Welcome Guests,or Inescapable Victims?

Abstract

The treatment of prisoners varies enormously across wars. Why are some prisoners horribly abused, while others are cared for humanely? The author argues key attributes of the belligerents, alongside the nature of the conflict itself, provides the most convincing explanation for differences in prisoner abuse. Democratic norms and domestic institutional incentives lead democracies to exhibit more restraint when dealing with prisoners. On the other hand, states caught up in drawn-out wars of attrition, or those seeking territorial conquest, are much more likely to resort to prisoner abuse. The author tests this argument against a variety of common alternative explanations using a new data set on prisoner abuse across all interstate wars from 1898 to 2003. The author finds strong support for the role of both the regime type and the nature of the conflict, while the results also suggest several points of difference from existing research on wartime conduct.

Keywords

prisoners of war prisoner abuse democracy

Winston Churchill offers a blunt definition for a prisoner of war (POW) as a “man who tries to kill you and fails, and then asks you not to kill him” (Tsouras 2000, 380).¹ Given the harsh realities often prevailing during wartime, states may be especially prone to abusing captured enemy soldiers. The brutal treatment inflicted upon prisoners falling into the hands of many belligerents during the Second World War shows how the prospects for captives are often quite bleak. Around 60 percent of Soviet soldiers captured by Nazi Germany would not live to see the end of the war with several becoming the first victims of the Auschwitz gas chambers (Bartov 2001, 107; Streit 1986, 12).

Taking a broader view of prisoner treatment over the past century reveals some perplexing patterns, which belies the notion that abuse is inescapable. The Russo-Japanese War of 1904–05 shares some striking similarities to several fronts during the Second World War, involving autocracies fighting foes of different races. Yet, the treatment of prisoners could not be more disparate as both sides cared for captives relatively humanely (Hata 1996, 262). In interstate wars since the turn of the twentieth century, just under one-quarter of autocratic belligerents have actually provided prisoners fairly good treatment. The Russo-Japanese War also highlights some intriguing differences in prisoner abuse over time. Japanese treatment of captives in this conflict and the First World War contrasted with subsequent increasingly brutal conduct, which culminated in systematic torture, backbreaking labor, and horrific living conditions endured by prisoners during the Second World War (Roland 1996, 165).

Although some of the most notorious cases of prisoner abuse involved autocratic perpetrators, taking stock of past conflicts suggests democracy is far from a panacea. As revelations continue to surface regarding US abuses of detainees at Abu Ghraib, Guantanamo, and elsewhere, democracies appear no more immune from desires to abuse prisoners. The War on Terror is far from an isolated instance, as French forces resorted to torture and other harsh tactics when confronting rebels during the 1954–62 Algerian War (Horne 2006, 195-97). In a certain symmetry to the pattern for autocracies, one-quarter of democratic belligerents resorted to fairly brutal levels of prisoner abuse during twentieth-century interstate warfare. Perusing past conflicts thus points to some puzzling trends in prisoner abuse, which calls for more systematic examination. What then explains differences in the treatment of prisoners during war?

Studies of prisoner abuse are limited for several reasons. First, much research on wartime conduct focuses on violence against civilians, given they represent the most visible victims of conflict. Surrendering soldiers, however, are entitled to similar protections, and prisoner abuse is considered a fundamental violation of the norms and laws of war (Walzer 2000, 46-47). The suffering and deaths attributed to prisoner abuse often constitute a substantial proportion of wartime fatalities—almost one-third of Soviet combatant deaths during the Second World War took place in captivity rather than during battle.²

A second and more pragmatic reason prisoner abuse is often overlooked involves the lack of adequate data (Downes 2006, 157n43). Most existing work comes from historical accounts, which tend to focus on a handful of cases, or a single country over time, but with less concern with identifying broader trends in prisoner abuse (e.g., Speed 1990; Springer 2010). In order to fill this gap, I construct a new data set on the treatment of prisoners across all interstate wars from 1898 to 2003 to provide a more comprehensive assessment of the determinants of prisoner abuse.

Although the literature seeking to explain prisoner treatment is relatively sparse, three factors are most commonly cited: whether both sides hold enemy prisoners, which conditions their conduct through threats of retaliation (MacKenzie 1994); whether the belligerents have ratified the prevailing laws of war (Morrow 2007); and whether the adversary is considered to be fundamentally beyond the pale of civilization for either racist or religious reasons, and thus undeserving of humane treatment (Dower 1986, 11).

