Hostname: page-component-7479d7b7d-68ccn Total loading time: 0 Render date: 2024-07-15T19:16:28.006Z Has data issue: false hasContentIssue false

Condemning or Condoning the Perpetrators? International Humanitarian Law and Attitudes Toward Wartime Violence

Published online by Cambridge University Press:  11 March 2019

Rights & Permissions [Opens in a new window]


What are the implications of international law for attitudes toward wartime violence? Existing research offers contrasting views on the ability of international legal principles to shape individual preferences, especially in difficult situations involving armed conflict. Employing cross-national survey evidence from several conflict and post-conflict countries, this article contributes to this debate by evaluating the relationship between individuals’ knowledge of the laws of war and attitudes toward wartime conduct. Findings show that exposure to international law is associated with a significant reduction in support for wartime abuses, though the results are stronger for prisoner treatment than for targeting civilians. Analysis further reveals that legal principles generate different expectations of conduct than alternative value systems that are rooted in strong moral foundations regarding the impermissibility of wartime abuses. The findings are relevant for understanding the relationship between international law and domestic actors, and how legal principles relate to the resort to violence.

Copyright © 2019 American Bar Foundation 


Can international laws regulating wartime conduct promote more humanitarian views among populations experiencing recent or ongoing armed conflict? Indeed, a central belief underlying instruments like the 1949 Geneva Conventions is that the dissemination of knowledge about the laws of war can play a central role in improving compliance during armed conflict.Footnote 1 Even if international law may offer salutary effects in some circumstances, one long-held view instead suggests that such agreements yield few, if any, benefits when core security issues are at stake (Mearsheimer Reference Mearsheimer1994–1995; Valentino, Huth, and Croco Reference Valentino, Huth and Croco2006). By contrast, other work points to the ability of international law to alter preferences and policies on controversial questions even when they involve the use of violence or the protection of human rights, though sometimes with important qualifications (Simmons Reference Simmons2009; Conrad and Ritter Reference Conrad and Ritter2013; Y. Lupu Reference Lupu2013; Fariss Reference Fariss2014).

Societies convulsed in, or emerging from, violence nevertheless represent a particularly difficult test for the potential value of international legal principles. Recent research does show that personally experiencing violence can lead to greater political and civic participation and, ultimately, a stronger sense of social cohesion (Bellows and Miguel Reference Bellows and Miguel2009; Blattman Reference Blattman2009; Voors et al. Reference Voors, Nillesen, Verwimp, Bulte, Lensink and van Soest2012; Gilligan, Pasquale, and Samii Reference Gilligan, Pasquale and Samii2014). However, much of this work focuses on ties among local communities rather than attitudes toward out-groups, where the latter sorts of intergroup relations are more pressing for questions about wartime conduct. The record of past fighting, as well as threats both real and perceived, provides a more somber accounting of individual attitudes toward violence and intergroup relations (Huddy et al. Reference Huddy, Feldman, Taber and Lahav2005; Grossman, Manekin, and Miodownik Reference Grossman, Manekin and Miodownik2015). The consequences of violence can be short term, but they can also have much more ingrained legacies lasting across generations (Balcells Reference Balcells2012; Child and Nikolova forthcoming; N. Lupu and Peisakhin Reference Lupu and Peisakhin2017).

Understanding mass attitudes toward violence is critical because many forms of abuse have at their core micro-level underpinnings (Kalyvas Reference Kalyvas2006). Far from simply being powerless observers or victims of warfare, civilians have been some of the most vocal proponents for brutal acts of violence (Goldhagen Reference Goldhagen1996; Reiter and Stam Reference Reiter and Stam2002, 158). Yet, in other instances segments of the public have proved crucial in restraining belligerents from resorting to extreme levels of violence (Merom Reference Merom2003, 19). Even accounts emphasizing the central importance of leaders and other elites acknowledge that such arguments are predicated on populations willing to condone, or at least turn a blind eye to, killings and other forms of abuse (Valentino Reference Valentino2004, 31). Ordinary people also often play key roles in the perpetration of violence separate from, or loosely in conjunction with, wider elites (Browning Reference Browning1998; Kalyvas Reference Kalyvas2006; Straus Reference Straus2006). It follows that two areas with some of the worst records for wartime compliance—the treatment of civilians and the treatment of prisoners—are also those where violations by individuals and small groups loom largest (Morrow Reference Morrow2007).

A comparable “bottom-up” turn is evident in the empirical study of international law with the incorporation of more ordinary domestic actors alongside the traditional focus on political and judicial elites (Simmons Reference Simmons2010; Linos Reference Linos2013). International law can introduce citizens to new information and ideas, offer greater legitimacy for emerging beliefs, or encourage mobilization to pressure governments and other domestic groups to alter their policies (Thomas Reference Thomas2002, 3; Dai Reference Dai2007, 90; Simmons Reference Simmons2009, 150–54). Yet reliable evidence disentangling some of these varied mechanisms linking international law to preferences and outcomes remains sparse (Simmons Reference Simmons2010).

Several recent experimental studies have sought to identify the impact of international law on mass attitudes (Wallace Reference Wallace2013; Chaudoin Reference Chaudoin2014; Chilton Reference Chilton2015). Although promising, such work continues to raise questions for fully understanding the individual-level consequences of international law due to the use of hypothetical vignettes, the one-shot nature of treatments, the short time lapse when measuring key outcomes, and challenges in assessing mechanisms for how international law actually has an impact. Moreover, much of this existing public opinion work has overwhelmingly focused on the United States, leaving open the question of whether international law operates in a similar manner in other countries.Footnote 2

This article seeks to bridge and build on existing micro-level approaches to the study of both international law and political violence. I examine the extent to which real-world exposure to international laws governing armed conflict is associated with any differences in individual attitudes toward conduct during war. Moving beyond a single-country design, I draw on evidence from a cross-national survey of eleven countries that was conducted by the International Committee of the Red Cross (ICRC) in 1999 as part of a larger study on the state of the laws of war (ICRC 1999). Despite being fielded a number of years ago, the ICRC survey is unique in the breadth of questions on attitudes toward international law and wartime violence, as well as providing the most comprehensive basis for comparison across a wider range of national contexts. Using survey items tapping into multiple distinct dimensions of wartime conduct, I construct a series of indices capturing individuals’ overall feelings about what sorts of actions should, or should not, be allowed in the midst of armed conflict.

The findings indicate that individuals’ knowledge of international law is associated with significant differences in their attitudes toward wartime violence, though with important qualifications. While an individual’s prior understanding of the core tenets of international humanitarian lawFootnote 3 (hereafter IHL) is correlated with a general decline in approval of abuses, the findings are strongest for the treatment of prisoners. Violence directed toward civilians instead shows little difference in support from those with a greater appreciation of international legal principles. On the other hand, the role of international law indicates comparable, and sometimes even more marked, changes in attitudes even for those groups that are more prone toward condoning violence or more distrustful in general of international legal commitments. The findings thus highlight the importance for empirical legal scholarship of taking into account the contextualized and conditional relationships between international law and individual attitudes, especially around charged issues like wartime violence.

This study also addresses calls to develop a better understanding of how international law shapes preferences and outcomes. The specific mechanisms through which international legal considerations influence outcomes remain hotly debated, with distinct contending theoretical perspectives put forward (Simmons Reference Simmons2010). Additionally, I offer evidence that individuals exposed to international law prioritize legal norms over alternative value systems, and that these principles exhibit strong moral foundations when forming policy attitudes rather than more instrumental considerations. Although mindful of the difficulties of directly establishing a causal relationship, the findings point to the potential, as well as limits, of international law in shaping attitudes toward wartime violence, alongside the importance of taking into account preexisting individual beliefs.

The article proceeds as follows. Section II offers a brief background on the laws of war and lays out theoretical expectations regarding the relationship between international law and mass attitudes. Section III describes the survey data and how the main variables are constructed. In Section IV, I present the empirical analysis, assess conditional relationships, and offer tests for specific underlying mechanisms, along with considering several likely threats to the validity of the overall results. The final section concludes by discussing the implications of the findings and making suggestions for further research.


Numerous international conventions have developed to regulate the conduct of armed actors during war (Howard, Andreopoulos, and Shulman Reference Howard, Andreopoulos and Shulman1994). IHL, or the laws of war, seeks to place certain limits on violence during armed conflicts. One of the most fundamental principles concerns the protection of certain groups, in particular civilians and prisoners. The 1949 Geneva Conventions and subsequent 1977 Additional Protocols outline many of the rights accorded to protected groups during war, alongside corresponding responsibilities for armed actors.Footnote 4 For instance, Article 27 of the Fourth Convention regulating the treatment of civilians states: “They shall at all times be humanely treated, and shall be protected especially against all acts of violence or threats thereof and against insults and public curiosity.” Similarly, Article 13 of the Third Convention dealing with soldiers who are captured declares: “Prisoners of war must at all times be humanely treated. Any unlawful act or omission by the Detaining Power causing death or seriously endangering the health of a prisoner of war in its custody is prohibited.” Both conventions, in addition to their first and second corollaries dealing with the sick and wounded on land and at sea, elaborate in detail the various rights and responsibilities of belligerents. The Geneva Conventions and related agreements thus codified and extended many earlier rules and customs regulating wartime conduct.

Although a great deal of attention has been paid to the development of the modern laws of war (Best Reference Best1980, Reference Best1994), much less settled is whether these rules actually restrain abuses (Evangelista and Tannenwald Reference Evangelista and Tannenwald2017). Several studies show that belligerents who have ratified the prevailing laws of war differ little in their treatment of civilians or prisoners during interstate wars from their counterparts who have refused to make comparable legal commitments (Valentino, Huth, and Croco Reference Valentino, Huth and Croco2006; Wallace Reference Wallace2012). Similarly, attempts by groups such as the ICRC to promote the laws of war show a similar questionable impact on the decisions of belligerents to target civilians during civil wars (Bussman and Schneider Reference Bussman and Schneider2016).Footnote 5 Other work suggests that the functioning of international treaties is conditional on reciprocity, in terms of both the legal commitment and compliance behavior by opponents. Yet even in these circumstances, adherence is generally much worse for the treatment of civilians or prisoners than in other areas (Morrow Reference Morrow2007).

These contested findings point to a frequently espoused view that international law plays little role in protecting those most vulnerable when core security interests are at stake. A long line of research demonstrates that when faced with severe military threats, governments often respond with higher levels of repression and violence both at home and abroad (Poe and Tate Reference Poe and Tate1994; Rummel Reference Rummel1994). Even the commitment to protect individual rights, often considered one of the pillars shared by democracies, largely falls by the wayside when democratic governments are confronted by grave threats (Davenport Reference Davenport2007, 173–74). The particular compulsion to reject international commitments similarly tends to increase dramatically when governments are dealing with armed challenges (Cardenas Reference Cardenas2007, 12).

The willingness to resort to abuse is not unique to political leaders and other elites, but also often finds strong agreement among their wider populations. Levels of political tolerance and acceptance of out-groups tend to worsen considerably when the public feels a large sense of threat (Merolla and Zechmeister Reference Merolla and Zechmeister2009, 76–78). Electoral support for right-wing political parties also tends to be higher when voters face more extreme threats to their security (Getmansky and Zeitzoff Reference Getmansky and Zeitzoff2014). Citizens are willing to endorse more aggressive policies, up to and including torture, when dealing with increased perils (Huddy et al. Reference Huddy, Feldman, Taber and Lahav2005; Rejali Reference Rejali2007, 22–23; Maoz and McCauley Reference Maoz and McCauley2008). Faced with such situations, individuals thus may not be open to the principles of humanitarianism and restraint at the heart of the modern laws of war.

Many of these problems are especially acute in societies wracked by conflict where many people have witnessed, or even perpetrated, extreme acts of violence. Some research points to the resilience of many populations in the face of extreme hardships caused by conflict, and the desire to work together to rebuild devastated societies (Bellows and Miguel Reference Bellows and Miguel2009; Blattman Reference Blattman2009), even if such commitments may be conditioned by contemporary circumstances (Freitag, Kijewski, and Oppold forthcoming).Footnote 6 In contrast, a number of other studies emphasize the pernicious effects of exposure to violence, especially with regard to the perception and treatment of outside groups. Across a wide range of countries, experiences with fighting reduce intergroup trust and lead to more exclusionary attitudes (Canetti-Nisim et al. Reference Canetti-Nisim, Halperin, Sharvit and Hobfoll2009; Rohner, Thoenig, and Zilibotti Reference Rohner, Thoenig and Zilibotti2013). A personal history rooted in violence can also result in more hard-line views with real policy consequences, such as lower levels of political trust, greater approval for reactionary politicians, reduced support for political compromise, and refusing to extend equal rights to out-groups (Berrebi and Klor Reference Berrebi and Klor2008; Beber, Roessler, and Scacco Reference Beber, Roessler and Scacco2014; Grossman, Manekin, and Miodownik Reference Grossman, Manekin and Miodownik2015; Hong and Kang Reference Hong and Kang2017). Of even more direct concern, the legacies of conflict can breed further violence, as past participants may be more prone to engage in even worse abuses down the road, such as ethnic cleansing (Jha and Wilkinson Reference Jha and Wilkinson2012). A large literature from social psychology confirms that traumatic events can lead victims to become perpetrators themselves, or to support increasingly destructive actions (Montada and Lerner Reference Montada and Lerner1998; Bandura Reference Bandura1999; Bar-Tal Reference Bar-Tal2004; Elcheroth Reference Elcheroth2006).

