Liberman suggests “comparative research might find greater popular enthusiasm for punitive wars in the United States than in Western Europe, as well as other cross-national variation in the moral-punitiveness effect.”12 If the abolition of the death penalty in other nuclear-armed democracies is indicative of weaker retributive attitudes among the public, we would expect these publics to be less supportive of using nuclear weapons. Thus, Målfrid Braut-Hegghammer contends that “American [nuclear] preferences are likely to be a poor guide for estimating beliefs elsewhere” because of broad US support for capital punishment.13 Unfortunately, there have been few efforts to test these assertions empirically.14 The only previous cross-national poll on attitudes about the use of nuclear weapons in these four countries, published in 2007, found that 40.4% of subjects in the United States, 48.8% in the UK, 43.4% in France, and 21.6% in Israel agreed that using nuclear weapons “would never be justified.”15 Regarding noncombatant immunity, a 2011 Gallup poll reported that 49% of Americans, 43% of Israelis, 33% of British citizens, and 15% of French citizens indicated “it was sometimes justified for the military to target and kill civilians.”16 Neither of these polls, however, presented scenarios in which using nuclear weapons or deliberate killing of civilians might realistically be contemplated. Other cross-national studies have compared support for real-world conflicts, often comparing American to European attitudes toward the use of force.17 Robert Kagan pointedly summarized the recurring finding that Americans are particularly hawkish: “On the all-important question of power … American and European perspectives are diverging … Americans are 11 Rachel M. Stein, Vengeful Citizens, Violent States: A Theory of War and Revenge (Cambridge: Cambridge University Press, 2019), 95–97. Peter Liberman, “An Eye for an Eye: Public Support for War against Evildoers,” International Organization 60, no. 3 (Summer 2006): 714. 13 Målfrid Braut-Hegghammer, “Proliferating Bias? American Political Science, Nuclear Weapons, and Global Security,” Journal of Global Security Studies 4, no. 3 (July 2019): 385. 14 For an exception, based on historical case studies, see T. V. Paul, The Tradition of Non-Use of Nuclear Weapons (Stanford, CA: Stanford University Press, 2009). 15 Global Public Opinion on Nuclear Weapons (Vancouver: Simons Foundation, 2007), http://www. thesimonsfoundation.ca/sites/default/files/2007%20Poll%20on%20Global%20Public%20Opinion%20on% 20Attitudes%20Towards%20Nuclear%20Weapons_0.pdf. 16 “Views on Violence,” Gallup, 2011, https://news.gallup.com/poll/157067/views-violence.aspx. 17 For instance, Benjamin E. Goldsmith, Yusaku Horiuchi, and Takashi Inoguchi, “American Foreign Policy and Global Opinion: Who Supported the War in Afghanistan?” Journal of Conflict Resolution 49, no. 3 (June 2005): 408–29; Philip Everts and Pierangelo Isernia, “Trends: The War in Iraq,” Public Opinion Quarterly 69, no. 2 (Summer 2005): 264–323. 12 KETTLES OF HAWKS 7 from Mars and Europeans are from Venus.”18 Unfortunately, these studies all suffer from a critical limitation. They cannot isolate whether differences in support for a particular military operation might be caused by differences in underlying norms about the use of force, or by the profoundly divergent implications the same military operation may have for different countries. Our study seeks to minimize this confounding dynamic by asking respondents in different countries to consider the same hypothetical military crisis. We selected the details of this crisis to ensure, to the extent possible, that the scenario would implicate all four countries in the same manner. This design allows us to isolate the extent to which publics have internalized norms against nuclear use and killing civilians and how they weigh the consequences of different military options. Categorical and Consequentialist Logics International norms, Martha Finnemore and Kathryn Sikkink note, are “collective expectations about proper behavior for a given identity.”19 They hence rest on beliefs about what actors in the international realm ought to do. Ultimately, ordinary citizens hold these beliefs. For a social norm to exist, many individuals in a society need to share a certain normative belief.20 International norms should, correspondingly, be visible in public opinion across different countries. Crucially, when norms are internalized, they should be reflected in ordinary citizens’ preferences without being explicitly invoked. Finnemore and Sikkink argue that “internalized” norms “achieve a ‘taken-for-granted’ quality that makes conformance with the norm almost automatic.”21 If categorical norms against nuclear use and targeting civilians are internalized, most respondents’ preferences should simply reflect their strictures. Echoing the long-standing dichotomization of interests and norms in IR theory, existing work on public attitudes about nuclear weapons and noncombatant immunity has taken cost–benefit analyses to be indicative of the absence or weakness of norms against the use of particular weapons or the victimization of civilians.22 We depart from the existing 18 Robert Kagan, Of Paradise and Power: America and Europe in the New World Order (New York: Vintage Books, 2004), 3–4. 19 Martha Finnemore and Kathryn Sikkink, “International Norm Dynamics and Political Change,” International Organization 52, no. 4 (Autumn 1998): 891. 20 Michelle Jurkovich, “What Isn’t a Norm? Redefining the Conceptual Boundaries of ‘Norms’ in the Human Rights Literature,” International Studies Review 22, no. 3 (September 2020): 703. 21 Finnemore and Sikkink, “International Norm Dynamics and Political Change,” 904. Also see Thomas Risse and Kathryn Sikkink, “The Socialization of International Human Rights Norms into Domestic Practices: Introduction,” in The Power of Human Rights: International Norms and Domestic Change, ed. Risse, Stephen C. Ropp, and Sikkink (Cambridge: Cambridge University Press, 1999), 5. 22 Koch and Wells, “Still Taboo?”; Sarah Kreps and Sarah Maxey, “Mechanisms of Morality: Sources of Support for Humanitarian Intervention,” Journal of Conflict Resolution 62, no. 8 (September 2018): 1814–42; Press et al., “Atomic Aversion.” 8 J. DILL ET AL. literature by arguing that public preferences that follow a cost–benefit analysis are not necessarily indicative of the irrelevance of norms. Recent research theorizing norms has noted how apparent “violations” of international norms and laws can be the result of tradeoffs between different individuals’ moral frameworks23 and the acceptance of perceived “exceptional compromises.”24 In short, ethical norms are not always categorical, and norm compliance can reflect tradeoffs between different predicted consequences. Yet, proponents and opponents of the nuclear taboo thesis alike have long assumed ethical reasoning must be shaped by categorical prohibitions. The oldest of these categorical prohibitions is that on the intentional killing of civilians during war.25 Tannenwald emphasizes that the nuclear taboo co-originated with the noncombatant immunity norm. She writes that at the core of the nuclear taboo is “the moral intuition that it is wrong to kill non-combatants” and argues that “any use of nuclear weapons is prohibited. The normative prohibition has come to be an absolute one: the weapons themselves are proscribed.”26 We argue that those who prefer using nuclear weapons when they are more likely to destroy an important military target can also be guided by normative beliefs about right and wrong. These individuals rely on a consequentialist logic that supports using nuclear weapons when they believe doing so will be more effective and thereby save more lives. The principle of proportionality, for example, is a long-standing consequentialist norm about the use of force that is also codified in international law. The norm demands that collateral civilian deaths in strikes against legitimate military targets must not be excessive relative to the military advantage of destroying the target. It is important to note that consequentialism in its broadest form does not prejudge what counts as a cost and what counts as a benefit. What counts as a consequence to be maximized versus one that ought to be avoided depends on an underlying substantive theory of moral value.27 Rathbun and Stein, “Greater Goods”; Smetana and Vranka, “How Moral Foundations Shape Public Approval of Nuclear, Chemical, and Conventional Strikes.” 24 Anette Stimmer, “Beyond Internalization: Alternate Endings of the Norm Life Cycle,” International Studies Quarterly 63, no. 2 (June 2019): 270–80. 