I show this conventional wisdom to be incomplete and only partially correct. Although the risk of retaliation, treaty ratification, and perceptions of the adversary’s identity play some role, they do not reliably account for how badly states treat prisoners. Drawing on research by scholars in the related field of civilian victimization and mass killing in war, I argue key factors instead have to do with the nature of the fighting, war aims, and regime type. Specifically, states are more likely to resort to high levels of prisoner abuse in protracted wars of attrition, or when seeking to annex enemy territory. On the other hand, democracy lessens the severity of prisoner abuse, despite notable exceptions and several conditional effects. The democracy finding is particularly intriguing in light of past research on civilian victimization. While some work shows democracy acts as a similar constraint (Harff 2003; Valentino, Huth, and Balch-Lindsay 2004), recent studies find regime type either has no effect or democracies are actually more likely to target civilians (Downes 2007; Valentino, Huth, and Croco 2006). Studying the determinants of prisoner abuse thus contributes to a growing body of scholarship examining the behavior of belligerents during war, and points to the need to consider different issue areas in wartime conduct.

The article proceeds as follows: the first section develops a theoretical framework for the determinants of prisoner abuse by drawing on the broader literature examining conduct during war. The second section then discusses the data and measurement issues for the main variables of interest. In the third section, I present the findings from the empirical analysis. I then conclude by assessing implications for understanding the treatment of prisoners, as well as the general study of wartime conduct.

Theories of Prisoner Abuse in War

Scholars have generated a number of theories to explain the resort to abuse during war. Much of the existing research investigates the targeting of civilians but provides several implications for prisoner abuse. The treatment of prisoners presents a dilemma for captors: although abuse may provide tangible benefits, such as valuable intelligence or coercing the adversary, it often involves significant drawbacks, including domestic or international condemnation and likely retaliation (Gross 2010, 122-44).³ The remainder of this section discusses some of the main factors influencing states when deciding how to treat prisoners.

Deterrence and Fears of Retaliation

States are often concerned their actions may lead adversaries to respond in unwanted ways. Abusing prisoners almost certainly increases the likelihood the adversary will commit equal or even higher levels of abuse in retaliation (Streit 1986, 9). When each side has the ability to hurt the other, reciprocity becomes a defining element characterizing their relations. This can lead to cooperative understandings, such as the “live-and-let-live” system that developed between some groups of soldiers during the First World War (Axelrod 1984, 73-87).

Similar restraint is evident in numerous instances of prisoner treatment. Prisoners often serve a comparable function to the exchange of hostages during earlier periods of warfare, ensuring each side would abide by agreed upon rules of war (Elliott 1995). The so-called Shackling Crisis between Britain and Germany during the Second World War illustrates the dangers of retaliation, and consequently the limits states often placed on prisoner abuse. The crisis would eventually lead both sides to shackle a number of each other’s prisoners in contravention of the laws of war. Fearing retaliation, both sides avoided more violent forms of abuse during this episode and eventually reached a settlement (MacKenzie 1995). This restraint contrasted with Germany’s brutal treatment of prisoners during its invasion of Poland where the decidedly one-sided nature of the fighting meant Germany had little fear of reprisals (Rossino 2003, 179). Even though it faced logistical difficulties, the decline in Allied treatment of German POWs at war’s end similarly illustrates captors’ willingness to care for enemy prisoners often erodes once the threat of retaliation disappears (MacKenzie 1994, 502-3). Generalizing across these cases suggests the following hypothesis:

Hypothesis 1: States are likely to commit lower levels of prisoner abuse when each holds prisoners from the other side.

International Law and Norms

Beginning concertedly with the 1899 Hague Conventions, states negotiated several treaties laying out obligations toward prisoner treatment. Despite legal protections, realist scholars argue the lack of enforcement mechanisms means international law exerts few constraints on states (Mearsheimer 1994–95). Others remark that states tend to commit to agreements reflecting preexisting interests, meaning their behavior would differ little in the absence of the agreement (Downs, Rocke, and Barsoom 1996). A large literature has developed countering that international law plays a significant role in shaping the preferences and conduct of states even in times of war. States possess a variety of noninstrumental motives for following international law as they gradually internalize an agreement’s principles as part of their view of appropriate conduct (Reus-Smit 2004, 21-24). Others point instead to fears of reputational costs for reneging on agreements and being perceived as untrustworthy (Guzman 2008, 34-35). Taken together, this suggests differences in the behavior of states based on whether or not they have committed to relevant treaties.⁴

Hypothesis 2: States that have ratified the prevailing laws of war are likely to commit lower levels of prisoner abuse than those states that have not ratified.