The prospects for international law to alter attitudes in the context of recent or ongoing conflict may thus appear quite dim. An emerging literature instead points to the potential for international legal instruments to transform the beliefs and behavior of domestic actors, even under trying circumstances. International law is thought to engender adherence through its legitimating qualities, including equality, consistency, or coherence, rather than through a reliance on threats of material punishment (Franck Reference Franck1990, 166–94). These attributes are not shared to the same extent by non-legal norms, and this may help account for the particular pull of legal principles (Fuller Reference Fuller1969, 40–65). This legitimating function can diffuse all the way down to individuals and encourage obedience even in the absence of direct monitoring and punishment (Tyler Reference Tyler1990, 19–27).

Enhanced by this authority, legal rules are well placed to transform the views of ordinary citizens on the ground. International law can transmit new information to domestic audiences, such as specific rights and responsibilities (Dai Reference Dai2007, 8). Yet what is perhaps most crucial is how these messages are received, interpreted, and absorbed, and the ways in which international law can change fundamental beliefs over what is considered appropriate. Ultimately, “[t]reaties matter because they potentially change the ideas that inspire political organization and activity” (Simmons Reference Simmons2009, 144). As with the focus of human rights agreements on governments’ treatment of their citizens, IHL may similarly have the potential to alter expectations over what is considered appropriate behavior by combatants during armed conflict. The existing literature thus points to distinctive pathways, both instrumental and more normatively oriented, whereby international law may influence preferences and outcomes. Yet determining the extent to which these different mechanisms are more or less likely to operate largely remains an open question in international legal scholarship (Simmons Reference Simmons2010).

A number of studies have shown that actors are often quite conscious of international commitments, and they alter their practices to appear more legitimate both domestically and internationally (Jo Reference Jo2015; Lasley and Thyne Reference Lasley and Thyne2015; Stanton Reference Stanton2016). Yet much of this research looks at the behavior of leaders and collective actors like states or rebel groups, or relies on rough proxy measures, to test claims that international law influences individual beliefs. A handful of recent studies have employed survey experiments to evaluate the effect of international law on public opinion and wartime conduct, such as torture or targeting civilians, more systematically (Wallace Reference Wallace2013; Chilton Reference Chilton2015). In these designs, respondents are randomly assigned to various prompts detailing international legal obligations, and then compared to a control group receiving no such legal information. Results from these studies show a clear causal relationship between exposure to information regarding international law and reduced public support for abusive conduct. Several of these studies also point to potential conditional relationships whereby international law can have a greater effect on some segments of society than others, such as political ideology or past military experience, the latter being especially relevant when thinking about IHL (Wallace Reference Wallace2013, Reference Wallace2014).

While promising and helpful in providing a more internally valid baseline for understanding the consequences of international law, these experimental studies contain a number of limitations. They construct generic hypothetical scenarios that may have little resonance with participants and the ways people would actually respond when faced with real-world situations.Footnote 7 The experimental manipulations are also extremely direct, in that respondents are given a clear statement that a specific practice (e.g., torture) is prohibited under international law, and then asked almost immediately afterward to indicate their approval or disapproval. Moreover, the emphasis in existing work has largely been to assess whether international law has an impact on public opinion, rather than the additional deeper question of the specific mechanisms through which legal principles may operate. Such designs thus offer a first cut for measuring the effects of broader exposure to the ideas underlying IHL, but they cannot assess whether there is a longer term or more diffuse relationship between people’s actual knowledge of the laws of war and their attitudes toward violence.

Much of this existing research is also limited to a single type of wartime conduct, and thus it can say little about the wider spectrum of abuses perpetrated during armed conflict, often in conjunction with one another. Furthermore, many of these studies solely involve respondents from the United States, many of whom have likely had no direct experience with armed conflict, which past research suggests should have important implications for attitudes toward wartime abuses. The next section seeks to build on the initial findings of these studies by examining the relationship between real-world knowledge of IHL and individual support for an array of violations across a wide range of post-conflict societies in which questions of law and violence are likely more salient and consequential.


A. Survey Data

The data used in this study to evaluate the relationship between international law and individual attitudes come from a cross-national survey, “The People on War Report” (POW) conducted at the behest of the ICRC (1999). The ICRC commissioned the survey in 1999 to mark the fiftieth anniversary of the signing of the Geneva Conventions, and it was intended to gauge general attitudes toward the laws of war writ large and the conduct these rules were designed to regulate. Many surveys contain at most a handful of items related to wartime conduct, such as a number of polls looking at public support for the use of torture against detainees (Gronke et al. Reference Gronke, Rejali, Drenguisa, Hicksa, Millera and Nakayama2010). Other surveys examining a broader range of behavior tend to focus more on government repression than on the direct wielding of violence by belligerents in the midst of armed conflict, which can be perpetrated by either state or non-state actors (Huddy et al. Reference Huddy, Feldman, Taber and Lahav2005; Merolla and Zechmeister Reference Merolla and Zechmeister2009). By contrast, the ICRC POW survey is unique in its wide coverage of questions related to many aspects of violence, as well as the legal rules and principles intended to regulate and restrain such abuses.

Beyond a battery of items on attitudes toward wartime conduct, another strength of the ICRC survey is its cross-national coverage. While many studies are limited to a single country,Footnote 8 the ICRC survey provides a useful comparative element by including participants from eleven different countries that had recently experienced, or were still experiencing, armed conflict. The countries surveyed are Afghanistan, Bosnia-Herzegovina, Cambodia, El Salvador, Georgia and the breakaway region of Abkhazia, Israel and the Palestinian occupied territories, Lebanon, Nigeria, Philippines, Somalia, and South Africa.Footnote 9 These countries also offer variation along several relevant dimensions, including geographic region, ethnicity and religion, conflict duration and severity, whether fighting was still ongoing, and length of time since hostilities ceased.Footnote 10

The survey was designed and fielded by the ICRC in conjunction with the survey firm Greenberg Research. A stratified multistage cluster sampling method was employed. The sample was first stratified to obtain an equal representation of the main geographic locations and groups involved in each conflict. For instance, in the case of Georgia-Abkhazia, around five-hundred respondents were selected from Georgia and another five-hundred from the Abkhazia region. In Bosnia-Herzegovina, the sample contained approximately five-hundred Muslims, five-hundred Croats, and five-hundred Serbs. The number of respondents surveyed in each country thus ranged between approximately one-thousand and fifteen-hundred depending on the number of parties to the relevant conflict. Each subsample was further stratified by geographic region. To ensure a representative sample to the greatest extent possible, respondents were then randomly selected within the resulting strata.

One further question concerns the timing of the survey, which was initially fielded around two decades ago. In the intervening years, other notable developments that have taken place include the emerging norm of the responsibility to protect (RtoP) and the formation of the International Criminal Court (ICC). There are several reasons for which potential insights gleaned from the ICRC survey remain relevant, especially in the case of evaluating the relationship between IHL and attitudes toward wartime violence. Although the ICC did not officially begin functioning until 2002, the Rome Statute establishing the court opened for ratification in July 1998, well before the time the survey was fielded. Precedents over possible prosecutions for violations had also been in place beforehand, including the special tribunals investigating and trying crimes committed during previous conflicts in Rwanda and the former Yugoslavia.Footnote 11 Similarly, the origins of the discourse surrounding RtoP developed in the middle and late 1990s in the aftermath of the Rwandan genocide and Balkan atrocities (Bellamy Reference Bellamy2009, 3–4), which also began before the survey was fielded. Although it is difficult to know a priori whether findings based on the ICRC survey would be stronger than, weaker than, or the same as similar polling conducted in the present day, the data nonetheless provide a useful reference point for thinking about the relationship between international law and individual attitudes toward wartime violence.Footnote 12

Of course, the ideal would be a cross-national panel survey of most countries of the world conducted across multiple waves in the manner of the World Values Survey, or similar regional projects such as the Afro- and Eurobarometers.Footnote 13 Yet the sheer breadth of coverage of questions asked about both wartime conduct and IHL makes the ICRC survey the most promising resource currently available for examining the sources of public attitudes toward the use of violence across a number of countries. The ICRC survey has also been successfully employed in other scholarly research looking into related questions on wartime violence. For instance, Meernik and King (Reference Meernik and King2014) examine differences in public support for transitional justice mechanisms, such as trials and truth commissions. Spini, Elcheroth, and Fasel (Reference Spini, Elcheroth and Fasel2008) look at attitudes toward wartime violence using a more limited set of items on support for abuse, though focusing specifically on the role of individual, group, and national context with little attention to the role of international law. Elcheroth (Reference Elcheroth2006) similarly looks at attitudes toward wartime violence using a more limited set of indicators, but he centers more on the role of individual and community-level trauma, while combining understandings of international law with support for abuse. My study thus seeks to build on this prior work by specifically examining the ways in which knowledge of IHL may directly influence attitudes toward wartime violence using a more systematic and disaggregated conceptualization of abuse.

The survey instrument contained a core set of questions measuring attitudes toward a wide range of types of wartime violence, as well as a variety of items dealing with individual attributes and experiences. Following common procedures used in many other cross-national surveys, question wording was sometimes modified slightly to ensure consistency in meaning. A full description of the sampling methods and questionnaire can be found in the final report for the survey (ICRC 1999, 87–88). The organization’s report also gave some indication of a possible relationship between a greater understanding of IHL and more restrained attitudes to what should be permitted during warfare (ICRC 1999, 19–20). Yet the original report’s discussion was in many respects preliminary and suggestive, given that it looked at only a few select abuses in isolation and did not take into account alternative factors or potential threats to validity. Most importantly, there still remains little sense in the existing literature of the relative explanatory power of specific mechanisms accounting for how awareness of international law may relate to individual attitudes toward abuses. The research design and analysis that follows thus seeks to address each of these issues to provide a fuller account of the relationship between international law and support for wartime violence.

B. Dependent Variable—Attitudes Toward Wartime Abuse

The main dependent variable involves the construction of a series of indices measuring each respondent’s general attitudes toward wartime abuse. The indices were created using a principal factor analysis of respondents’ answers to twelve separate items detailing various types of violations. Factor analysis allows for uncovering a common dimension, or set of dimensions, underlying a given set of variables (Kim and Mueller Reference Kim and Mueller1978b, 9–10). The component items fall into two main categories based on the identity of the victimized group—civilians or prisoners.Footnote 14 Table 1 summarizes the full set of items, which were rescaled so that higher values indicate greater levels of support for abuse. The six civilian measures focus on preferences toward attacks against civilian versus military targets, as well as strategies that disproportionately harm noncombatants, such as the use of land mines. For instance, the phrasing for the first civilian item is as follows: “When combatants attack to weaken the enemy, should they ….” Respondents were then given the following options to complete the sentence in declining order of willingness to accept violence against civilians: “Attack enemy combatants and civilians” / “Attack enemy combatants and avoid civilians as much as possible” / “Attack only enemy combatants and leave the civilians alone.”

Table 1. Questions Used to Construct Support for Wartime Abuse Indices

The six component prisoner abuse measures correspondingly cover a variety of common violations associated with the capture and detention of enemy combatants, ranging from torture to denying access to outside monitors. For example, in the first prisoner item listed in Table 1, respondents were asked: “Would you save the life of a surrendering enemy combatant who killed a person close to you?” Respondents could then answer either yes, they “Would save,” or no, they “Would not save.” Importantly, across all of the items used to construct the indices, the phrasing does not make any direct reference to international law, but focuses on the nature of the physical acts directed toward either civilians or prisoners. The resulting outcome variables thus seek to capture general attitudes toward wartime violence, while remaining distinct from the main explanatory variable centered on individual knowledge of IHL. The full text for these and other relevant questions is included in Supplementary Appendix A.