25 Helen M. Kinsella, The Image before the Weapon: A Critical History of the Distinction between Combatant and Civilian (Ithaca, NY: Cornell University Press, 2011); Seth Lazar, Sparing Civilians (Oxford: Oxford University Press, 2015); David Traven, “Moral Cognition and the Law and Ethics of Armed Conflict,” International Studies Review 17, no. 4 (December 2015): 556–87. This categorical principle of noncombatant immunity (or the principle of “distinction”) also forms a central part of the laws of war today: Article 48 of the 1977 Additional Protocol I to the Geneva Conventions stipulates that combatants “shall at all times distinguish between the civilian population and combatants.” See Article 48 of Protocol I to the Geneva Conventions of 1949, 8 June 1977, United Nations Treaty Collection, https://treaties.un.org/doc/Publication/UNTS/Volume%201125/volume1125-I-17512-English.pdf. 26 Tannenwald, Nuclear Taboo, 58–59, 62. Italics in original. 27 R. M. Hare, Moral Thinking: Its Levels, Method, and Point (Oxford: Oxford University Press, 1981); Walter SinnottArmstrong, “Consequentialism,” Stanford Encyclopedia of Philosophy, last modified 3 June 2019, https://plato. stanford.edu/entries/consequentialism/. 23 KETTLES OF HAWKS 9 Only on some accounts does moral consequentialism require that good consequences be defined in impartial reference to the common good, rather than an individual’s or a group’s particular values or egoistic aims.28 Individuals who have internalized consequentialist norms about the use of force could, therefore, still differ dramatically in their substantive preferences, but their preferences would follow a similar logic, shifting with differences in the projected consequences of the use of force. The most immediately obvious consequence of using nuclear weapons or targeting civilians is the loss of life among foreign civilians. At the same time, nuclear use and violations of noncombatant immunity are most often contemplated when these tactics are deemed necessary to protect civilians at home from impending attack. Preferences that follow a cost–benefit analysis, therefore, likely depend significantly on how individuals weigh the lives of compatriots versus those of foreign civilians.29 We call the extent to which individuals favor the lives of their own citizens over those of noncitizens “compatriot partiality.” A significant body of research suggests that variations in out-group hostility and in-group biases affect hawkishness.30 Different measures of what Kathleen E. Powers calls “unity nationalism,” for example, have been found to increase willingness to use force or escalate violence in a conflict.31 Higher degrees of cosmopolitanism, conversely, are associated with less hawkish attitudes.32 It is likely that compatriot partiality is related to certain forms of nationalism and in-group bias, but it provides a more direct measure of how individuals whose attitudes follow a consequentialist logic weigh the most significant and direct consequences of nuclear use and targeting civilians. More partial consequentialists are likely more hawkish than those who weigh human life more equitably. A second factor that has been demonstrated to influence individual preferences about the use of force and that may affect consequentialist calculations is retributiveness. While compatriot partiality might lead some to give less consideration to expected foreign civilian casualties as a cost of using nuclear weapons, the desire for retribution could lead others to consider foreigners’ suffering as a benefit of the use of force. As noted above, scholars have shown that variations in individuals’ retributiveness or vengefulness (often measured by support for the death penalty) help explain Rathbun and Stein, “Greater Goods.” For the finding that Americans prioritize sparing compatriot soldiers over foreign civilians, see Janina Dill and Livia I. Schubiger, “Attitudes toward the Use of Force: Instrumental Imperatives, Moral Principles, and International Law,” American Journal of Political Science 65, no. 3 (July 2021): 612–33. 30 Henri Tajfel and John C. Turner, “The Social Identity Theory of Intergroup Behavior,” in Political Psychology: Key Readings, ed. John T. Jost and Jim Sidanius (New York: Psychology Press, 2004), 276–93. 31 Kathleen E. Powers, Nationalisms in International Politics (Princeton, NJ: Princeton University Press, forthcoming); Richard K. Herrmann, “How Attachments to the Nation Shape Beliefs about the World: A Theory of Motivated Reasoning,” International Organization 71, no. S1 (2017): S61–S84. 32 A. Burcu Bayram, “Due Deference: Cosmopolitan Social Identity and the Psychology of Legal Obligation in International Politics,” International Organization 71, no. S1 (2017): S137–S163. 28 29 10 J. DILL ET AL. variations in the willingness to use force both in the real world and in experimental surveys.33 Stein argues that “vengeful individuals may be more willing to tolerate high cost conflicts and to disregard the standards of international law in order to give evildoers their just deserts.”34 Liberman similarly notes that “outrage at the guilty should carry over to punitiveness against others, such as civilians vulnerable to ‘collateral damage.’”35 This research suggests that variations in retributiveness might explain why some consequentialists are more hawkish than others. Hypotheses about Individual-Level Variation These differing logics yield several testable implications for public opinion on nuclear weapons and noncombatant immunity. If individuals have internalized a nuclear taboo, they should reject the use of nuclear weapons in favor of conventional military options, regardless of their respective effects. Likewise, if public opinion reflects an internalized categorical prohibition against targeting civilians, most citizens should oppose such attacks, regardless of other available options. Hypothesis 1a (H1a (nuclear taboo)). Changes in the relative military effectiveness of nuclear and conventional strikes or the number of foreign civilian casualties associated with either strike option should have no effect on individuals’ preferences for nuclear use. Hypothesis 2a (H2a (noncombatant immunity)). Individuals will prefer not to attack a military target rather than launch either a nuclear or conventional strike that intentionally targets foreign civilians. We also consider the possibility that some individuals have not fully internalized categorical norms that prohibit the use of nuclear weapons or the intentional targeting of civilians, but they nonetheless recognize these norms as salient to ethical judgements. If so, we expect that although individuals may support violating these prohibitions, they always designate nuclear use as unethical, even if a nuclear strike has better projected consequences than conventional alternatives. Moreover, we would expect that respondents would designate military strikes targeted against civilians as less ethical than those with similar consequences, but where civilians are unintended victims of the strike.36 Peter Liberman and Linda J. Skitka, “Vicarious Retribution in US Public Support for War against Iraq,” Security Studies 28, no. 2 (April–May 2019): 189–215. Stein, Vengeful Citizens, Violent States, 172. 35 Liberman, “An Eye for an Eye,” 695. 36 We do not assume that interests drive subjects’ preferences when these diverge from ethical assessments. H1b and H2b examine whether individuals recognize that categorical prohibitions exist. There are a variety of reasons why their preferences (H1a and H2b) may nonetheless not reflect the strictures of these norms, including that they consider other norms even more important. 33 34 KETTLES OF HAWKS 11 Hypothesis 1b (H1b (nuclear taboo ethics)). Changes in the relative military effectiveness of nuclear strikes should have no effect on individuals’ assessments of the ethics of the nuclear strike. Hypothesis 2b (H2b (noncombatant immunity ethics)). Individuals will assess any military strike that intentionally targets foreign civilians as less ethical than strikes that do not intentionally target foreign civilians. In contrast, if public opinion follows a consequentialist logic, we should find that most citizens’ preferences for a nuclear strike increase when it is more effective than a conventional strike and decrease when the strike is projected to kill more foreign civilians. Hypothesis 3 (H3 (nuclear consequentialist)). Individuals will be more likely to prefer a nuclear strike the more effective nuclear weapons are projected to be in destroying a target compared to a conventional strike. Hypothesis 4 (H4 (civilian consequentialist)). Individuals will be less likely to prefer a nuclear strike the greater the number of foreign civilians that nuclear weapons are projected to kill compared to a conventional strike. Among individuals whose preferences follow a consequentialist logic, we also expect differences in preferences depending on their degree of compatriot partiality and retributiveness. Hypothesis 5 (H5a (compatriot partiality)). Individuals with high compatriot partiality are more likely to prefer a nuclear strike than individuals with low compatriot partiality. Hypothesis 5b (H5b (retribution)). Individuals who are highly retributive are more likely to prefer a nuclear strike than individuals who are less retributive. Hypotheses about Cross-National Variation We also evaluate two hypotheses about cross-national variation in attitudes toward nuclear use and noncombatant immunity. First, some societies may have internalized the nuclear taboo and the categorical prohibition on intentional attacks against civilians more thoroughly than others. In particular, as noted above, scholars have offered various reasons why the nuclear taboo and the prohibition on noncombatant immunity may be weaker in the United States than in other nuclear-armed democracies. If so, we should find that the American public is more hawkish than citizens of other nuclear-armed democracies. Hypothesis 6 (H6 (American exceptionalism)). American citizens are more likely to prefer nuclear weapons and support attacks that intentionally target foreign civilians than citizens of other nuclear-armed democracies. 