Unilateral ratification of the laws of war might be insufficient to ensure protections for prisoners. Joint ratification increases the likelihood both sides understand what types of actions count as noncompliance and makes it more likely they will coordinate their actions and refrain from prisoner abuse (Morrow 2007, 561).⁵

Hypothesis 3: States are likely to commit lower levels of prisoner abuse when both sides have ratified the prevailing laws of war.

Principles embodied in international law may also have influence beyond the formal parties to the treaty. Higher levels of support for norms contained in existing international treaties may shape the preferences and behavior of all states, irrespective of whether they ratified the original agreement (Finnemore and Sikkink 1998, 900-901). Principles enshrined in the 1949 Geneva Conventions are frequently viewed as a watershed in protections for war victims, particularly POWs (Best 1994, 80). As new normative orders take hold, states may come to see prisoner abuse as an illegitimate strategy.

Hypothesis 4: States are likely to commit lower levels of prisoner abuse in wars taking place after 1949.

Perceptions of the Enemy’s Identity

One of the most common explanations in the existing literature claims that prisoner abuse is motivated by barbaric images of the adversary based on racial, religious, or other cultural differences (Beaumont 1996, 280). Contempt for opposing cultures is evident in an Italian officer’s remarks during the 1895–96 war against Ethiopia, “A bit of Pizarro [referring to the Spanish conquistador] does not hurt; with some people terror works better than kindness” (Baudendistel 2006, 221). Similarly, Dower (1986, 11) argues that racial hatreds led both the United States and Japan to commit atrocities against each other’s prisoners on the Pacific Front during the Second World War. In contrast, states from similar cultural backgrounds, such as Bolivia and Paraguay during the 1932–35 Chaco War, did not view each other as barbaric and correspondingly treated prisoners decently (Zook 1960, 101).

The reasoning underlying these arguments shares similarities to Huntington’s clash of civilizations thesis that wars are more frequent and brutal between groups from different civilizational origins (1993, 25). Social identity theorists propose a broader process of in-grouping and out-grouping, whereby groups are prone to treat those from outside groups in a more hostile manner (Rabbie 1989, 141). This leads to the following hypothesis linking cultural differences and resulting perceptions of barbarism to prisoner abuse.

Hypothesis 5: States are likely to commit higher levels of prisoner abuse when their adversary is viewed as barbaric and from a different culture.

The Nature of the Fighting and Wars of Attrition

In contrast to these common explanations for prisoner treatment, recent research on civilian victimization highlights several other factors that shape incentives for abusing prisoners. Belligerents facing drawn-out or severe fighting that devolves into a war of attrition may be especially attracted to potential benefits arising from prisoner abuse. Following a logic of coercion, prisoner abuse provides a means to test the enemy’s capability and resolve to continue fighting by increasing the costs of war (Pape 1996, 12-13). Of course, abuse poses the risk of becoming counterproductive by simply increasing the determination of those troops still on the battlefield to continue fighting rather than surrender (Rees 1999, 67). Nevertheless, abuse against defenseless groups has proven effective in certain circumstances, and in desperate situations, states may turn to any strategy increasing their chances for victory, however remote (Downes 2008, 38-39).

An extractive logic provides further motives to abuse prisoners during wars of attrition where resources are often stretched thin. Properly caring for prisoners can be expensive in terms of finances and personnel. During portions of the First World War, Austria-Hungary spent more taking care of its 1.8 million prisoners than it did on explosives (Davis 1977, 629). Reallocating funds away from maintaining prisoners frees up resources to serve more pressing military and civilian needs. Furthermore, the incentives to ruthlessly exploit captives as a slave labor force are also higher, as was evident in the atrocious conditions suffered by prisoners of Russia in building the strategically important Murman railway during the same war (Rachamimov 2002, 112). The coercive and extractive motives both suggest that prisoner abuse will be an appealing strategy in particularly hard-fought conflicts.