Analyzing these twelve items revealed one common underlying factor, having an eigenvalue of 2.04. Additional tests confirmed that a single dimension of support for wartime violence best described the relationship among the component variables.Footnote 15 The loadings generated from the factor analysis were then used to create an index of general attitudes toward wartime violence, Overall Abuse. To facilitate subsequent estimation and interpretation, the measure was rescaled to have a range between 0 and 1, where higher values indicate greater support for wartime violence.Footnote 16 Because past research suggests there may be important differences in the treatment of various groups during war (Morrow Reference Morrow2007; Downes Reference Downes2008; Wallace Reference Wallace2012), separate indices were also created for violence toward civilians and prisoners. Civilian Abuse and Prisoner Abuse are corresponding indices generated by separate factor analyses on the six specific items for each type of abuse, respectively, and their scales were then normalized in a similar manner.

Figure 1 presents a histogram of the distribution of values across the three indices, overlaid with a smoothed kernel density plot. Figure 1a indicates that attitudes are weighted toward the lower end of support for wartime violence, which should offer some relief to activists and policymakers worried over the general hardening of attitudes that can occur in conflict and post-conflict societies (Kaufmann Reference Kaufmann1996). Nevertheless, only around 10 percent of respondents professed the lowest levels of objection to the resort to abuse, meaning that the vast majority of those surveyed felt at least some situations were warranted where protections could be denied to civilians and captured combatants. When distinguishing between the particular victim of abuse, lower end values are more prevalent for the separate civilian and prisoner indices. Figures 1b and 1c also show that attitudes toward prisoner abuse tend to be more concentrated, not only at the lower end of the scale, but also sometimes at the upper ends, while opinions toward the treatment of civilians are relatively more dispersed across the full range of values.

Note: Histogram for each relevant abuse index is overlaid with a kernel density plot (black line).

Figure 1. Summary of Support for Wartime Abuse Indices

Those respondents indicating the greatest levels of abhorrence for wartime abuse are not necessarily the same when it comes to the treatment of civilians versus combatants. Although the two victim-specific indices are positively correlated overall (r = 0.39), they are actually negatively related (r = –0.29) at the lower ends for values less than or equal to 0.1 on the normalized scale. This pattern points to the importance of considering the possibility that the correlates of support for wartime violence may differ along certain dimensions depending on the identity of the prospective victim. This localized negative relationship between the two indices also helps account for why the proportion of cases falling at the lower ends is more pronounced for each index individually in Figures 1b and 1c than in the overall abuse index in Figure 1a.

Figure 2 offers a better sense of patterns in support for different forms of wartime violence by plotting the cumulative distribution function for each index. While all three indices follow a similar pattern, opposition to wartime violence actually seems slightly more noticeable for abuses directed toward prisoners than toward civilians. This may seem surprising given the general emphasis in much of the discourse on the rights of civilians, along with the lesser threat or blame generally attributed to noncombatants that could be used to justify violence. However, the different distributions may partly be a function of the constituent items used to construct each index—several of the prisoner questions were much more direct, such as torturing or killing combatants, while a number of the civilian items involved less direct actions (denying food, planting land mines) or emphasized “attacking” rather than the more charged language of “killing.” It is thus difficult to compare the absolute levels across the various abuse indices directly. Of course, it might be preferable if the questions asked were exactly the same except for the group being victimized. However, closer congruence would come at the cost of ignoring the most relevant types of violence directed against civilians versus prisoners, which often differ importantly in numerous respects (Morrow and Jo Reference Morrow and Jo2006). Ultimately, for the purposes of this study, overall levels of support for various forms of wartime violence are generally less central than are differences in individual attitudes based on key attributes of respondents, in particular their understanding of IHL.

Figure 2. Cumulative Distribution Functions of Support for Wartime Abuse Indices

C. Main Explanatory Variable—Knowledge of IHL

The main independent variable concerns individuals’ knowledge of IHL and the consequent limits these rules seek to place on the resort to abuse in times of armed conflict. While some experimental studies have used some form of prompt stating the illegality of a given violation to proxy for exposure to international law, the POW survey offers a more direct measure of individuals’ actual knowledge of IHL, which is intended to be separate from their policy preferences.

The IHL regime complex has expanded greatly over the last century to regulate a wide range of behavior.Footnote 17 The 1949 Geneva Conventions nonetheless remain the foundation for much of contemporary IHL (Best Reference Best1994, 136; Evangelista and Tannenwald Reference Evangelista and Tannenwald2017). While mainly addressing international armed conflicts, Common Article 3 of the conventions also guarantees a fundamental set of protections for victims in civil wars, or “conflicts not of an international character” (Solis Reference Solis2010, 96–104). Protections afforded under the Geneva Conventions to civilians and prisoners, among others, codified many prior rules and customs, while continuing to serve as a basis for later agreements, including the conventions’ two Additional Protocols of 1977 and the 1998 Rome Statute establishing the ICC (Schabas Reference Schabas2011, 131–33). As such, understanding the conventions introduces individuals to basic principles of restraint and humanity at odds with the violations at the core of the various abuse outcome indices.

To capture exposure to IHL, I rely on a pair of questions posed to all respondents in the survey. Respondents were first asked: “Have you ever heard of the Geneva Conventions?” If they answered affirmatively, this was followed by an open-ended question: “Could you tell me what the Geneva Conventions are about?” Enumerators for the survey then recorded whether these respondents accurately identified the main purpose of the conventions based on mentioning any reference to the placement of limits on wartime conduct. I then used these items to construct the dichotomous variable, Know About Geneva Conventions, which equals 1 if a respondent had heard of the conventions and correctly identified their purpose, and 0 if respondents had either not heard of the conventions or incorrectly described what the laws were about. This variable exhibits clear variation, as around 25 percent of respondents demonstrated a general understanding of the Geneva Conventions, suggesting exposure to basic IHL principles is evident but far from universal.Footnote 18

As a measure of IHL knowledge, the variable admittedly is unable to make fine-grained distinctions in a person’s relative level of knowledge or awareness of particular rules of warfare. Nevertheless, the measure offers a number of distinct advantages. First, the second item is open-ended, requiring respondents to provide an answer in their own words, making guessing and false positives more difficult and less of a worry than a closed-ended format. Second, the generality of the construct is beneficial as it is not directly linked to specific items gauging preferences for or against abuse, mitigating concerns over consistency bias between responses. Third, the two Geneva Conventions questions were asked well after the key items used to construct the abuse indices outcome variables, helping to minimize worries over question ordering and contamination across responses.Footnote 19

D. Additional Explanatory Variables

A number of other variables are also included that past research suggests should have some bearing on individual support for wartime abuse. The first category of variables concerns the experiences of respondents during the conflict given that such events can fundamentally change their political attitudes (Blattman Reference Blattman2009). Lived in warzone is a dichotomous variable that equals 1 if fighting took place in the area where the respondent lived, and 0 otherwise.Footnote 20 In addition to their general proximity to the fighting, the next two variables capture respondents’ personal experiences during the conflict. Although suffering any form of abuse or deprivation is deplorable, these variables seek to distinguish between severe and more moderate forms of violence that respondents may have encountered, as each type may have different implications for their attitudes (Balcells Reference Balcells2012). Respondents were presented with a list of twelve wartime experiences that might have happened to them. More severe experiences are those that involved the infliction or grave threat of direct physical violence, while experiences considered more moderate included indirect deprivations or where violence was experienced secondhand against a close relation.

Severe wartime experience is a dichotomous variable that equals 1 if the respondent reported experiencing any of the following acts—being imprisoned, kidnapped, or taken hostage; tortured; or wounded by the fighting; and 0 otherwise. Moderate wartime experience is a corresponding dichotomous variable that equals 1 if the respondent instead reported experiencing any of the following—felt humiliated, lost contact with a close relative, property was seriously damaged, food was taken away, house was looted, forced from home, an immediate family member was killed, or a person they knew well was raped by combatants, and 0 otherwise. This variable was further coded 0 for those respondents who reported that one or more of the severe acts had also happened to them to ensure the two wartime experience variables do not overlap.

While the previous three variables capture acts or situations encountered by respondents, the next two variables examine the nature of an individual’s contributions and actions taken during the war. Took sides is a dichotomous variable that equals 1 if the respondent actively supported one of the warring parties during the war, and 0 otherwise. Unfortunately, the question does not provide an opportunity to distinguish between supporters of the government versus opposition forces. Using religion or ethnicity as a rough proxy for support is inappropriate, since it is not uncommon for co-ethnics to choose or fight on opposing sides (Kalyvas Reference Kalyvas2008; Lyall Reference Lyall2010), not to mention that the ethnic identity measures in the ICRC survey suffer from a number of problems (Spini, Elcheroth, and Fasel Reference Spini, Elcheroth and Fasel2008). Nevertheless, the variable provides a reasonable measure of whether individuals who are more politically engaged in the conflict may be more or less likely to support the use of violence (Balcells Reference Balcells2010). To assess whether those with past military experience possess different attitudes (Feaver and Kohn Reference Feaver and Kohn2001), Veteran is a dichotomous variable that equals 1 if the respondent took up arms, and 0 if he or she remained a noncombatant throughout the conflict.

The second set of indicators covers several socio-demographic characteristics that tend to be associated with a variety of political attitudes (Holsti Reference Holsti2004). Age is a categorical variable that separates respondents into a series of age ranges.Footnote 21 Gender is a dichotomous variable that equals 1 if a respondent is male, and 0 if female. Has children is a dichotomous variable that equals 1 if a respondent has children, and 0 otherwise.

Education is a particularly important variable for evaluating the theoretical argument put forward. Higher levels of education have been associated with support for more liberal policies, such as free trade and immigration (Hainmueller and Hiscox Reference Hainmueller and Hiscox2006, Reference Hainmueller and Hiscox2007). On the other hand, the role of education in armed conflict and political violence is much less settled (Thyne Reference Thyne2006; Brockhoff, Krieger, and Meierrieks Reference Brockhoff, Krieger and Meierrieks2015). If respondents who know more about the Geneva Conventions also tend to be the most educated, then any observed relationship between attitudes toward violence and awareness of IHL may simply be a function of educational attainment. The survey recorded the number of years of schooling attained by each respondent. Countries often have different national systems for primary school, secondary school, and postsecondary education, which makes a single common educational classification system problematic. To rectify these concerns and still allow a comparable cross-national measure, Education is a four-level categorical variable that divides the full sample of respondents into quartiles based on the number of years of schooling completed.Footnote 22

Along with these socio-demographic items, the survey also included an experimental element in which respondents were presented with slightly different wordings for a number of items, including some of the abuse questions, such as a generic scenario versus a partisan situation involving combatants from the respondent’s own side.Footnote 23 Partisan wording is a dichotomous variable that equals 1 if the respondent received the more partisan scenario for relevant questions, and 0 if he or she instead received the more nonpartisan wording. While there are differences in overall responses for some items based on the particular version asked, they are not always in the same direction—in some cases, the partisan version is formatted to elicit responses more supportive of abuses while in others it leads to greater opposition. Additionally, the relative distribution of values across the various indices do not differ markedly when subset by either of the two experimental groupings, suggesting that the different wordings should not overly affect the results of interest for this study.Footnote 24 Nevertheless, the variable is included to take into account that those receiving the partisan wording may differ in their responses. Table 2 provides summary statistics for the main variables used in the analysis that follows.Footnote 25

Table 2. Summary Statistics for Main Variables Used in the Analysis


A. Overall Findings

Because the indices forming the main dependent variables dealing with attitudes toward wartime violence are continuous, regression models were estimated using ordinary least squares (OLS). To control for any country-specific effects, separate dichotomous variables were included for each country in the survey.Footnote 26 Standard errors were clustered by country to take into account that observations are unlikely to be independent within each country.

Figure 3 reports results from a series of OLS models estimating individual attitudes toward the use of violence across each of the three indices—overall, civilian, and prisoner abuse. The figure plots coefficients for the main variables of interest across each abuse index, with horizontal lines indicating 95 percent confidence intervals.Footnote 27 Since each index was normalized to values ranging between 0 and 1, interpreting the coefficients is relatively straightforward. Positive (negative) coefficients indicate a given percentage point increase (decrease) in the level of support for abuse. Supplementary Appendix C presents more complete regression results from this figure as well as for the subsequent analyses discussed in further detail below.

Note: Coefficient plot (OLS) indicating percentage point change in support for each abuse index. Lines indicate 95 percent confidence intervals. Results for partisan wording and country variables not shown.