12 J. DILL ET AL. Second, we explore the possibility that the Israeli public may be more hawkish than the publics of other nuclear-armed democracies. There are at least two reasons to expect that Israelis might be more willing to use nuclear weapons and support intentional attacks on civilians. First, polls show that nationalism is higher among Israelis than citizens of most other Western countries. For example, a 2003 poll revealed that 79.0% of Israelis said they felt “very close” to their country, compared to 57.0% of French, 52.4% of Americans, and 33.7% of Britons.37 Further evidence suggests Israelis may be particularly partial toward their compatriots, especially in the context of the use of force. Public support in Israel is high, for example, for prisoner of war (POW) exchanges, which may reflect compatriot partiality. For instance, 78% of Israelis supported the 2011 swap of one Israel POW (Gilad Shalit) for 1,000 Palestinian prisoners.38 By comparison, during the 1985 TWA Flight 847 hijacking incident, only 27% of US respondents favored exchanging 700 Shiite Muslim prisoners for the 40 mostly American passengers and crew aboard the aircraft.39 As noted above, compatriot partiality is related to nationalism and may explain variation in individual attitudes toward the use of force. Israelis, therefore, may be more willing to use force in ways that sacrifice foreign lives to protect Israeli compatriots from harm. Second, just as some have asserted that US history may have rendered Americans less likely to internalize the nuclear taboo, Israeli perceptions of their history, including the Holocaust and Israel’s existential conflicts with its neighbors, may make Israelis more hawkish when responding to security threats. According to Daniel Bar-Tal and Dikla Antebi, Israeli Jews have developed a “siege mentality” that encourages them “to take drastic measures in order to prevent possible dangers.”40 Bar-Tal and Antebi conclude that “Israeli Jews feel that the goal of survival is so important that any means can be used to insure security. It sometimes may lead to courses of action considered extreme and unacceptable by the international community.”41 Indeed, a 1989 poll found that 64 percent of Jewish adults in Israel agreed that “all means are acceptable to secure the existence of the State of Israel.”42 Israel’s long history with terrorism may also make Israelis International Social Survey Programme, “National Identity II,” 2003, https://www.gesis.org/en/issp/modules/isspmodules-by-topic/national-identity/2003. 38 “The Peace Index: October 2011,” Dahaf Institute, Tel Aviv, Israel, October 2011, http://www.peaceindex.org/ indexMonthEng.aspx?num=235&monthname = October#anchor245. Also see Shelly Aviv Yeini, “Weighing Lives: Israel’s Prisoner-Exchange Policy and the Right to Life,” Minnesota Journal of International Law 27, no. 2 (Summer 2018): 493–526, https://scholarship.law.umn.edu/cgi/viewcontent.cgi?article=1308&context=mjil. 39 CBS News Poll, #1985-JUNE HIJK2: Flight 847, Hijack Survey #2 [Roper #31090687], https://ropercenter.cornell. edu/ipoll/study/31090687. 40 Daniel Bar-Tal and Dikla Antebi, “Siege Mentality in Israel,” International Journal of Intercultural Relations 16, no. 3 (Summer 1992): 251–75. See also Gil Merom, “Israel’s National Security and the Myth of Exceptionalism,” Political Science Quarterly 114, no. 3 (Autumn 1999): 409–34. 41 Bar-Tal and Antebi, “Siege Mentality in Israel,” 268. 42 Ibid. 37 KETTLES OF HAWKS 13 more hawkish when dealing with terrorist threats. Exposure to terrorism and political violence in Israel has been linked to support for the use of force and opposition to peace treaties.43 Oded L€ owenheim and Gadi Heimann’s study of the Second Lebanon War (Operation Just Desert) argues that Israel’s experience with terrorism encourages “revengeful retaliation,” including “excessive force, to harm civilians.”44 These factors suggest that Israelis may feel more vulnerable than citizens of other nations when exposed to the same threats, and that Israelis may be more willing to take extreme measures to counter such threats. Hypothesis 7 (H7 (Israeli exceptionalism)). Israeli citizens are more likely to prefer nuclear weapons and support attacks that intentionally target foreign civilians than citizens of other nuclear-armed democracies. Research Design To test these hypotheses, we designed and fielded an original survey experiment in the United States, the UK, France, and Israel. The online survey firm YouGov administered the experiment in June and July 2018.45 An average of 1,154 subjects, all citizens over the age of 18, completed the survey in each country.46 We selected these four countries because all are democratic, nuclear-weapons states that permit unrestricted public opinion polling. We focus on democratic states because mass public opinion is more likely to influence national policy in such states. We randomly assigned subjects in each country to one of four treatment groups or to a control group (each treatment group in each country constituted an average of 166 subjects and the control condition an average of 324 subjects per country). In each of the four treatment groups, we asked respondents to read a mock news story (see Figure 1 for a sample). Subjects in the control condition read no story. The stories were clearly identified as fictional but were designed to be equally realistic from the respondents’ points of view in each of the four countries. All subjects were encouraged to “imagine how you would feel about these events if they were happening in the real world today.” We constructed the stories to mimic typical newspaper articles of between 650 and 700 words. This format allowed us to emphasize and repeat key elements of the experimental Daphna Canetti, Julia Elad-Strenger, Iris Lavi, Dana Guy, and Daniel Bar-Tal, “Exposure to Violence, Ethos of Conflict, and Support for Compromise: Surveys in Israel, East Jerusalem, West Bank, and Gaza,” Journal of Conflict Resolution 61, no. 1 (January 2017): 84–113. 44 Oded L€owenheim and Gadi Heimann, “Revenge in International Politics,” Security Studies 17, no. 4 (2008): 685–724. 45 YouGov utilizes a technique called “active sampling” to approximate a representative sample in each of the four countries we surveyed. All results presented in this paper are weighted to match the age, gender, race, and education distributions of the target populations. See “Panel Methodology,” YouGov, https://yougov.co.uk/ about/panel-methodology/. 46 For more information about our sample, see Online Appendix 5. 43 14 J. DILL ET AL. Figure 1. Mock news story (US version, Condition 2). treatment in the story headlines, pull quotes, and summary tables, and to provide sufficient contextual information to limit potentially confounding assumptions subjects might make across different conditions.47 This format is also intended to heighten the realism of the experience for subjects, increasing the external validity of any findings.48 In all four versions of the news story, subjects read that their country’s political leadership had received credible information that a “previously unknown Islamist terrorist group” based in the remote town of Daraya in northern Libya was planning a chemical weapons attack on buses in the nation’s capital. The story reported that the information about the attack was initially uncovered by Swedish intelligence agents who had intercepted specialized chemicals and equipment used to produce sarin gas, a deadly nerve toxin. This scenario was designed to maximize the credibility of the intelligence regarding the terrorist plot, decreasing the likelihood that subjects would question whether the threat was genuine. To identify the town’s residents as noncombatants, all stories reported that “citizens of Daraya have not provided material aid or recruits to the terrorists.” All respondents were asked two “manipulation check” questions to determine whether they comprehended the treatments. Subjects who answered incorrectly were asked to reread the story. Over 85% of subjects answered correctly the first time and 100% answered correctly the second time. 48 The stories for the different national samples were identical with the exceptions of the relevant national actors and the threatened capital cities. See the online appendix. 