Hypothesis 6: States engaged in a costly war of attrition are likely to commit higher levels of prisoner abuse.

War Aims and Territorial Annexation

States do not enter into war without seeking to achieve specific goals (Holsti 1991, 12-20). Some aims create greater incentives for prisoner abuse than for others, in particular, the desire to annex enemy territory. Civilians are frequently targeted when territory is at stake because the populace represents a potential “fifth column” that could foment rebellion (Downes 2008, 35-36; Valentino, Huth, and Croco 2006, 356). If conquerors perceive civilians as threats, then enemy prisoners present a particularly menacing danger. Soldiers are usually of prime military age and have received at least a minimum level of training, making them a serious hazard for occupying powers (Stachura 2004, 132-33). One of the primary motives behind the 1940 Katyn massacre, when the Soviet Union executed over 20,000 Polish officers, was to exterminate any future resistance to Soviet control (Cienciala et al. 2007, 141-42).

The defining trait of wars of annexation is the desire of prospective conquerors to achieve permanent rule over the territory in question. Indefinitely holding enemy combatants in captivity remains an option but would become extremely costly and divert resources from solidifying control over the civilian populace and exploiting newly conquered lands. Furthermore, the danger of prisoners escaping and taking up arms would always exist. Rather than taking such chances, states seeking to conquer enemy territory possess strong motives to abuse prisoners.

Hypothesis 7: States seeking to annex territory from an adversary are likely to commit higher levels of prisoner abuse.

Other expansive war aims, such as unconditional surrender or regime change, might not provide as clear incentives. The coercive potential of prisoner abuse provides one route toward forcing an adversary to surrender without preconditions. As noted earlier, however, extreme levels of abuse can actually prove detrimental for achieving unconditional surrender because enemy soldiers may prefer to continue fighting rather than lay down their arms. The aims of the United States to achieve unconditional surrender on both the European and the Pacific Fronts during the Second World War were followed by concerted attempts later in the war to employ good prisoner treatment to induce surrenders (Ferguson 2004, 188-91).

A similar ambivalence in the resort to prisoner abuse is evident in wars involving regime change, since military elites have played stabilizing roles in many postconflict situations (O’Donnell and Schmitter 1986, 39–40). Committing high levels of prisoner abuse risks undermining both the capability and the willingness of the adversary’s military to play a constructive role in any future regime. Incentives for prisoner abuse are not necessarily absent in other cases of ambitious war aims, but the prospects for prisoner abuse are less clear compared to territorial motives.

Democracy and Prisoner Treatment

The general question of whether or not democracies act differently in foreign affairs has numerous implications for wartime conduct. Many of these arguments owe their inspiration to the democratic peace thesis that democracies do not fight one another (Russett 1993, 30-42). Translating insights from the democratic peace requires moving beyond the focus on dyadic interactions between states sharing similar political cultures or institutions. Given that democracies do not fight against each other, theories relating democracy to wartime conduct focus instead on a monadic relationship, whereby the characteristics of democracies lead them to fight in distinctive ways. While many earlier studies found little support for monadic effects, recent work suggests that the overall distinctiveness of democracies rests on firmer theoretical and empirical grounds (Lipson 2003, 21).

Arguments linking democracy to the humane treatment of civilians provides several parallels to caring for POWs by emphasizing the role of liberal and democratic norms inculcated in the political culture of democratic regimes (Rummel 1995, 4). Cultural restraints against violence suggest that democracies are unwilling to “play rough” on or off the battlefield (Engelhardt 1992, 53). The exact source of domestic norms governing wartime conduct continues to be debated. Doyle argues that the tendency for democracies to externalize liberal norms of respect for individual rights necessitates they comply with all aspects of the laws of war (1983, 344). Locke follows a similar logic in asserting that liberal principles mandate democracies treat enemy prisoners in a humane manner ([1690]/1980, 93).

Other scholars emphasize the role of democratic norms, such as equality and tolerance, for explaining the conduct of democracies (Weart 1998, 59-60). Valentino, Huth, and Croco propose that “If democratic values promote tolerance, nonviolence, and respect for legal constraints, then democracies should wage their wars more humanely than other forms of government” (2004, 382). In contrast, autocracies tend to treaty their own citizens in a harsher manner, and should be more likely to resort to similar brutalities during war. Irrespective of its liberal or democratic origins, the political culture of democracies leads to similar expectations regarding prisoner treatment.