Figure 3. Substantive Effects of Support for Wartime Abuse

To give a better initial sense of raw levels of support, I calculated the predicted absolute values for each abuse index, while holding all explanatory variables constant at their median values.Footnote 28 The resulting value for overall abuse was 0.23, which is generally in line with the patterns reported in Figure 1, where respondents on average demonstrate a relatively low propensity to countenance abuse, but they are nonetheless willing to condone violations in a variety of circumstances. Based on their corresponding models, the values were 0.18 for civilian abuse and 0.27 for prisoner abuse. While this may seem to reflect a lower willingness in general to support wartime violence against civilians than against prisoners once other factors are taken into account, as noted earlier, any direct comparisons can be suggestive only given the difference in question wordings and the normalized scales across the two victim-specific abuse indices.

Figure 3 instead focuses on the first differences associated with each independent variable, since our main quantity of interest is the relationship between knowledge of international law and support for wartime abuse. The results show that prior awareness of IHL is associated with greater restraint toward violence, though this relationship is not consistent across all forms of abuse. Looking at the combined overall abuse index, individuals who have prior knowledge of the Geneva Conventions are 3 percentage points less supportive of the resort to violence, a modest though statistically significant finding. Importantly, this result holds even when controlling for educational attainment, suggesting that IHL knowledge is not simply a function of those respondents who happen to have more schooling and, presumably, a greater level of general knowledge.Footnote 29 By contrast, in the specific case of violence against civilians, the association with IHL knowledge disappears; the coefficient is still negative but greatly diminishes in size and statistical significance. The initial results for the overall abuse index appear instead largely driven by attitudes toward the treatment of enemy combatants. Individuals with greater IHL knowledge are more than 5 percentage points less supportive of abuse toward prisoners. This same pattern holds across various alternative specifications of explanatory variables, as well as combinations of the component items used to construct the abuse indices.Footnote 30

The consistent non-finding for the association between IHL knowledge and civilian abuse does not necessarily imply that respondents do not care about the plight of noncombatants. These weaker results may instead actually point to the possibility of a more widespread acceptance of underlying norms of civilian protections, as suggested in the absolute values noted above, and along the lines enunciated in what later became known as the RtoP norm. Of course, individuals may still condone the targeting of civilians for strategic reasons, even while acknowledging that limits likely constrain such actions in principle. The discrimination between combatants and noncombatants and, by extension, the protection of the innocent remains one of the foundational tenets of just war theory (Walzer Reference Walzer2000, 151), and it may thus be more commonly accepted as such in mass opinion. Having more specific IHL knowledge may bring little new to the minds of individuals in conflict situations and, in turn, have little discernable ability to change their attitudes toward civilian targeting.

By contrast, the appropriate treatment of prisoners may be less settled given the uneasy status of combatants who had been fighting, or potentially even perpetrating abuses, not long before capture (Kinsella Reference Kinsella2005; Wallace Reference Wallace2015, 42–45). The War on Terror highlights how states can treat civilians relatively humanely, while simultaneously brutalizing enemy combatants (Kahl Reference Kahl2007, 8 n3). To the extent that the Geneva Conventions and similar laws can disseminate new beliefs about appropriate treatment of detainees, or bolster existing suppositions with greater legitimacy, it follows that IHL may exert greater restraint on attitudes toward prisoner abuse. Of course, it is important to note that this interpretation of the different results for civilian versus prisoner abuse is only suggestive. The available data unfortunately do not provide the ability to conclude definitively that individuals’ baseline understandings of varying forms of abuse are driving the results.

Although differences in support for both overall abuse and prisoner abuse associated with greater knowledge of IHL are modest, they are of comparable size to those found in related experimental studies (Wallace Reference Wallace2013; Chilton Reference Chilton2015). Caution must also be exercised before drawing any conclusions of a causal effect attributable to international law based on the observational data used in this study; a point discussed in greater detail below. It is nonetheless promising that the size of the observed differences in attitudes toward wartime violence based on a real-world measure of exposure to the laws of war closely tracks the causal estimates gleaned from prior experimental studies, which exhibit high levels of internal validity.

Looked at from a relative standpoint, the coefficient for knowledge about the Geneva Conventions is also at least of equal or greater size to the other explanatory variables measuring many other common explanations for attitudes toward violence. In fact, in the prisoner abuse model, IHL knowledge represents the single largest coefficient, with a greater associated change in attitudes toward wartime violence than direct personal experiences, gender, or being a veteran. The substantive importance of prior awareness of international law is thus in line with existing research and is impressive when compared to alternative factors.Footnote 31

Turning to the other explanatory variables, many are generally consistent with past research. For wartime experiences, only the gravest types of past events have any systematic relationship with support for abuse or similarly harsher viewpoints (Balcells Reference Balcells2012). Living in the midst of fighting has no discernable association on its own with an individual’s willingness to condone or condemn wartime abuses. More moderate (though nonetheless serious) experiences, such as having one’s house looted or losing contact with close relatives, similarly appears to have little bearing on attitudes toward appropriate conduct. On the other hand, severe wartime experiences like being wounded or tortured are positively and significantly associated with greater levels of approval for all three of the abuse outcomes. The finding thus confirms that exposure to violence can breed an escalating cycle of abuse, as those who are victimized are more likely to support similarly brutal actions in the future (Elcheroth Reference Elcheroth2006).

Partisans are also more likely to approve of greater levels of violence, suggesting that those who are more supportive of armed groups are also more willing to condone the resort to abuse. Similarly, those who actually took up arms were generally more supportive of abuse, which is consistent with prior research showing the greater proclivities of those with military experience in approving higher levels of violence (Feaver and Gelpi Reference Feaver and Gelpi2004, 60). Interestingly, veterans are no more likely to support the abuse of captured combatants, perhaps suggesting that the prospect of retaliation may make soldiers more hesitant to support violations against their enemy counterparts in ways that are more salient than when the victims would instead be civilians (Axelrod Reference Axelrod1984, 73–76).

Looking at the remaining socio-demographic control variables, those who are older are generally less likely to support violence, though the coefficient does not attain statistical significance in the case of civilian abuse. Given past research pointing out that countries with demographic shifts experiencing so-called youth bulges are more conflict prone (Urdal Reference Urdal2006), the results suggest that younger populations may also be more supportive of higher levels of violence in the event of war. Unlike other research pointing to a fairly robust relationship between gender and the use of force (Eichenberg Reference Eichenberg2003; Brooks and Valentino Reference Brooks and Valentino2011), there are less definitive differences between men and women when it comes to explaining attitudes toward wartime conduct, with the exception of civilian abuse. Education similarly shows few consistent findings—the coefficient is negative and significant for prisoner abuse as expected, but it fails to achieve statistical significance in the other models and it is actually slightly positive for civilian abuse. Although education may be a strong predictor of more liberal attitudes for a number of other foreign and domestic policies, this does not appear to be the case for wartime conduct. On the other hand, having children does appear to be associated with a small though persistent decline in approval of wartime violence, which corresponds to evidence examining general attitudes toward the use of force (Smith Reference Smith2005; Horowitz and Levendusky Reference Horowitz and Levendusky2011). Perhaps because of concerns over possible future negative consequences for their children and later generations, respondents who are parents appear less likely to condone the resort to extreme forms of violence during wartime.Footnote 32

B. Who Might the Law Influence Most? Assessing Conditional Relationships

One potential issue with the findings discussed so far is that the relationship between IHL knowledge and support for wartime violence may be masking important conditional relationships, or lack thereof, among specific groups of respondents. This may be of particular concern if the role of IHL is greatest simply for those who already espouse fairly humanitarian ideals, while having little influence on those most skeptical of such constraints. International law might be succeeding in moving attitudes, but the effect may be limited to likeminded individuals, while playing no role for harder cases. Indeed, one of the most frequent charges against international law is that such rules and principles only work in easier situations, or when actors are already positively predisposed, but they fail under more trying circumstances (Mearsheimer Reference Mearsheimer1994–1995).

To evaluate this possibility, I conducted a series of analyses distinguishing individual respondents by key traits commonly associated with different baseline attitudes toward wartime violence. Each model was estimated with the same general specification of explanatory variables as in the original analysis, but it was then limited to respondents from the relevant subsample—for instance, civilians versus veterans.Footnote 33 Figure 4 summarizes the results of these subsample models across four covariates—civilian-veteran status, type of wartime experience, whether the respondent took sides during the war, and gender. Although each of these covariates finds some (though varied) support for overall differences in attitudes toward wartime abuse, past research also suggests differences in terms of sensitivity to international legal appeals (Brooks and Valentino Reference Brooks and Valentino2011; Meernik and King Reference Meernik and King2014; Wallace Reference Wallace2014).

Note: Plot for Knowledge of Geneva Conventions coefficient (OLS) indicating percentage point change in support for each abuse index by relevant subgroup. Lines indicate 95 percent confidence intervals.

Figure 4. Knowledge of Geneva Conventions and Support for Wartime Abuse

Figure 4 reports the coefficient (measuring the percentage point change in support for abuse) for knowledge of the Geneva Conventions for the relevant subsample, while holding all other explanatory variables constant. The subfigures suggest that the constraining relationship between international law and support for wartime abuse is not only limited to those already predisposed to international legal principles, but also evident even for those who at first glance may not be very welcoming of humanitarian appeals. If anything, the results are slightly stronger for seemingly harder cases like veterans and those with more severe wartime experiences, though differences between subgroups sometimes do not attain standard levels of statistical significance.Footnote 34 This is especially apparent for the civilian abuse measure, which is negative for veterans, severe wartime experiences, and those who took sides, while actually slightly positive for their counterparts (civilians, individuals with moderate or no serious wartime experiences, and those who did not choose sides). Consistency in the role of international law across certain groups is evident in Figure 4d, which shows similar results across both men and women. Although the differences between the two subgroups are not statistically significant, they demonstrate that the role of IHL knowledge is comparable, and potentially even more pronounced, across subpopulations that may approach both violence and legal principles in very different ways.

C. Discussion and Implications

Although the results point to an understanding of IHL being associated with a decline in support for wartime violence, a number of concerns should be kept in mind. Perhaps most importantly, in the preceding analysis I have been careful not to assert that a definitive causal relationship exists between knowledge of IHL and attitudes toward wartime violence. Assessing causality on the basis of observational data is inherently problematic and suffers from a number of threats to validity (Winship and Morgan Reference Winship and Morgan1999)—several of which I discuss in turn.

First is the possibility of reverse causality, where those who already possess more humanitarian inclinations are more likely to seek out sources of rules and principles that are consistent with their beliefs. Rather than international law playing any independent role in shaping individual preferences, knowing about these rules may simply be a function of preexisting attitudes toward appropriate wartime conduct. Although by no means definitive, the conditional relationships shown in Figure 4 mitigate some of the worst worries that reverse causality is driving the results. If prior preferences were decisive, the role of international law should have been especially prominent for subgroups tending to be approving of more benevolent practices in general. Despite veterans, partisans, and those experiencing traumatic wartime events exhibiting higher overall support for violent conduct, the association with IHL knowledge in these subgroups was of similar magnitude as that of their more humanitarian counterparts.

This points to a second and more general inferential concern over the nonrandom nature of the key explanatory variable in the study—knowledge of IHL. Unlike in an experimental setting, respondents were not randomly assigned to whether or not they were exposed to the key principles governing the laws of war. While the ICRC survey may offer a novel and effective measure for individuals’ actual IHL knowledge, those respondents who gained knowledge about the laws of war may differ in important ways from those with little or no awareness of international legal principles. Especially problematic is whether those individuals with a greater propensity to be aware of IHL also have systematically different preexisting attitudes toward wartime violence. This suggests an individual-level corollary to the larger problem plaguing many observational studies of international law. It can be difficult in such studies to discern whether international legal principles are actually exerting any independent effect, or whether some other confounding factor is explaining both commitment to international law and the outcome behavior the agreement is intended to regulate (Simmons and Hopkins Reference Simmons and Hopkins2005; von Stein Reference Von Stein2005). In the case of the survey data analyzed here, perhaps there is some unmeasured individual factor, such as pacifism, accounting for both lower support for wartime abuses and a person’s greater desire to seek out and understand more about the very laws intended to regulate such violence.

Although the evidence reported here may only indicate a correlation between international law and policy attitudes, as noted earlier, the results track closely with the findings of a number of past survey experiments that have demonstrated with greater confidence a causal relationship, whereby exposure to an international legal treatment prompt affects public opinion toward controversial policies (Tomz Reference Tomz2008; Wallace Reference Wallace2013; Chilton Reference Chilton2015). Moreover, further experimental work shows that the effects of international law may actually be weakest for those with strong policy preferences (Chaudoin Reference Chaudoin2014), suggesting that the current results may, if anything, underestimate the importance of international law.