47 KETTLES OF HAWKS 15 We selected a terrorist threat, rather than a state actor, because all four nations face threats of terrorism emanating from the Middle East, because we could more easily exclude the terrorist group’s ability to retaliate in kind, and because previous survey experiments on the nuclear taboo and noncombatant immunity have focused on terrorist targets. Although it is possible that subjects might be less willing to resort to nuclear weapons when the target is a nation-state, evidence from the experiments in the United States suggests that a majority of the American public is willing to use nuclear weapons against state adversaries, at least under certain circumstances.49 We selected Libya as the location for the terrorist facility because it provided the most plausible location from which a terrorist group might operate and strike against any of the four states and because Libya is within range of both the nuclear and conventional weapons systems of all four states. In addition, none of the four states surveyed had a manifestly adversarial or friendly relationship with Libya in 2018. We believe the scenario does not closely resemble any military operations any of the four countries conducted in recent years. Although we designed this scenario to describe a threat that would be as similar as possible for all four countries, we cannot entirely eliminate the possibility that citizens of different countries might perceive the terrorist threat differently given their unique histories and experiences with terrorism. The story states that political leaders are currently deciding how to respond to the report of the imminent terrorist attack. The news story features a leaked report, written by the country’s top military advisers, outlining two potential military options to preempt the attack. Subjects read that because the precise location of the underground bunker that the terrorists are using to assemble the weapons within the Libyan town is unknown, “destroying it will require either a large conventional strike or the use of a nuclear weapon.” The first option would use fifty conventionally armed cruise missiles to destroy the facility. The second option would use a single nuclear-tipped cruise missile. Because both strikes use pilotless cruise missiles, the report states that neither option would risk the lives of compatriot soldiers. The military report emphasizes that if no action is taken, the terrorists will carry out the attack, which is projected to kill three thousand civilian compatriots. The country “only has one shot at destroying the facility,” since the terrorists are likely to relocate if it is not destroyed in the 49 Sagan and Valentino, “Revisiting Hiroshima in Iran”; Alida R. Haworth, Scott D. Sagan, and Benjamin A. Valentino, “What Do Americans Really Think about Conflict with Nuclear North Korea? The Answer Is Both Reassuring and Disturbing,” Bulletin of the Atomic Scientists 75, no. 4 (July 2019): 179–86. 16 J. DILL ET AL. first strike.50 The article also states clearly that the “report does not recommend which option the president [or prime minister] should choose.” The first two treatment groups are designed to explore the prevalence and strength of categorical opposition to the use of nuclear weapons. In Condition 1, which we label “effectiveness 90/90,” the consequences of the nuclear and conventional strikes are identical. Each strike has a 90 percent probability of destroying the terrorist facility and preventing the planned attack. The report states that although the military “will attempt to minimize civilian loss of life” in both strikes, 2,700 Libyan civilians will be “unavoidably killed” as a “regrettable side-effect” of either strike, “including immediate deaths and deaths resulting from long term consequences” of the strikes. Thus, the only difference between the two strikes in this condition is the type of weapon used.51 Condition 2 (“effectiveness 90/45”) is identical to Condition 1 (“effectiveness 90/90”), except that the conventional strike is described as having only a 45% chance to destroy the terrorist target, whereas the nuclear option remains 90% effective. This condition, therefore, forces subjects to consider whether avoiding the use of nuclear weapons is more important than doubling the likelihood of preventing the terrorist attack. Condition 3 (“nuclear 100k deaths”) allows us to explore how subjects weigh the negative consequences of using nuclear weapons against foreign civilians against the positive benefit of nuclear weapons being more likely to destroy the target. This condition mirrors Condition 2 (“effectiveness 90/45”), except that the more effective nuclear strike now unavoidably kills 100,000 Libyan civilians instead of the estimated 2,700 deaths in the conventional strike. This condition is a test of how sensitive respondents are to rising foreign civilian casualties: Would they be willing to kill 100,000 foreign civilians to have a greater chance of saving 3,000 compatriot civilians? Condition 4 (“target civilians”) is designed to test the strength of the categorical noncombatant immunity norm. It mirrors Condition 1 (“effectiveness 90/90”), with both strikes estimated to be equally effective and Libyan civilian fatalities projected at 2,700 for either military option. The civilian fatalities resulting from both strikes in this condition, however, are explicitly described as intentional, rather than a “regrettable side-effect” as in Condition 1 (“effectiveness 90/90”). The article states that the attacks will “target the facility and surrounding civilian population” and that the military report concludes that by doing so, “both strikes would send a 50 In Conditions 1 (“effectiveness 90/90”), 2 (“effectiveness 90/45”), and 4 (“target civilians”), the nuclear strike is estimated to kill 2,700 Libyan civilians and has an estimated 90% probability of destroying the chemical weapons facility and preventing the terrorist attack. Thus, the nuclear strike is expected to save 2,700 compatriots. 51 Modern low-yield nuclear weapons create significantly reduced collateral damage. KETTLES OF HAWKS 17 Table 1. Details of experimental conditions. 3: Nuclear 100k deaths Treatment condition 1: Effectiveness 90/90 2: Effectiveness 90/45 Military effectiveness of nuclear and conventional strikes Estimated Libyan civilian deaths Intentional targeting of Libyan civilians? Compare condition with: 4: Target civilians Nuclear: 90% Nuclear: 90% Nuclear: 90% Conventional: 90% Conventional: 45% Conventional: 45% Nuclear: 90% Conventional: 90% Nuclear: 2,700 Nuclear: 2,700 Nuclear: 100k Conventional: 2,700 Conventional: 2,700 Conventional: 2,700 No No No Nuclear: 2,700 Conventional: 2,700 2 and 4 1 1 and 3 2 Yes strong message to terrorist sympathizers everywhere to reject ideologies of terror.”52 After reading the news stories, respondents were asked which of the two military options they would prefer if they had to choose. This basic design has been used in a variety of previous experiments on public opinion and the use of nuclear weapons. Requiring subjects to choose between two options—even if many subjects may find both choices distasteful—is the most objective way to elicit attitudes about tradeoffs between competing imperatives. This approach forms the foundation of the widely used “trolley car” experiments, which moral philosophers designed to assess moral intuitions about killing.53 To measure respondent attitudes in greater detail, however, respondents were also given the opportunity to rank order the options, including the option not to strike the target at all. Moreover, we asked respondents to indicate how ethical or unethical they considered each option to be (giving subjects the chance to rate both options unethical if they wished).54 To search for additional evidence on subjects’ reasoning, we asked subjects to explain their choice in their own words in openresponse questions. After responding to this series of questions (our primary dependent variables), subjects were asked a series of additional questions designed to 52 Some authors have suggested that the identity of the victims of war may influence decisions about civilian targeting. See Tanisha M. Fazal and Brooke C. Greene, “A Particular Difference: European Identity and Civilian Targeting,” British Journal of Political Science 45, no. 4 (October 2015): 829–51; Scott D. Sagan and Benjamin A. Valentino, “Weighing Lives in War: How National Identity Influences American Public Opinion about Foreign Civilian and Compatriot Fatalities,” Journal of Global Security Studies 5, no. 1 (January 2020): 25–43. To explore whether anti-Muslim sentiments influenced attitudes in the four countries, we also fielded a fifth condition, in which half of the residents who would be killed in the Libyan village were described as Christian. This treatment did not significantly change support for the nuclear strike. Results for this treatment are included in Figure 1A in Online Appendix 4. 