The conditional relationships from Figure 4 continue to offer further insight into whether some hidden confounder, such as pacifism, is explaining the observed results. To the extent that pacifism is interrelated with other types of traits, such as gender, civilian-veteran status, or being a partisan during the war, it would be expected that the role of international law in support for abuses should be significantly different for these subgroups. However, the results from Figure 4 indicated relatively similar coefficients for knowledge of international law between these different subgroups.

Additional statistical analyses also suggest unobserved confounding is unlikely to be driving the main results. Matching has been increasingly offered as an approach to lessen the risk that any observed differences in outcomes between a treatment and control group is a function of nonrandom assignment to the treatment (in this case, knowledge of the Geneva Conventions). Matching pairs a “treated” observation (here, someone who knows about the Geneva Conventions) with an observationally similar “control” unit (someone who does not know about the Geneva Conventions) based on a defined set of covariates. The procedure generates more evenly balanced data that minimize observed differences on the covariates between the treatment and control groups (Ho et al. Reference Ho, Imai, King and Stuart2007). This helps increase confidence in an observed relationship between a treatment and an outcome—in this case, between knowledge of international law and support for wartime violence.

I employed a genetic matching routine using all the main explanatory survey variables and country indicators as the baseline covariates. Genetic matching is a multivariate matching method designed to achieve maximum balance in the covariates between treated and control groups (Diamond and Sekhon Reference Diamond and Sekhon2013). Although not reported here, the results using matching are not substantially different from those reported earlier based on more conventional OLS regression models.Footnote 35 Of course, the reliability of any matching findings depends on the assumption that assignment to the treatment group (knowledge of the Geneva Conventions) is captured by the observed covariates (Rosenbaum Reference Rosenbaum2010, 76–79). However, further sensitivity tests indicate that the results for attitudes toward abuse are robust to the potential exclusion of a fairly strong unobserved confounder that could account for people’s propensity to know about international law.Footnote 36

Third, and building in a more specific direction off the prior point, the positive finding for awareness of IHL may simply reflect respondents’ general levels of knowledge, which presumably should be higher for those who have completed more years of education. However, the relationship between education and attitudes toward violence is far from clear in the existing literature. While some work suggests that education serves as a constraint on the resort to violence (Collier and Hoeffler Reference Collier and Hoeffler2004; Thyne Reference Thyne2006), other research shows that education can serve as a basis for heightened conflict (Posen Reference Posen1993; King Reference King2013, 7–11).

Looking at the survey evidence, Figure 3 shows that the role of IHL knowledge continued to hold even after controlling for educational attainment. Yet it may instead be the case that IHL is only able to shape individual preferences among those with a sufficient baseline level of general knowledge and the ability to grasp relevant humanitarian principles. To test for the possibility of a conditional relationship with education, I separated the analysis based on those with greater (upper two quartiles) and lesser (lower two quartiles) levels of educational attainment (Figure 5). The results indicate that respondents across education levels are similarly influenced by IHL, and for civilian abuse the coefficient is actually slightly stronger (though not statistically distinguishable) for those with lower levels of educational attainment.

Note: Plot for Knowledge of Geneva Conventions coefficient (OLS) indicating percentage point change in support for each abuse index by relevant subgroup. Lines indicate 95 percent confidence intervals.

Figure 5. Knowledge of Geneva Conventions and Support for Wartime Abuse by Level of Education

Fourth, concerns of social desirability bias loom large in studies of controversial topics (Podsakoff et al. Reference Podsakoff, MacKenzie, Lee and Podsakoff2003), especially with the present survey’s focus on the victimization of vulnerable populations. Faced with sensitive questions, respondents may be motivated to give answers they expect to be socially appropriate and what interviewers want to hear, rather than revealing their sincere preferences. Such biases may be even more severe for respondents with knowledge of the Geneva Conventions, since they are already more aware of existing constraints.

Several reasons suggest social desirability concerns are not unduly influencing the results.Footnote 37 Despite the direct nature of some questions, many respondents remained willing to condone fairly brutal behavior explicitly. For instance, looking at the raw results for some of the component abuse items, more than 25 percent of respondents approved using land mines that would endanger civilians, and over one-third reported that they would refuse to save an injured or surrendering soldier. A similar lack of concern for prisoners was evident even among those with prior knowledge of IHL—over 30 percent still supported using torture. Moreover, the prior conditional plots again illustrate that individuals who should be more sensitive to social desirability dynamics, such as civilians, exhibit similar influences concerning IHL as former combatants and others normally espousing more hard-line perspectives. In sum, although there are a number of reasons to be cautious about inferring a direct relationship between international law and attitudes toward wartime violence, the preceding discussion has sought to address a number of the more serious likely threats to the validity of the findings.

D. Assessing Potential Mechanisms

The main analyses presented above offer only a general sense of the pathways and reasoning through which international law may be related to individual attitudes. Yet one of the strongest calls has been for empirical scholarship to understand better how international law and institutions shape preferences and outcomes (Martin and Simmons Reference Martin and Simmons1998). Building on the prior analysis establishing an association between international law and public attitudes, I turn below to assessing some of the specific mechanisms that may account for this relationship.

For instance, while the content of the conventions appears to matter, it is unclear whether individuals are thinking specifically about the international legal nature of those rules. Constraints on the resort to wartime violence reflect contemporary IHL, but they may also be rooted in other sources, such as domestic law, or religious and other values (Howard, Andreopoulos, and Shulman Reference Howard, Andreopoulos and Shulman1994). Persons exposed to IHL may become aware of the accompanying principles, but they may in fact be motivated by unrelated belief systems. The ICRC survey largely focuses on IHL by design, and it does not contain a comparable set of extensive questions tapping into alternative value systems. However, an item is included where respondents were asked what they believed was the basis of rules the violation of which perpetrators should be held accountable. Given the nominal nature of the outcome variable, a multinomial logit model was estimated. Figure 6 summarizes how IHL knowledge is associated with differences in the probability a respondent gave each answer choice.Footnote 38 The results indicate a close and statistically significant correspondence between knowledge of IHL and beliefs that wartime limits are based on international legal rules. By contrast, respondents with prior exposure to IHL are less likely to believe that such rules are rooted in domestic legal systems; the result is similarly negative and statistically significant for religious values.

Note: Multinomial logit estimation. Average marginal effect (percentage point change in the predicted probability of the outcome) of Knowledge of Geneva Conventions, holding all other variables at their median values. Lines indicate 95 percent confidence intervals. Model is limited to respondents who believed there were rules for which people should be held accountable if they violated them.

Figure 6. Knowledge of the Geneva Conventions and Beliefs on the Basis of Rules for Which Violators Should Be Held Accountable

The lack of findings for religious or other values does not necessarily imply that individuals exposed to international law view such principles as divorced from normative belief systems. Indeed, debate continues over whether international law primarily influences beliefs and behavior through normative appeals, or more instrumental motives; although the line between the two is often blurred (Simmons Reference Simmons2010, 277). A few other items from the ICRC survey offer an opportunity to examine the specific ways in which IHL may inform individual attitudes.

Respondents were presented with a question asking: “Is there anything that combatants should not be allowed to do in fighting their enemy?” Estimating a logit model, the first point estimate in Figure 7 shows the change in the predicted probability a respondent answered affirmatively that limits do exist on soldiers’ conduct when fighting.Footnote 39 Not surprisingly, respondents with prior knowledge of IHL are over 15 percentage points more likely to say such limits exist. Those respondents who answered affirmatively were then asked a follow-up question about precisely why limits on wartime conduct should exist. The two available answer choices tap into related but distinct reasons underlying these beliefs—one more normatively centered on a sense that such actions are morally wrong, and one more instrumentally focused on the problems that may result from violations. The second point estimate in Figure 7 from a corresponding logit model shows that those who know about the Geneva Conventions are 7 percentage points more likely to believe limits exist because violations are morally wrong rather than for more instrumental reasons.

Note: Logit estimation separately for each outcome. Average marginal effect (percentage point change in the predicted probability of the outcome) of Knowledge of Geneva Conventions, while holding all other variables at their median values. Lines indicate 95 percent confidence intervals. Model for ‘Why Limits?’ limited to respondents who answered affirmatively to prior question on ‘Any Limits on Fighting?’

Figure 7. Knowledge of Geneva Conventions and Beliefs of Limits on Wartime Conduct

Delving further, those who offered a more normative rationale were asked an additional follow-up question inquiring into why they believed such violations were morally wrong. Similar to the earlier item on reasons for holding violators accountable, some options centered on more universal ideals, such as the importance of the law and human rights, while others highlighted religion, culture, or a personal code.Footnote 40 For this question, however, answers were not mutually exclusive, as respondents were allowed to give multiple reasons. Since a conventional multinomial logic model could not be used, I instead estimated a series of separate logit models for each outcome. The results in Figure 8 show that normative beliefs in the limits of wartime conduct for those with prior exposure to IHL tend to be rooted in more universal appeals to law and human rights. Although the confidence intervals for the law/human rights first differences overlap with some of the other answer choices, neither cultural principles nor personal codes were statistically distinguishable from 0, or no apparent effect. Moreover, those with prior IHL knowledge were significantly less likely to attribute any moral beliefs on the resort to abuse to religious ideals.

Note: Logit estimation separately for each outcome. Average marginal effect (percentage point change in the predicted probability of the outcome) of Knowledge of Geneva Conventions, while holding all other variables at their median values. Lines indicate 95 percent confidence intervals. Models limited to respondents who answered ‘It’s wrong’ on previous item of ‘Why are there limits on fighting?’

Figure 8. Knowledge of Geneva Conventions and Beliefs for Why Breaking Limits on Wartime Conduct Is Wrong

Taken together, the previous discussion responds to a number of different concerns arising over various claims regarding the nature of the relationship between knowledge of IHL and attitudes toward wartime violence. The analysis indicates that exposure to the Geneva Conventions does have a discernable relationship with support for abuse, the specific humanitarian content espoused in the agreements plays a key role, and these principles appear to be associated with heightened moral calculations among respondents that are rooted in universal legal norms of restraint on the use of force and protections of individual rights.


This study offers evidence that international law is related to differences in attitudes over controversial issues like the resort to wartime abuse, and even in the difficult circumstances of countries coming out of, or still experiencing, armed conflict. While the constraining role of IHL is more evident for issues like prisoners than the targeting of civilians, the relationship is still consistent overall as well as for subgroups that may have different baseline preferences toward both violence and the law. Care needs to be taken in ascribing any causal effect of international law on public attitudes toward abuses, but an assessment of several likely inferential threats suggests that some of the most serious concerns are mitigated. The survey evidence also helped indicate the ways in which international law may inform attitudes, in particular by activating normative considerations related to the moral limits on the resort to violence during armed conflict.

At their founding, the Geneva Conventions envisioned the dissemination of knowledge about the laws of war as central for promoting adherence to their underlying humanitarian principles. The analysis indicates one route through which IHL may do so by influencing mass attitudes toward appropriate wartime conduct. The findings also lend some credence to contemporary programs by the ICRC and other organizations that specifically seek to transform the beliefs of both civilians and combatants through informing the wider population about the principles and obligations under IHL.Footnote 41

The results also raise additional questions for future research. The survey data necessarily focused on respondents’ preferences toward violence rather than actual behavior observed during wartime. Of course, the resort to violence may be driven by a variety of processes separate from specific individual preferences, such as the broader context of the fighting, structures of military organizations, and small-group dynamics (Browning Reference Browning1998; Kalyvas Reference Kalyvas1999; Weinstein Reference Weinstein2007). Although posing several methodological challenges, assessing whether knowledge of IHL can impact not only attitudes toward abuse, but also actions, represents an important area for further inquiry. Individual preferences remain an important building block for theories seeking to account for mass violence, since they often serve as a necessary basis for decisions ultimately to implement and engage in abuse (Valentino Reference Valentino2000).

The study also centered on a general awareness of IHL, but it did not fully incorporate concerns over the concrete legal obligations involved. A number of studies find that international treaties can shape behavior in wartime, especially for democracies, but only if the country or countries have ratified the prevailing laws of war (Morrow Reference Morrow2014; Wallace Reference Wallace2015). To what extent will knowledge of IHL have different implications for citizens from countries that have committed to a treaty versus those that have not? Unfortunately, the survey evidence presented here is unable to assess whether results are conditional on the ratification status of the respondent’s home country. The Geneva Conventions have been among the most successful instruments regulating the laws of war, enjoying near universal support. There is consequently no variation in ratification status across the countries surveyed—all had ratified the conventions, with most being parties for decades.Footnote 42 Examining knowledge of other instruments exhibiting more varied international support, such as the Ottawa Treaty banning land minesFootnote 43 or the Rome Statute establishing the ICC, may be able to provide greater leverage in this regard. Issues surrounding the ICC also raise questions of the extent to which more salient threats of prosecution and punishment of violators would supplant, or supplement, the more normatively rooted dynamics between IHL and individual attitudes uncovered in this study (Chapman and Chaudoin Reference Chapman and Chaudoin2017). These different avenues collectively provide opportunities to further improve our understanding of the conditions under which international law can influence individual attitudes and contribute toward greater adherence to humanitarian principles.

Supplementary Material

To view supplementary material for this article, please visit


For their helpful comments at various stages, he thanks Kristine Eck, Robert Keohane, Giovanni Mantilla, Krzysztof Pelc, Beth Simmons, Sophia Jordán Wallace, the participants in the 2015 workshop “The Laws of War as an International Regime: History, Theory, and Prospects” at the Niehaus Center for Globalization and Governance at Princeton University, and the three anonymous reviewers. Earlier versions of this article were presented at the 2011 annual meetings of the International Studies Association and the Midwest Political Science Association, as well as at the 2017 annual meeting of the American Political Science Association. All errors remain the author’s own.

1. For instance, each of the four agreements making up the Geneva Conventions contains a specific provision requiring contracting parties to disseminate information about the regulations within their national societies, such as Article 127 of the Third Convention governing prisoners of war.

2. For one exception, see Y. Lupu and Wallace (Reference Lupu and Wallace2018).

3. I use the terms “international humanitarian law” and the “laws of war” interchangeably.

4. For further information, see

5. Though other work finds that dependence on international networks of democratic states can influence belligerent conduct, at least with respect to civilians (Prorok and Appel Reference Prorok and Appel2014).

6. Though for a skeptical take on the more general relationship between engagement in civil society and attitudes toward political violence, see Chapman (Reference Chapman2008).

7. On general challenges of external validity for experimental studies, see McDermott (Reference McDermott2002).

8. For instance, see Blattman (Reference Blattman2009) on Uganda, Grossman, Manekin, and Miodownik (Reference Grossman, Manekin and Miodownik2015) on Israel, and Humphreys and Weinstein (Reference Humphreys and Weinstein2004) on Sierra Leone.

9. The ICRC pretested a pilot study in Mozambique and Colombia, but these are excluded from the analysis because the questionnaire differed markedly from the final version employed for the main set of countries subsequently surveyed. Also included in a separate study were surveys conducted in four of the five permanent members of the UN Security Council (France, Russia, the United Kingdom, and the United States), along with Switzerland, where the ICRC is headquartered. However, the instrument for these countries also differed in many respects for the set of wartime conduct questions asked, along with several other key covariates, both by necessity and design. It is thus difficult to compare results directly for these countries to the main set of conflict and post-conflict countries examined in this study.

10. For further details on the ICRC’s country selection criteria, see ICRC (1999, iv n2).

11. These are the International Criminal Tribunal for the former Yugoslavia (ICTY), established in 1993, and the International Criminal Tribunal for Rwanda (ICTR), begun in 1994.

12. Temporal trends in attitudes toward wartime violence are by no means straightforward. The history of norms surrounding civilian immunity remains fragile, with progressive shifts, but also reversals, across different eras (Bellamy Reference Bellamy2012, 1–3; Barnett 2013, 12–14). Moreover, in some instances, attitudes toward violence may have become more permissive over time, such as support for torture in the US case (Gronke et al. Reference Gronke, Rejali, Drenguisa, Hicksa, Millera and Nakayama2010). This points to the importance of examining more systematically, and in a cross-national context, attitudes toward wartime violence and their relationship to the wider socio-legal environment, in particular the role of international law.

13. In actuality, the ICRC conducted a similar follow-up survey in 2009 as well as 2016, though using different samples of countries and questionnaires with little overlap to the original 1999 study. Even in those countries that were included in each study, such as Afghanistan, respondents from the original survey were not necessarily surveyed in the second wave. Despite some of these limits, it would nonetheless be interesting to be able to compare patterns in responses across the various waves. Unfortunately, the ICRC has no plans currently in place to release the raw data for either the 2009 or 2016 surveys (based on personal correspondence with ICRC staff). For further information on these later studies, see, as well as

14. Other work, such as Elcheroth (Reference Elcheroth2006) and Spini, Elcheroth, and Fasel (Reference Spini, Elcheroth and Fasel2008), also draws on some of these wartime abuse questions. However, both used a more limited subset involving eight of the component items, and they did not differentiate between the nature of the victimized groups. In constructing their outcome variables, both also created a simple additive score rather than conducting a more formal factor analysis.

15. A Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy of 0.75 suggests the component variables have enough in common to justify factor analysis (Kaiser Reference Kaiser1974). Moreover, following Kaiser’s widely used criterion, only this single factor had an eigenvalue greater than 1 (Kim and Mueller Reference Kim and Mueller1978a, 43). Scree plots further confirmed that retaining one sole factor was most appropriate.

16. I rescaled the abuse index for each observation i based on the following formula, which employed the minimum (Min Value) and maximum (Max Value) values for the abuse index observed across the entire set of respondents: (Abusei – Abuse(Min Value))/(Abuse(Max Value) – Abuse(Min Value)). For a similar approach, see Humphreys and Weinstein (Reference Humphreys and Weinstein2006).

17. For a review of the many agreements and instructions seeking to regulate warfare, see

18. More specifically, around 44 percent of respondents answered yes to the first question, indicating that they had heard of the Geneva Conventions. Of these, just over 58 percent correctly identified the purpose of the laws.

19. Additional discussion of the merits of the IHL knowledge variable measure compared to possible alternative constructs is provided in further detail in Supplementary Appendix B.

20. Unlike in other countries, this specific question of whether the war took place in the area where a respondent lived was not asked as part of the Israeli-Palestinian questionnaire. Due to the area’s relatively small geographic size, as well as the nature of the violence in the occupied territories on the one hand and the widely distributed terrorist attacks in Israel on the other, this variable was coded 1 for all respondents. Nevertheless, the overall results do not change substantially if Israeli-Palestinian respondents are excluded from the analysis.

21. There are six categories: 18–24 years, 25–34 years, 35–44 years, 45–54 years, 55–64 years, and 65 or older. Following usual human subjects concerns, the survey was intended to be conducted only with individuals 18 years of age or older. However, a small number of observations involve respondents who reported ages less than 18. This may have been the result of misinterpreting the question, or simply errors in data entry, but all these cases were excluded.

22. Results for education remain largely the same when substituting the original continuous measure for years of schooling.

23. The different question wordings for relevant items are included in Supplementary Appendix A. Balance tests indicate that the two groups were comparable across a range of baseline covariates.

24. A Kruskal-Wallis equality-of-populations rank test was conducted and the null hypothesis that the two samples were drawn from the same population could not be rejected (p = 0.58).

25. Supplementary Appendix B also reports summary statistics separately for each of the countries surveyed.

26. South Africa represented the excluded reference category. To ensure the results were not driven by any particular country, the analysis was also rerun excluding each of the countries in turn, and the results remained substantially the same. A parallel study was also fielded with a local survey firm in two of the countries included in the study—Bosnia-Herzegovina and the Philippines—to investigate any potential role for enumerator identity. When substituting the parallel responses for the relevant countries, the results do not substantially change.

27. All analysis was performed using Stata 14. For ease of presentation, coefficients for the partisan wording and country indicators are not displayed.

28. Predicted values calculated using Stata’s -margins- command.

29. Knowledge of the Geneva Conventions and education are also only moderately correlated (r = 0.25).

30. Full results available from the author upon request.

31. As a further robustness check, the analysis was rerun using multiple imputation to deal with missing values across several of the variables. The results remained substantially the same and are discussed in further detail in Supplementary Appendix C.

32. To ensure that the results from having children were not simply a function of being married, I included an indicator for whether the respondent was married. The results for having children do not change significantly, while the coefficient for being married on its own does not attain standard thresholds of statistical significance.

33. Conditional relationships were also estimated in a series of models, including an interaction term between knowledge of the Geneva Conventions and the relevant individual characteristic. The results remained substantially the same.

34. Wald tests for equality of coefficients for the IHL knowledge variable across different subgroups were generated by first estimating a seemingly unrelated regression model incorporating relevant subsample models.

35. Analysis conducted using the Matching package in R. Supplementary Appendix C reports the full results from the matching procedure. Results also do not change substantially when using alternative matching routines, such as nearest-neighbor matching using the Mahalanobis distance metric.

36. Unfortunately, this assumption of selection on observables cannot be tested directly. However, I performed a sensitivity analysis based on Rosenbaum’s guidelines using the R package rbounds (Keele n.d.). The procedure estimates the degree to which an unobserved confounder would have to change the odds (denoted by Γ) of receiving the treatment compared to the control to cancel out the observed effect estimated for the treatment (here, knowledge of the Geneva Conventions). Looking at the largest finding for prisoner abuse, the analysis suggests that the difference in support for abuse due to knowledge of the Geneva Conventions is unchanged up to a Γ value of 1.6. The corresponding value for the overall abuse measure is Γ = 1.4 (civilian abuse is not reported since it did not yield any statistically significant first difference). This means that an observed confounder would need to change the odds of being knowledgeable about the Geneva Conventions by a factor of 60 percent to overturn the finding for prisoner abuse (and correspondingly 40 percent for overall abuse).

37. A number of strategies for designing and implementing surveys have been proposed to mitigate social desirability bias, from self-administration of questionnaires to employing list experiments and other randomized response techniques (Nederhof Reference Nederhof1985; Blair, Imai, and Lyall Reference Blair, Imai and Lyall2014). Implementing such strategies is beyond the scope of the current article, which relies on existing survey evidence, but it would be a valuable avenue in future research.

38. As in other models, predicted probabilities were generated using Stata’s -margins- command. First differences were calculated holding all other variables constant at their median values. This question was asked only of respondents who answered in a prior item that they believed rules existed for which violators should be held accountable. For the prior outcome, those with knowledge of the Geneva Conventions were also more likely to believe such limits exist (p < 0.01).

39. The logit model exhibited problems of separation where some country indicators (Afghanistan, Bosnia-Herzegovina, Georgia, and Israel) perfectly predicted the binary outcome, resulting in those observations being dropped. A penalized likelihood correction model offers a solution that addresses separation problems (Zorn Reference Zorn2005). As an alternative, I employed Firth’s penalized logit estimator on the full set of country data and results remain substantially the same.

40. The law and human rights answer choices overlap conceptually in the survey and were thus combined.

41. See For similar initiatives by the International Institute of Humanitarian Law, see Some of these initiatives are controversial (Massoud Reference Massoud2011).

42. Even for two of the countries who formally ratified most recently—Bosnia-Herzegovina (1992) and Georgia (1993)—their predecessor states (Yugoslavia and the Soviet Union, respectively) had committed to the conventions years before.

43. The full title is the Convention on the Prohibition of the Use, Stockpiling, Production and Transfer of Anti-Personnel Mines and on their Destruction.