53 David Edmonds, Would You Kill the Fat Man? The Trolley Problem and What Your Answer Tells Us about Right and Wrong (Princeton, NJ: Princeton University Press, 2014). 54 Though rating questions were asked on a 6- or 7-point interval scale, we present our results in binary form for ease of interpretation. Unless otherwise noted, our results remain consistent using either the interval or binary form. All question wordings are provided in Online Appendix 2. 18 J. DILL ET AL. Figure 2. Respondent preferences and ethical assessments of nuclear and conventional strikes. gauge their political beliefs, their compatriot partiality, and support for the death penalty to measure retributiveness. Subjects in the control group only received this second set of questions. The relevant details of each condition are summarized in Table 1. Note that the conditions are not fully crossed, so it is not possible to experimentally compare each condition to every other condition in the experiment. Results Four Consequentialist Publics The results presented in Figure 2 below show the percentages of subjects in each country who preferred the nuclear strike (Panel A) and who agreed that it would be ethical to launch the nuclear strike (Panel B) and the conventional strike (Panel C). In addition to the four countries, a fifth category (designated with an “x” symbol) represents the average of responses, aggregated across all countries. Our results provide little support for a widespread internalization of a categorical taboo against nuclear use. Although only 12% of respondents (averaging across the four countries) preferred the nuclear strike when the consequences of nuclear and conventional weapons were described as equal (Condition 1, Panel A), subjects’ willingness to use nuclear weapons KETTLES OF HAWKS 19 increased over four times (to 56%) when the article reported that the nuclear strike would be twice as likely as the conventional strike to destroy the terrorist target (Condition 2). These increases were strongly statistically significant in every country. Indeed, a majority of the public in every country preferred the nuclear strike in Condition 2 except in the UK, where 48% preferred it. This sensitivity to military effectiveness is not consistent with an internalized categorical taboo on the use of nuclear weapons and therefore does not support H1a (nuclear taboo). Rather, the greater willingness of subjects to support the use of nuclear weapons when they are more likely to achieve a military objective suggests most respondents view the decision to use nuclear weapons in consequentialist terms, in support of H3 (nuclear consequentialism). When subjects were asked to make normative assessments about the strikes, a more complicated picture emerged, suggesting that many subjects do see nuclear weapons as raising unique ethical concerns. Subjects in every country and condition rated the conventional strike as more ethical than the nuclear strike, usually by a very substantial margin. This result reflects the fact that most subjects (72.4%) who preferred the nuclear strike also judged the conventional strike to be ethical, whereas only 22.5% of subjects who preferred the conventional strike considered the nuclear strike ethical. Indeed, in the UK in Condition 2, only 29% deemed the nuclear strike ethical, even though 48% preferred it. These results suggest that many subjects acknowledge that using nuclear weapons constitutes an ethical transgression, but this concern is not enough to overcome their desire to minimize the risks of harm to their compatriots by maximizing the strike’s military effectiveness.55 Nevertheless, although subjects always judged the conventional strike as more ethical, judgments about the ethics of the nuclear strike were still sensitive to concerns about military effectiveness, providing no support for H1b (nuclear taboo ethics). On average, subjects rated the nuclear strike significantly more ethical in Condition 2 (48%) than in Condition 1 (34%, Panel B), suggesting that normative concerns also follow a consequentialist rather than a categorical logic. This pattern can also be seen by comparing subjects’ relative ratings of the conventional and nuclear strikes across the two conditions. Whereas 41.9% more subjects deemed conventional weapons more ethical than nuclear weapons in Condition 1, that difference fell by more than half (to 19.4%) in Condition 2. The broad consequentialist pattern in the responses is reinforced by the results from Condition 3, in which subjects read that nuclear weapons 55 As noted above, this does not imply that subjects followed interests instead of norms. There are a variety of reasons why a subject’s preferences might diverge from one or more norms that they consider salient. The overriding desire to keep compatriots safe, for instance, could itself be a normative concern. 20 J. DILL ET AL. Figure 3. Respondent preferences for not striking terrorist target. would be twice as effective in destroying the terrorist target (identical to Condition 2) but would now kill 100,000 Libyan civilians instead of the 2,700 who would die in the conventional strike. Providing strong support for H4 (civilian consequentialist), subjects appeared to weigh the increased effectiveness of nuclear weapons against the increased fatalities they produced. The four-country average preference for the nuclear strike (Panel A) declined dramatically from 56% in Condition 2 to 30% in Condition 3. The declines were strongly statistically significant for each of the countries. The comparison between Conditions 1 and 4 represents our main test of the internalization and salience of the categorical norm against intentionally killing civilians. In both conditions, the nuclear and conventional strikes were described as equally effective. In Condition 1, however, the civilian fatalities expected in both strikes were described as “unavoidable” and a “regrettable side-effect” of the attack on the military target. In contrast, in Condition 4, subjects read that both the nuclear and conventional strikes would intentionally target local civilians to “send a message” to terrorist supporters. We asked subjects to rank order their preferences for three options—launching the nuclear strike, launching the conventional strike, and not striking the target at all. Subjects who have internalized that deliberately targeting civilians is never acceptable should have preferred not launching any attack at all in Condition 4, and they should have been much more likely to prefer the no-attack option in Condition 4 than in Condition 1. KETTLES OF HAWKS 21 Figure 3 shows the percentage of subjects who ranked the “do not attack” option first in Conditions 1 and 4. Averaging across all four countries, we found that subjects in Condition 4 were slightly, but statistically significantly, more likely (4.7%) to rank not striking as their most preferred option than were subjects in Condition 1. It is important to note, however, that the disaggregated results reveal that the internalization of the categorical prohibition on deliberate attacks against civilians appears to be limited largely to UK citizens. British respondents’ preferences for “no strike” increased by 11% between Conditions 1 and 4, whereas no other country changed by more than three points. Moreover, despite this change between conditions, most subjects across countries (88.2%), including the UK (82%), still preferred some form of military attack, even when the strikes intentionally targeted civilians. These results suggest there is no widespread internalization of a categorical norm against targeting civilians and thus little support for H2a (noncombatant immunity). Subjects do, however, rate the intentional nuclear attack against civilians in Condition 4 as slightly, but significantly, less ethical than the strike that kills civilians unintentionally in Condition 1 (34% to 28%).56 Although ethical ratings decline for citizens of each country, the effect is small and significant only for the four-country average, providing only modest support for H2b (noncombatant immunity ethics).57 To gain a deeper understanding of the logic behind respondents’ choices, we asked all subjects to briefly describe in their own words “the single most important reason they preferred” the nuclear or conventional strike. Subjects who ranked not striking at all first when presented with all three options were additionally asked to describe the most important reason for that preference. Using a team of three human coders, we systematically coded all responses for whether they expressed reasons consistent with the nuclear taboo or civilian immunity.58 The open-ended responses provide little indication of categorical reasoning in any of the four countries. Among people who opposed using nuclear weapons, very few subjects referenced a taboo logic, with less than 4%, on average, expressing a categorical rejection of nuclear weapons. Subjects most frequently explained their opposition to nuclear weapons using taboo rationales in Condition 2, in which nuclear weapons were most attractive on consequentialist terms. Even in this condition, though, only 6.3% of respondents expressed reasoning compatible with a nuclear taboo. 56 The four-country average ethical rating for the conventional strike was also lower in Condition 4 (“target civilians”) than Condition 1 (“effectiveness 90/90”) (p ¼.07). 57 The decline for the UK between Conditions 1 (“effectiveness 90/90”) and 4 (“target civilians”) is significant at p ¼.09. 58 Coding procedures for open-ended responses are described in Online Appendix 3. 22 J. DILL ET AL. Instead, subjects were much more likely to reference one of three main consequentialist reasons for opposing nuclear weapons use. In Condition 3, in which the nuclear strike would kill many more civilians, 69.5% of subjects cited the higher civilian casualties of the nuclear strike as the primary reason they preferred the conventional attack. Concern about the effects of radiation was the single most common reason expressed in the open responses among subjects who preferred the conventional strike in Conditions 1, 2, and 4 (38.4%, 38.0%, and 39.4%, respectively). This suggests that fears about nuclear weapons use are produced by perceptions about unique radiological consequences, a finding supported by research on nuclear waste.59 A significant group of subjects in Conditions 1, 2, and 4 (9.7%, 15.7%, and 12.2%, respectively) also referenced the fear that using nuclear weapons might set a precedent that would lead to their use against their own country. Together, these three consequences—civilian casualties, radiation, and precedent setting—accounted for over 60% of responses in every condition. We also found little evidence of categorical reasoning about attacks on civilians when we examined the open responses from subjects in Conditions 1 and 4 who preferred not striking the terrorist facility at all. Of subjects who preferred not to strike the terrorist facility, 8.6% cited the idea that targeting civilians was wrong as the reason for preferring no strike in Condition 1, whereas 12.8% cited it in Condition 4. This difference, however, was not statistically significant. In sum, respondents’ attitudes across conditions and for all outcome variables (nuclear preference, nuclear ethics, and no-strike preference) follow largely a consequentialist logic, with little evidence that large portions of the public in any country have internalized or consider salient categorical norms against nuclear use or deliberately targeting civilians. Compatriot Partiality and Retributiveness As mentioned, consequentialists may have different substantive preferences regarding nuclear use and killing civilians depending on how they weigh the consequences of using force. We examined subjects’ compatriot partiality (H5a) and retributiveness (H5b) as potential explanations for individual variation in hawkishness in our experiments. To measure the strength of compatriot partiality we asked all subjects to use standard 100-point “feeling thermometers” to rate their levels of “warmth” or “coldness” toward their own country and toward Libya, where the terrorist facility in our scenario was located and whose civilians would 59 Paul Slovic, James H. Flynn, and Mark Layman, “Perceived Risk, Trust, and the Politics of Nuclear Waste,” Science 254, no. 5038 (13 December 1991): 1603–7. KETTLES OF HAWKS 23 Figure 4. Compatriot partiality, retributiveness, and nuclear strike preferences. be the victims of the strikes.60 We then subtracted subjects’ Libya feeling thermometer score from their “own country” score. Figure 4, Panel A shows how subjects (pooled across all four countries) who rated their own country at least fifty points higher than Libya compare to subjects with lower relative ratings of their own country on their preferences for using nuclear weapons.61 Because compatriot partiality is likely correlated with several other relevant individual-level attributes, the results in Figure 4 are adjusted after controlling for measures of age, gender, education, political conservatism, and religiosity. Subjects who scored high on our compatriot partiality measure were more likely to prefer using nuclear weapons in three of the four conditions (although the effect was only significant in Conditions 2 and 3), providing 60 Due to an error involving the way the feeling thermometer questions rendered on mobile devices in Israel, 46.5% of Israeli observations for these questions were missing. Although missingness was not correlated with our dependent variables, it was more common among women, subjects under sixty years of age, and those with no college education. To help account for this issue, YouGov provided separate survey weights using only subjects who answered the feeling thermometer questions. Our results remain consistent and we use the revised weights in all analyses that include feeling thermometer scores. In addition, as a robustness check, we used an alternate measure of compatriot partiality (with less than 1% missingness in all countries) based on the following question: “How much do you agree or disagree with the following statements? The world would be a better place if people from other countries were more like people from [my country].” Our results remained consistent using this measure. 61 Averaging across all four countries, 41% of subjects scored above 50 and 59% scored below 50. We use this dichotomous measure for ease of interpretation. The patterns remain consistent when using the continuous version of the feeling thermometer variable. 24 J. DILL ET AL. at least some support for H5a. In addition, we found evidence that highly partial subjects responded differently to the increasing military effectiveness of nuclear weapons between Conditions 1 and 2. Highly partial subjects increased their support for the nuclear strike by 53 percentage points across the two conditions, compared to 38 points for less partial subjects.62 Subjects who place a higher value on their own compatriots, in other words, were more willing to use nuclear weapons to increase the chances of protecting those compatriots from attack. We found no significant interactions between compatriot partiality and Conditions 2 and 3, in which we varied the number of Libyan civilians killed by the nuclear strike. One potential explanation for this result is that, at this very high level of foreign civilian fatalities, even subjects who place a high value on compatriots are equally swayed by the sheer number of deaths on the other side. Nevertheless, compatriot partiality was evident in many respondents’ explanations of their preferences for using nuclear weapons in Condition 3. One American, for example, wrote: “Unfortunately, loss of live [sic] in Libya will be terrible but it’s us or them. Sorry, but I’ll take ‘them.’” In Israel, one respondent acknowledged that “we need to prevent as much as possible the killing of innocent people,” but nevertheless asserted that Israel must place the “lives of our civilians before the lives of our enemies.” Two British respondents wrote: “The first duty of a government is the safety of its citizens,” and “I value my country and its safety more than anything else in this world.” Finally, one French citizen who preferred the nuclear option killing 100,000 Libyans justified the choice by the obligation to “protect our compatriots from barbarism.” Many respondents who favored the conventional attack in Condition 3 also focused on the attack’s consequences but judged that the nuclear attack killed too many Libyan civilians. One American’s explanation for preferring the conventional option suggested an explicitly cosmopolitan utilitarian view: “100,000 dead people is more than 2,700. The probability of the mission successfully removing the threat of chemical weapons may be higher, but the cost is much greater. Even if the U.S. fails to destroy the facility and the chemical attack happens which kill[s] 3,000 Americans, 100,000 PEOPLE dead is still a larger number than 5,700.” In France, one respondent wrote: “The conventional strike would kill 2,700 Libyan civilians to save 3,000 French civilians which is fair, even if it is still horrible, whereas the nuclear strike affects 100,000 civilians which is far too many,” and another maintained that “a nuclear strike is disproportionate and will kill a large number of civilians.” In Israel, one respondent who preferred the conventional strike explained that “I don’t think that [saving] 3,000 Israelis 62 The interaction between compatriot partiality and condition was significant at p ¼.037. KETTLES OF HAWKS 25 equals the killing of 