Axelrod, Robert. The Evolution of Cooperation. New York: Basic Books, 1984.Google Scholar
Balcells, Laia. “Rivalry and Revenge: Violence Against Civilians in Conventional Civil Wars.” International Studies Quarterly 54, no. 2 (2010): 291313.CrossRefGoogle Scholar
Balcells, Laia. “The Consequences of Victimization on Political Identities: Evidence from Spain.” Politics and Society 40, no. 3 (2012): 311–47.CrossRefGoogle Scholar
Bandura, Albert. “Moral Disengagement in the Perpetration of Inhumanities.” Personal and Social Psychology Review 3, no. 3 (1999): 193209.CrossRefGoogle ScholarPubMed
Barnett, Michael. Empire of Humanity: A History of Humanitarianism. Ithaca, NY: Cornell University Press, 2013.Google Scholar
Bar-Tal, Daniel. “The Necessity of Observing Real Life Situations: Palestinian-Israeli Violence as a Laboratory for Learning About Social Behaviour.” European Journal of Social Psychology 35, no. 6 (2004): 677701.CrossRefGoogle Scholar
Beber, Bernd, Roessler, Philip, and Scacco, Alexandra. “Intergroup Violence and Political Attitudes: Evidence from a Dividing Sudan.” Journal of Politics 76, no. 3 (2014): 649–65.CrossRefGoogle Scholar
Bellamy, Alex J. Responsibility to Protect: The Global Effort to End Mass Atrocities. Cambridge: Polity Press, 2009.Google Scholar
Bellamy, Alex J. Massacres and Morality: Mass Atrocities in an Age of Civilian Immunity. Oxford: Oxford University Press, 2012.CrossRefGoogle Scholar
Bellows, John, and Miguel, Edward. “War and Local Collective Action in Sierra Leone.” Journal of Public Economics 93, no. 11–12 (2009): 1144–57.CrossRefGoogle Scholar
Berrebi, Claude, and Klor, Esteban F.. “Are Voters Sensitive to Terrorism? Direct Evidence from the Israeli Electorate.” American Political Science Review 102, no. 3 (2008): 279301.CrossRefGoogle Scholar
Best, Geoffrey. Humanity in Warfare: The Modern History of the International Law of Armed Conflicts. New York: Columbia University Press, 1980.Google Scholar
Best, Geoffrey. War and Law Since 1945. Oxford: Clarendon Press, 1994.Google Scholar
Blair, Graeme, Imai, Kosuke, and Lyall, Jason. “Comparing and Combining List and Endorsement Experiments: Evidence from Afghanistan.” American Journal of Political Science 58, no. 4 (2014): 1043–63.CrossRefGoogle Scholar
Blattman, Christopher. “From Violence to Voting: War and Political Participation in Uganda.” American Political Science Review 103, no. 2 (2009): 231–47.CrossRefGoogle Scholar
Brockhoff, Sarah, Krieger, Tim, and Meierrieks, Daniel. “Great Expectations and Hard Times: The (Nontrivial) Impact of Education on Domestic Terrorism.” Journal of Conflict Resolution 59, no. 7 (2015): 11861215.CrossRefGoogle Scholar
Brooks, Deborah J., and Valentino, Benjamin A.. “A War of One’s Own: Understanding the Gender Gap in Support for War.” Public Opinion Quarterly 75, no. 2 (2011): 270–86.CrossRefGoogle Scholar
Browning, Christopher R. Ordinary Men: Reserve Police Battalion 101 and the Final Solution in Poland. New York: Harper, 1998.Google Scholar
Bussman, Margit, and Schneider, Gerald. “A Porous Humanitarian Shield: The Laws of War, the Red Cross, and the Killing of Civilians.” Review of International Organizations 11, no. 3 (2016): 337–59.CrossRefGoogle Scholar
Canetti-Nisim, Daphna, Halperin, Eran, Sharvit, Keren, and Hobfoll, Stevan E.. “A New Stress-Based Model of Political Extremism: Personal Exposure to Terrorism, Psychological Distress, and Exclusionist Political Attitudes.” Journal of Conflict Resolution 53, no. 3 (2009): 363–89.CrossRefGoogle ScholarPubMed
Cardenas, Sonia. Conflict and Compliance: State Responses to International Human Rights Pressure. Philadelphia, PA: University of Pennsylvania Press, 2007.CrossRefGoogle Scholar
Chapman, Terrence L.Unraveling the Ties Between Civic Institutions and Attitudes Toward Political Violence.” International Studies Quarterly 52, no. 3 (2008): 515–32.CrossRefGoogle Scholar
Chapman, Terrence L., and Chaudoin, Stephen. “Public Reactions to International Legal Institutions: The ICC in a Developing Democracy.” Unpublished manuscript, University of Texas at Austin and University of Illinois, 2017.Google Scholar
Chaudoin, Stephen. “Promises or Policies? An Experimental Analysis of International Agreements and Audience Reactions.” International Organization 68, no. 1 (2014): 235–56.CrossRefGoogle Scholar
Child, Travers B., and Nikolova, Elena. “War and Social Attitudes.” Conflict Management and Peace Science (forthcoming).Google Scholar
Chilton, Adam S.The Laws of War and Public Opinion: An Experimental Study.” Journal of Institutional and Theoretical Economics 171, no. 1 (2015): 181201.CrossRefGoogle Scholar
Collier, Paul, and Hoeffler, Anke. “Greed and Grievance in Civil War.” Oxford Economic Papers 56, no. 4 (2004): 563–95.CrossRefGoogle Scholar
Conrad, Courtenay R., and Ritter, Emily H.. “Treaties, Tenure, and Torture: The Conflicting Domestic Effects of International Law.” Journal of Politics 75, no. 2 (2013): 397409.CrossRefGoogle Scholar
Dai, Xinyuan. International Institutions and National Policies. Cambridge: Cambridge University Press, 2007.CrossRefGoogle Scholar
Davenport, Christian. State Repression and the Domestic Democratic Peace. Cambridge: Cambridge University Press, 2007.CrossRefGoogle Scholar
Diamond, Alexis, and Sekhon, Jasjeet S.. “Genetic Matching for Estimating Causal Effects: A General Multivariate Matching Method for Achieving Balance in Observational Studies.” Review of Economics and Statistics 95, no. 3 (2013): 932–45.CrossRefGoogle Scholar
Downes, Alexander B. Targeting Civilians in War. Ithaca, NY: Cornell University Press, 2008.Google Scholar
Eichenberg, Richard C.Gender Differences in Public Attitudes Toward the Use of Force by the United States, 1990–2003.” International Security 28, no. 1 (2003): 110–41.CrossRefGoogle Scholar
Elcheroth, Guy. “Individual-Level and Community-Level Effects of War Trauma on Social Representations Related to Humanitarian Law.” European Journal of Social Psychology 36, no. 6 (2006): 907–30.CrossRefGoogle Scholar
Evangelista, Matthew, and Tannenwald, Nina, eds. Do the Geneva Conventions Matter? Oxford: Oxford University Press, 2017.CrossRefGoogle Scholar
Fariss, Christopher J.Respect for Human Rights Has Improved Over Time: Modeling the Changing Standard of Accountability.” American Political Science Review 108, no. 2 (2014): 297318.CrossRefGoogle Scholar
Feaver, Peter D., and Gelpi, Christopher. Choosing Your Battles: American Civil-Military Relations and the Use of Force. Princeton, NJ: Princeton University Press, 2004.Google Scholar
Feaver, Peter D., and Kohn, Richard H., eds. Soldiers and Civilians: The Civil-Military Gap and American National Security. Cambridge, MA: MIT University Press, 2001.Google Scholar
Franck, Thomas M. The Power of Legitimacy Among Nations. New York: Oxford University Press, 1990.Google Scholar
Freitag, Markus, Kijewski, Sara, and Oppold, Malvin. “War Experiences, Economic Grievances, and Political Participation in Postwar Societies: An Empirical Analysis of Kosovo.” Conflict Management and Peace Science (forthcoming).Google Scholar
Fuller, Lon L. The Morality of Law. New Haven, CT: Yale University Press, 1969.Google Scholar
Getmansky, Anna, and Zeitzoff, Thomas. “Terrorism and Voting: The Effect of Rocket Threat on Voting in Israeli Elections.” American Political Science Review 108, no. 3 (2014): 588604.CrossRefGoogle Scholar
Gilligan, Michael J., Pasquale, Benjamin J., and Samii, Cyrus. “Civil War and Social Cohesion: Lab-in-the-Field Evidence from Nepal.” American Journal of Political Science 58, no. 3 (2014): 604–19.CrossRefGoogle Scholar
Goldhagen, Daniel J. Hitler’s Willing Executioners: Ordinary Germans and the Holocaust. New York: Knopf, 1996.Google Scholar
Gronke, Paul, Rejali, Darius, Drenguisa, Dustin, Hicksa, James, Millera, Peter, and Nakayama, Bryan. “U.S. Public Opinion on Torture, 2001–2009.” PS: Political Science and Politics 43, no. 3 (2010): 437–44.Google Scholar
Grossman, Guy, Manekin, Devorah, and Miodownik, Dan. “The Political Legacies of Combat: Attitudes Towards War and Peace Among Israeli Ex-Combatants.” International Organization 69, no. 4 (2015): 9811009.CrossRefGoogle Scholar
Hainmueller, Jens, and Hiscox, Michael J.. “Learning to Love Globalization: Education and Individual Attitudes Toward International Trade.” International Organization 60, no. 2 (2006): 469–98.CrossRefGoogle Scholar
Hainmueller, Jens, and Hiscox, Michael J.. “Educated Preferences: Explaining Attitudes Toward Immigration in Europe.” International Organization 61, no. 2 (2007): 399442.CrossRefGoogle Scholar
Ho, Daniel E., Imai, Kosuke, King, Gary, and Stuart, Elizabeth A.. “MatchIt: Nonparametric Preprocessing for Parametric Causal Inference.” Journal of Statistical Software 42, no. 8 (2007): 128.Google Scholar
Holsti, Ole R. Public Opinion and American Foreign Policy. Ann Arbor, MI: University of Michigan Press, 2004.Google Scholar
Hong, Ji Yeon, and Kang, Woo Chang. “Trauma and Stigma: The Long-Term Effects of Wartime Violence on Political Attitudes.” Conflict Management and Peace Science 34, no. 3 (2017): 264286.CrossRefGoogle Scholar
Horowitz, Michael C., and Levendusky, Matthew S.. “Drafting Support for War: Conscription and Mass Support for Warfare.” Journal of Politics 73, no. 2 (2011): 111.CrossRefGoogle Scholar
Howard, Michael, Andreopoulos, George J., and Shulman, Mark R., eds. The Laws of War: Constraints on Warfare in the Western World. New Haven, CT: Yale University Press, 1994.Google Scholar
Huddy, Leonie, Feldman, Stanley, Taber, Charles, and Lahav, Gallya. “Threat, Anxiety, and Support of Antiterrorism Policies.” American Journal of Political Science 49, no. 3 (2005): 593608.CrossRefGoogle Scholar
Humphreys, Macartan, and Weinstein, Jeremy M.. What the Fighters Say: A Survey of Ex-Combatants in Sierra Leone June–August 2003. New York: Earth Institute, Colombia University, 2004.Google Scholar
Humphreys, Macartan, and Weinstein, Jeremy M.. “Handling and Manhandling Civilians in Civil War.” American Political Science Review 100, no. 3 (2006): 429–47.CrossRefGoogle Scholar
International Committee of the Red Cross. The People on War Report: ICRC Worldwide Consultation on the Rules of War. Geneva: ICRC, 1999.Google Scholar
Jha, Saumitra, and Wilkinson, Steven I.. “Does Combat Experience Foster Organizational Skill? Evidence from Ethnic Cleansing During the Partition of South Asia.” American Political Science Review 106, no. 4 (2012): 883907.CrossRefGoogle Scholar
Jo, Hyeran. Compliant Rebels: Rebel Groups and International Law in World Politics. Cambridge: Cambridge University Press, 2015.CrossRefGoogle Scholar
Kahl, Colin H.In the Crossfire or the Crosshairs? Norms, Civilian Casualties, and U.S. Conduct in Iraq.” International Security 32, no. 1 (2007): 746.CrossRefGoogle Scholar
Kaiser, Henry F.An Index of Factorial Simplicity.” Psychometrika 39, no. 1 (1974): 3136.CrossRefGoogle Scholar
Kalyvas, Stathis N.Wanton and Senseless? The Logic of Violence in Civil Wars.” Rationality and Society 11, no. 3 (1999): 243–85.CrossRefGoogle Scholar
Kalyvas, Stathis N. The Logic of Violence in Civil War. Cambridge: Cambridge University Press, 2006.CrossRefGoogle Scholar
Kalyvas, Stathis N. “Ethnic Defection in Civil War.” Comparative Political Studies 41, no. 8 (2008): 1043–68.CrossRefGoogle Scholar
Kaufmann, Chaim. “Possible and Impossible Solutions to Ethnic Civil Wars.” International Security 20, no. 4 (1996): 136–75.CrossRefGoogle Scholar
Keele, Luke J. “Rbounds: An R Package for Sensitivity Analysis with Matched Data.” n.d.Google Scholar
Kim, Jae-On, and Mueller, Charles W.. Factor Analysis: Statistical Methods and Practical Issues. Beverly Hills, CA: Sage, 1978a.CrossRefGoogle Scholar
Kim, Jae-On, and Mueller, Charles W.. Introduction to Factor Analysis: What it Is and How to Do It. Beverly Hills, CA: Sage, 1978b.CrossRefGoogle Scholar
King, Elizabeth. From Classrooms to Conflict in Rwanda. Cambridge: Cambridge University Press, 2013.CrossRefGoogle Scholar
Kinsella, Helen M.Discourses of Difference: Civilians, Combatants, and Compliance with the Laws of War.” Review of International Studies 31, Supplement S1 (2005): 163–85.CrossRefGoogle Scholar
Lasley, Trace, and Thyne, Clayton L.. “Secession, Legitimacy and the Use of Child Soldiers.” Conflict Management and Peace Science 32, no. 3 (2015): 289308.CrossRefGoogle Scholar
Linos, Katerina. The Democratic Foundations of Policy Diffusion: How Health, Family, and Employment Laws Spread Across Countries. Oxford: Oxford University Press, 2013.CrossRefGoogle Scholar
Lupu, Noam, and Peisakhin, Leonid. “The Legacy of Political Violence Across Generations.” American Journal of Political Science 61, no. 4 (2017): 836851.CrossRefGoogle Scholar
Lupu, Yonatan. “The Informative Power of Treaty Commitment: Using the Spatial Model to Address Selection Effects.” American Journal of Political Science 57, no. 4 (2013): 912–25.Google Scholar
Lupu, Yonatan, and Wallace, Geoffrey P. R.. “Violence, Non-Violence, and the Effects of International Human Rights Law.” Unpublished manuscript, George Washington University and University of Washington, 2018.Google Scholar
Lyall, Jason. “Are Coethnics More Effective Counterinsurgents? Evidence from the Second Chechen War.” American Political Science Review 104, no. 1 (2010): 120.CrossRefGoogle Scholar
Maoz, Ifat, and McCauley, Clark. “Threat, Dehumanization, and Support for Retaliatory Aggressive Policies in Asymmetric Conflict.” Journal of Conflict Resolution 52, no. 1 (2008): 93116.CrossRefGoogle Scholar
Martin, Lisa L., and Simmons, Beth A.. “Theories and Empirical Studies of International Institutions.” International Organization 52, no. 4 (1998): 729–58.CrossRefGoogle Scholar
Massoud, Mark F.Do Victims of War Need International Law? Human Rights Education Programs in Authoritarian Sudan.” Law and Society Review 45, no. 1 (2011): 132.CrossRefGoogle Scholar
McDermott, Rose. “Experimental Methods in Political Science.” Annual Review of Political Science 5 (2002): 3161.CrossRefGoogle Scholar
Mearsheimer, John. “The False Promise of International Institutions.” International Security 19, no. 3 (1994–1995): 549.CrossRefGoogle Scholar
Meernik, James, and King, Kimi. “A Psychological Jurisprudence Model of Public Opinion and International Prosecution.” International Area Studies Review 17, no. 1 (2014): 320.CrossRefGoogle Scholar
Merolla, Jennifer L., and Zechmeister, Elizabeth J.. Democracy at Risk: How Terrorist Threats Affect the Public. Chicago, IL: Chicago University Press, 2009.CrossRefGoogle Scholar
Merom, Gil. How Democracies Lose Small Wars: State, Society, and the Failures of France in Argelia, Israel in Lebanon, and the United States in Vietnam. Cambridge: Cambridge University Press, 2003.CrossRefGoogle Scholar
Montada, Leo, and Lerner, Melvin J., eds. Responses to Victimizations and Belief in a Just World. New York: Springer, 1998.CrossRefGoogle Scholar
Morrow, James D.When Do States Follow the Laws of War?American Political Science Review 101, no. 3 (2007): 559–72.CrossRefGoogle Scholar
Morrow, James D. Order Within Anarchy: The Laws of War as an International Institution. New York: Cambridge University Press, 2014.CrossRefGoogle Scholar
Morrow, James D., and Jo, Hyeran. “Compliance with the Laws of War: Dataset and Coding Rules.” Conflict Management and Peace Science 23, no. 1 (2006): 91113.CrossRefGoogle Scholar
Nederhof, Anton J.Methods of Coping with Social Desirability Bias: A Review.” European Journal of Social Psychology 15, no. 3 (1985): 263–80.CrossRefGoogle Scholar
Podsakoff, Philip M., MacKenzie, Scott B., Lee, Jeong-Yeon, and Podsakoff, Nathan P.. “Common Method Biases in Behavioral Research: A Critical Review of the Literature and Recommended Remedies.” Journal of Applied Psychology 88, no. 5 (2003): 879903.CrossRefGoogle ScholarPubMed
Poe, Steven C., and Tate, C. Neal. “Repression of Human Rights to Personal Integrity in the 1980s: A Global Analysis.” American Political Science Review 88, no. 4 (1994): 853–72.CrossRefGoogle Scholar
Posen, Barry R.Nationalism, the Mass Army, and Military Power.” International Security 18, no. 2 (1993): 80124.CrossRefGoogle Scholar
Prorok, Alyssa K., and Appel, Benjamin J.. “Compliance with International Humanitarian Law: Democratic Third Parties and Civilian Targeting in Interstate War.” Journal of Conflict Resolution 58, no. 4 (2014): 713–40.CrossRefGoogle Scholar
Reiter, Dan, and Stam, Allan C.. Democracies at War. Princeton, NJ: Princeton University Press, 2002.CrossRefGoogle Scholar
Rejali, Darius. Torture and Democracy. Princeton, NJ: Princeton University Press, 2007.Google Scholar
Rohner, Dominic, Thoenig, Mathias, and Zilibotti, Fabrizio. “Seeds of Distrust: Conflict in Uganda.” Journal of Economic Growth 18, no. 3 (2013): 217–52.CrossRefGoogle Scholar
Rosenbaum, Paul R. Design of Observational Studies. New York: Springer, 2010.CrossRefGoogle ScholarPubMed
Rummel, R. J. Death by Government. New Brunswick, NJ: Transaction, 1994.Google Scholar
Schabas, William A. An Introduction to the International Criminal Court. Cambridge: Cambridge University Press, 2011.CrossRefGoogle Scholar
Simmons, Beth A. Mobilizing for Human Rights: International Law in Domestic Politics. Cambridge: Cambridge University Press, 2009.CrossRefGoogle Scholar
Simmons, Beth A. “Treaty Compliance and Violation.” Annual Review of Political Science 13 (2010): 272–96.CrossRefGoogle Scholar
Simmons, Beth A., and Hopkins, Daniel J.. “The Constraining Power of International Treaties: Theory and Methods.” American Political Science Review 99, no. 4 (2005): 623–31.CrossRefGoogle Scholar
Smith, Hugh. “What Costs Will Democracies Bear? A Review of Popular Theories of Casualty Aversion.” Armed Forces and Society 31, no. 4 (2005): 487512.CrossRefGoogle Scholar
Solis, Gary D. The Law of Armed Conflict: International Humanitarian Law in War. Cambridge: Cambridge University Press, 2010.CrossRefGoogle Scholar
Spini, Dario, Elcheroth, Guy, and Fasel, Rachel. “The Impact of Group Norms and Generalization of Risks Across Groups on Judgments of War Behavior.” Political Psychology 29, no. 6 (2008): 919–41.CrossRefGoogle Scholar
Stanton, Jessica A. Violence and Restraint in Civil War: Civilian Targeting in the Shadow of International Law. Cambridge: Cambridge University Press, 2016.CrossRefGoogle Scholar
Straus, Scott. The Order of Genocide: Race, Power, and War in Rwanda. Ithaca, NY: Cornell University Press, 2006.Google Scholar
Thomas, Daniel C. The Helsinki Effect: International Norms, Human Rights, and the Demise of Communism. Princeton, NJ: Princeton University Press, 2002.Google Scholar
Thyne, Clayton L.ABC’s, 123’s, and the Golden Rule: The Pacifying Effect of Education on Civil War, 1980–1999.” International Studies Quarterly 50, no. 4 (2006): 733–54.CrossRefGoogle Scholar
Tomz, Michael. “Reputation and the Effect of International Law on Preferences and Beliefs.” Unpublished manuscript, Stanford University, 2008.Google Scholar
Tyler, Tom R. Why People Obey the Law. New Haven, CT: Yale University Press, 1990.Google Scholar
Urdal, Henrik. “A Clash of Generations? Youth Bulges and Political Violence.” International Studies Quarterly 50, no. 3 (2006): 607–29.CrossRefGoogle Scholar
Valentino, Benjamin A. “Final Solutions: The Causes of Mass Killing and Genocide.” Security Studies 9, no. 3 (2000): 159.CrossRefGoogle Scholar
Valentino, Benjamin A. Final Solutions: Mass Killing and Genocide in the Twentieth Century. Ithaca, NY: Cornell University Press, 2004.Google Scholar
Valentino, Benjamin A., Huth, Paul K., and Croco, Sarah E.. “Covenants Without the Sword: International Law and the Protection of Civilians in Times of War.” World Politics 58, no. 3 (2006): 339–77.CrossRefGoogle Scholar
Von Stein, Jana. “Do Treaties Constrain or Screen? Selection Bias and Treaty Compliance.” American Political Science Review 99, no 4 (2005): 611–22.CrossRefGoogle Scholar
Voors, Maarten J., Nillesen, Eleonora E. M., Verwimp, Philip, Bulte, Erwin H., Lensink, Robert, and van Soest, Daan P.. “Violent Conflict and Behavior: A Field Experiment in Burundi.” American Economic Review 102, no. 2 (2012): 941–64.CrossRefGoogle Scholar
Wallace, Geoffrey P. R.Welcome Guests, or Inevitable Victims? The Causes of Prisoner Abuse in War.” Journal of Conflict Resolution 56, no. 6 (2012): 955–81.CrossRefGoogle Scholar
Wallace, Geoffrey P. R. “International Law and Public Attitudes Toward Torture: An Experimental Study.” International Organization 67, no. 1 (2013): 105–40.CrossRefGoogle Scholar
Wallace, Geoffrey P. R. “Martial Law? Military Experience, International Law, and Support for Torture.” International Studies Quarterly 58, no. 3 (2014): 501–14.CrossRefGoogle Scholar
Wallace, Geoffrey P. R. Life and Death in Captivity: The Abuse of Prisoners during War. Ithaca, NY: Cornell University Press, 2015.Google Scholar
Walzer, Michael. Just and Unjust Wars. New York: Basic Books, 2000.Google Scholar
Weinstein, Jeremy M. Inside Rebellion: The Politics of Insurgent Violence. Cambridge: Cambridge University Press, 2007.Google Scholar
Winship, Christopher, and Morgan, Stephen L.. “The Estimation of Causal Effects from Observational Data.” Annual Review of Sociology 25 (1999): 659706.CrossRefGoogle Scholar
Zorn, Christopher. “A Solution to Separation in Binary Response Models.” Political Analysis 13, no. 2 (2005): 157–70.CrossRefGoogle Scholar
Figure 0