100,000 people from Libya.” In the UK, one respondent wrote: “If 2,700 odd Libyans were killed and the strike failed then the 3,000 UK citizens that were killed afterwards would still be far less damage than the nuclear strike.” We also found evidence that differences in retributiveness drive some variation in subjects’ hawkishness. To measure retributiveness, we used respondents’ support for the death penalty for convicted murderers, as previous studies have done. Our conditions were not designed to vary the degree to which the use of nuclear weapons might be seen as retribution or vengeance, but Figure 4, Panel B shows how subjects’ (pooled across all four countries) preferences for the nuclear strike compare between those who approve or oppose the death penalty.63 In support for H5b, subjects who approve of the death penalty were significantly more likely to prefer the nuclear attack in all conditions except Condition 4.64 Although the majority of nuclear hawks explained their preferences on military effectiveness or deterrence grounds, the presence of vengeful or retributive comments in some of the nuclear hawks’ explanations across different conditions suggests retribution or revenge was an important motive for some respondents.65 In Condition 1, a British respondent wanted to “leave the long lasting pain that nuclear weapons will have on the area, as this is basically what they’re planning on London.” A French respondent simply wanted to produce “more suffering.” In Condition 3, one French hawk complained that “there would be a 90% chance of killing the terrorists and most of the civilians support them in their actions,” and one American hawk justified the nuclear attack because “half of the people were cheering them on in Libya.” One Israeli said: “Whoever messes with us and wants to harm us [should be killed].” In Condition 4, one American who preferred the nuclear strike explained, “If a terrorist wants to attack the us [sic] they deserve the full might of the USA.” One British citizen simply wrote: “Fight fire with fire.” Cross-National Variation Regarding patterns of cross-national differences, first, and most significantly, we find no support for H6 (American exceptionalism). Averaging across the four conditions, Americans’ preferences for nuclear weapons (25.5%) were significantly lower than Israelis’ (32%), significantly higher than UK citizens’ (20.7%), and indistinguishable from French preferences 63 Again, we control for age, gender, education, political conservatism, and religiosity. The difference between Conditions 1 and 4 is significant at p ¼.08. 65 For a discussion of the distinction between reciprocity and revenge and their role in American public opinion on the use of force, see Scott D. Sagan and Benjamin A. Valentino, “On Reciprocity, Revenge, and Replication: A Rejoinder to Walzer, McMahan, and Keohane,” Ethics & International Affairs 33, no. 4 (Winter 2019): 473–79. 64 26 J. DILL ET AL. (28.3%). Importantly, we found no significant interactive effects between nationality and the relevant conditions for nuclear preferences, nuclear ethical ratings, or preferences for not striking the target at all. In other words, citizens from different countries did not respond differently to variations in military effectiveness, civilian fatalities, or violations of noncombatant immunity. This strongly suggests the cross-national differences we observe are not due to the differential internalization of a nuclear taboo or categorical norms against targeting civilians. Indeed, we found a remarkably stable pattern of responses between countries. Across all four conditions, citizens of the UK were the least willing to use nuclear weapons, the most likely to register ethical reservations about a nuclear attack, and among the most likely to prefer not attacking at all. Israelis, on the other hand, were almost always the most willing to use nuclear weapons, the most likely to say that doing so was ethical, and the least likely to prefer not striking at all. Israelis were statistically significantly more likely than British citizens to prefer nuclear weapons in every condition except Condition 4, more likely to judge the nuclear option as ethical in every condition, and less likely to prefer not striking in every condition except Condition 3. American and French citizens usually fell below Israel and above Britain and are statistically indistinguishable from one another in every condition on all three measures. These results provide strong support for H7 (Israeli exceptionalism) but cast doubt on explanations for nuclear attitudes that focus on America’s nuclear history or its uniquely retributive national culture. To revise Kagan’s metaphor, it appears that Israelis are from Mars and British citizens are from Venus, whereas the French and Americans are from somewhere in between (perhaps here on Earth). Our experiment was not designed to test the specific causes of cross-national variation, apart from differences in the internalization of categorical prohibitions.66 Nevertheless, our results do provide some suggestive evidence that retributiveness and compatriot partiality may play a role in explaining the cross-national patterns we observe. Patterns of compatriot partiality and death penalty support roughly track the hawkish Israeli and dovish British results we observe. Israelis exhibited by far the highest degree of compatriot partiality, with 74.6% of Israelis scoring their country at least fifty points higher than Libya, compared to 43.9% 66 Using causal mediation analysis to measure the indirect effects of citizenship on hawkish preferences through other mediating variables was not possible, since this would violate the sequential ignorability assumption required for causal mediation analyses (see Avidit Acharya, Matthew Blackwell, and Maya Sen, “Analyzing Causal Mechanisms in Survey Experiments,” Political Analysis 26, no. 4 (October 2018): 357–78). In particular, since potential mediators such as compatriot partiality or retributiveness were not randomly assigned, we cannot assume that there are no unmeasured confounders for the relationship between these mediators and military strike preferences. KETTLES OF HAWKS 27 Figure 5. Marginal effects of experimental conditions and respondent attributes on preference for the nuclear strike. in the United States, 37.2% in the UK, and 23.2% in France. On our measure of retributiveness, 70.2% of Israelis supported the death penalty, compared to 63.3% in the United States, 61.4% in France, and 54.6% in the UK. Variation in the degree to which subjects fear a terrorist attack, on the other hand, does not appear to track hawkishness closely. This casts doubt on explanations for Israeli exceptionalism that focus on Israelis’ unique history with terrorism. Although 60.1% of Israelis deemed a terrorist attack against their homeland in the next ten years that kills more than one thousand civilians somewhat or very likely, so did 58.7% in the UK, and even greater numbers did so in America (69.2%) and France (73.7%).67 67 When including the expectation of terrorism in models in Figure 5, described below, it is positively correlated with preferences for and ethical ratings of the nuclear strike, but our results remain consistent. The relationship with nuclear preference, however, is not significant. 28 J. DILL ET AL. To examine further the connection between individual-level variation and cross-national differences, Figure 5 presents the marginal effects for nuclear preference by country, age, sex, religiosity, college education, political conservatism, support for the death penalty, and compatriot partiality on subjects’ preferences for and normative assessment of the nuclear strike. We also include variables for each experimental condition to show the relative sizes of the treatment effects. The United States and Condition 1 serve as baseline categories in these models, so the marginal effects of the country and condition variables in Figure 5 represent the relative effect of each country compared to the United States and each condition compared to Condition 1. After controlling for these individual-level variables, the nuclear preferences of the US population are statistically indistinguishable from those of British citizens. France and Israel remain more likely to prefer nuclear weapons than the UK, but the differences are relatively small. The differences between the countries on ethical ratings remain much larger, however, suggesting that the factors in our model explain less of the variation between countries on this measure. Variables representing standard demographic characteristics such as age, sex, and education have relatively small, and mostly statistically insignificant, effects for both nuclear preferences and ethical assessments. Political conservativism, support for the death penalty, and high compatriot partiality have much stronger effects, increasing the probability of preferring nuclear weapons and rating the nuclear strike as ethical by between 6% and 16%. Implications for Policy and Further Research American attitudes toward nuclear weapons and noncombatant immunity are not exceptional; citizens of other nuclear-armed democracies do not hold categorical prohibitions on using nuclear weapons or targeting civilians either. This article has demonstrated that, contrary to many scholars’ expectations, the publics of two other nuclear-armed democracies are equally willing (France) as or more willing (Israel) than Americans to use nuclear weapons and deliberately target civilians. Only the British public shows slightly greater reluctance to violate categorical prohibitions on nuclear use and targeting civilians. In all four countries, public opinion largely follows a consequentialist logic: support for nuclear use increases the more effective the strike is and decreases the higher the projected foreign civilian casualties, regardless of whether they are inflicted intentionally. At the same time, we find a remarkably consistent pattern of cross-national differences in hawkishness. These differences are not due to divergently internalized taboos or categorical prohibitions, however. Instead, we KETTLES OF HAWKS 29 identified two potential causes of support for nuclear use and killing civilians, particularly in Israel: high compatriot partiality and high degrees of retributive beliefs. Our findings have important implications for a central debate in IR theory. Mainstream rationalist theories of IR portray all states as driven by the same logic of reasoning: they minimize costs and maximize the projected benefits of action, often simply defined in terms of power or prosperity. Constructivist theories, in contrast, allow for the internalization of norms to influence states’ logic of reasoning, and their perceptions of what constitutes a cost and a benefit. In our study, rationalist theories are half right. Public opinion in different states is driven by the same logic: a cost–benefit analysis of consequences, largely unconstrained by categorical norms. However, mainstream theories are wrong to assume that publics all maximize the same goals or utilities. In fact, Israelis perceive the costs and benefits of using military force quite differently than British citizens. Our findings raise four questions for future research. First, our study was designed to detect cross-national differences in the willingness to use nuclear weapons and target civilians, and to test whether any observed differences were driven by differences in the degree of consequentialist versus categorical reasoning. Although we show that compatriot partiality and retributiveness help explain individual variation, we can only offer suggestive evidence for their role in explaining why Israelis are so much more hawkish than the British. Future cross-national studies could experimentally manipulate opportunities for retribution (by varying, for example, whether victims supported the terrorists in the past) and the compatriot status of expected friendly civilian casualties (allies or national compatriots) to confirm these findings. Second, further research could shed light on the origins of compatriot partiality (and its relationship to other measures of nationalism) and the causes of variation in retributive beliefs. We show that support for the death penalty is stronger than previously known even in abolitionist democracies, but we need to deepen our understanding of the role of history, racism, and domestic politics in promoting retributiveness.68 Why and how individuals in different countries move from supporting retribution (punishing the guilty in the name of justice) to punishing uninvolved foreign civilians for terrorist threats is also not well understood. Third, although there are practical challenges to polling in nondemocracies, understanding the attitudes of citizens in nuclear-armed autocracies is also critical. 68 See David Garland, Peculiar Institution: America’s Death Penalty in an Age of Abolition (Cambridge, MA: Belknap Press of Harvard University Press, 2012); James D. Unnever, Francis T. Cullen, and Cheryl Lero Jonson, “Race, Racism, and Support for Capital Punishment,” Crime and Justice 37 (2008): 45–96. 30 J. DILL ET AL. Finally, we need more research on how individuals in different countries think about tradeoffs between violations of conflicting norms. That we find little evidence of an international nuclear taboo does not mean that citizens of nuclear-armed democracies consider nuclear use ethically unproblematic. Rather, their ethical judgements also follow a consequentialist rather than categorical logic. Moreover, many respondents appear willing to override their normative concerns when not using nuclear weapons means risking the lives of compatriot civilians. In Condition 2, for example, where the conflict between the imperatives to protect compatriots and to avoid the use of nuclear weapons was strongest, 56% of all respondents preferred the nuclear strike, and 32% (averaging across countries) who judged the nuclear strike to be unethical preferred it anyway. Indeed, when asked if they agreed that their country “must protect its most vital interests even if that sometimes requires doing things that are morally wrong,” 52.9% of respondents across all countries and conditions said yes.69 These results significantly enrich our understanding of prior works’ recurring finding that democratic publics are willing to kill large numbers of foreign noncombatants (both intentionally and incidentally) to reduce compatriot soldier deaths and help win wars.70 More comparative work on this critical subject is needed. Our findings also have two important nuclear policy implications. First, they suggest an answer to the puzzle of nuclear nonuse. If the political elites responsible for the decision to use nuclear weapons reason as their publics do, the nonuse of nuclear weapons since 1945 has likely been driven by consequentialist calculations about the prospects for retaliation, the relative effectiveness of nuclear weapons compared to conventional weapons, and a fear about setting a dangerous precedent. Second, our findings are instructive for how we can best prevent nuclear use in the future. Antinuclear activists have expended considerable energy on strengthening the perceived public taboo on nuclear weapons use. The 2017 Treaty on the Prohibition of Nuclear Weapons, for example, bans nuclear strikes on the grounds that there is a categorical difference between conventional and nuclear weapons, casting the latter as “abhorrent to the principles of humanity and the dictates of public conscience.”71 Our research suggests that to mobilize public opinion against these weapons, efforts might be better spent not appealing to the better angels of our nature, but instead educating the public on the consequences of nuclear use, including the 69 Even in the control groups, 46% of subjects agreed with this statement. Alexander B. Downes, Targeting Civilians in War (Ithaca, NY: Cornell University Press, 2008); Christopher Gelpi, Peter D. Feaver, and Jason Reifler, Paying the Human Costs of War: American Public Opinion and Casualties in Military Conflicts (Princeton, NJ: Princeton University Press, 2009). 71 Treaty on the Prohibition of Nuclear Weapons, 20 September 2017, United Nations Office of Disarmament Affairs, https://treaties.un.org/doc/Treaties/2017/07/20170707%2003-42%20PM/Ch_XXVI_9.pdf. 70 KETTLES OF HAWKS 31 possibility that nuclear use against adversaries will create a precedent that will increase the likelihood of future nuclear attacks on themselves and their compatriots. Acknowledgments For helpful comments, we thank the participants of workshops at the Department of Political Science at Duke University, the School of Political Science at Haifa University, the Belfer Center at Harvard University, the Hertie School, the Institute for National Security Studies Tel Aviv, and the Security Studies program at MIT, the James Martin Center for Nonproliferation Studies at the Middlebury Institute for International Studies at Monterey, the International Security Center at University of Notre Dame, the University of Pennsylvania, the Royal United Service Institute, and the Center for International Security and Cooperation at Stanford University. We thank members of panels at the annual conventions of the American Political Science Association in 2018 and the International Studies Association in 2018. For exceptional research assistance we thank Chelsea Green, Alida Haworth, Katherine McKinney, and Nitish Vaidyanathan. Funding We also thank the John D. and Catherine T. MacArthur Foundation, Nuffield College Oxford, and the Stanton Foundation for their financial support for this project. ORCID Janina Dill http://orcid.org/0000-0002-5872-5018 Data availability statement The data and materials that support the findings of this study are available in the Security Studies Dataverse at https://doi.org/10.7910/DVN/PAY8IA.