Table 1. Questions Used to Construct Support for Wartime Abuse Indices

Figure 1

Figure 1. Summary of Support for Wartime Abuse Indices

Note: Histogram for each relevant abuse index is overlaid with a kernel density plot (black line).
Figure 2

Figure 2. Cumulative Distribution Functions of Support for Wartime Abuse Indices

Figure 3

Table 2. Summary Statistics for Main Variables Used in the Analysis

Figure 4

Figure 3. Substantive Effects of Support for Wartime Abuse

Note: Coefficient plot (OLS) indicating percentage point change in support for each abuse index. Lines indicate 95 percent confidence intervals. Results for partisan wording and country variables not shown.
Figure 5

Figure 4. Knowledge of Geneva Conventions and Support for Wartime Abuse

Note: Plot for Knowledge of Geneva Conventions coefficient (OLS) indicating percentage point change in support for each abuse index by relevant subgroup. Lines indicate 95 percent confidence intervals.
Figure 6

Figure 5. Knowledge of Geneva Conventions and Support for Wartime Abuse by Level of Education

Note: Plot for Knowledge of Geneva Conventions coefficient (OLS) indicating percentage point change in support for each abuse index by relevant subgroup. Lines indicate 95 percent confidence intervals.
Figure 7

Figure 6. Knowledge of the Geneva Conventions and Beliefs on the Basis of Rules for Which Violators Should Be Held Accountable

Note: Multinomial logit estimation. Average marginal effect (percentage point change in the predicted probability of the outcome) of Knowledge of Geneva Conventions, holding all other variables at their median values. Lines indicate 95 percent confidence intervals. Model is limited to respondents who believed there were rules for which people should be held accountable if they violated them.
Figure 8

Figure 7. Knowledge of Geneva Conventions and Beliefs of Limits on Wartime Conduct

Note: Logit estimation separately for each outcome. Average marginal effect (percentage point change in the predicted probability of the outcome) of Knowledge of Geneva Conventions, while holding all other variables at their median values. Lines indicate 95 percent confidence intervals. Model for ‘Why Limits?’ limited to respondents who answered affirmatively to prior question on ‘Any Limits on Fighting?’
Figure 9

Figure 8. Knowledge of Geneva Conventions and Beliefs for Why Breaking Limits on Wartime Conduct Is Wrong

Note: Logit estimation separately for each outcome. Average marginal effect (percentage point change in the predicted probability of the outcome) of Knowledge of Geneva Conventions, while holding all other variables at their median values. Lines indicate 95 percent confidence intervals. Models limited to respondents who answered ‘It’s wrong’ on previous item of ‘Why are there limits on fighting?’
Supplementary material: File

Wallace supplementary material

Appendices A-C

Download Wallace supplementary material(File)
File 104.9 KB