Hypothesis 8: Democracies are likely to commit lower levels of prisoner abuse than nondemocracies.

Normative principles may be reinforced by domestic institutional constraints when democracies decide how to treat enemy prisoners. Democratic institutions make leaders more accountable to their publics compared to autocracies. Past studies find democracies are highly sensitive to casualties and resulting declines in public support for war (Mueller 1973, 62). If democracies are more casualty averse, then democratic leaders should be especially wary of abusing prisoners in light of prospects for retaliation. Although not immune from the popular will, autocratic leaders are less vulnerable to any resulting costs of retaliation. Even in cases where the adversary begins abusing prisoners, democracies may shrink away from responding in kind out of fear retaliating would only lead to further escalation in violence. During the Second World War, Churchill justified proper care for Axis soldiers fearing a “general countermassacre of prisoners” could be sparked by British abuses (Moore 2000, 90). The greater sensitivity of democracies to concerns over retaliation leads to the following two hypotheses:

Hypothesis 9: Democracies are likely to commit lower levels of prisoner abuse than nondemocracies when both sides hold prisoners.

Hypothesis 10: Democracies are likely to commit lower levels of prisoner abuse than nondemocracies even when the adversary resorts to prisoner abuse.

Institutional constraints are far from limitless. Under certain conditions, public accountability may even increase the willingness of democracies to employ brutal methods. While democracies may be more casualty averse, research also shows democratic leaders are more likely to be removed from office if their side loses a war (Bueno de Mesquita and Siverson 1995). Democracies thus exhibit greater caution when deciding to go war and tend to select wars they are likely to win quickly and cheaply (Reiter and Stam 2002, 10-11). When encountering a costly conflict, however, democratic leaders should be more likely to fight hard and do whatever it takes to win and remain in power (Bueno de Mesquita et al. 1999). Given the costly nature of attritional warfare, Downes (2007, 877) argues that the greater domestic demands placed upon democracies make them more willing to fight in nastier ways to prevail, such as victimizing enemy civilians. When facing threats to national security, democracies have also proved enthusiastic in employing harsh methods against prisoners (Rejali 2007, 46-47). This leads to an interactive relationship between regime type and the severity of the conflict.

Hypothesis 11: Democracies are likely to commit higher levels of prisoner abuse than nondemocracies in costly wars of attrition.

The institutional argument that democratic leaders tend to select wars they are favored to win quickly and decisively also suggests that democracies may be less likely to need to employ prisoner abuse in wars they initiate. This implies that democracies may act differently in wars of their own choosing than when they are targeted.

Hypothesis 12: Democracies are likely to commit lower levels of prisoner abuse than nondemocracies in wars they initiate.

Alongside the wartime context, two other arguments point to further conditional relationships between democracy and prisoner abuse. The centrality of the rule of law domestically means democracies should also be more likely to be rule-bound in their foreign relations (Slaughter 1995, 532-33). Democratic leaders may also be more vulnerable to audience costs from domestic groups should they fail to uphold international commitments (Lipson 2003, 81). Taken together, democracies should be more likely to abide by international rules if they have ratified the relevant international treaty (Morrow 2007, 561).

Hypothesis 13: Democracies are likely to commit lower levels of prisoner abuse than nondemocracies when they have ratified the prevailing laws of war.

Even if they have not committed to a formal international agreement, democracies may also be increasingly sensitive to international norms. Small but important segments of the public espousing humanitarian ideals have gained a growing influence within democracies, shifting attention toward the prevention of abuses (Merom 2003, 77-78). In the midst of normative changes both internationally and domestically, the 1949 Geneva Conventions served as a crucial focal point for the conduct of democracies during war. This leads to the final hypothesis:

Hypothesis 14: Democracies are likely to commit lower levels of prisoner abuse than nondemocracies in wars taking place after 1949.

Data and Measurement

To test these hypotheses, I constructed a data set examining the treatment of prisoners during all interstate wars from 1898 to 2003. I do not include wars before this period because of the limited availability and reliability of information on prisoner treatment for earlier years. An additional advantage of this period is it matches closely with the modern codification of the laws of war. Although many states may not have ratified the relevant treaties, there was at least a common baseline understanding of prevailing international obligations concerning prisoner treatment.

The list of wars is based on the Correlates of War (COW) Interstate War data set version 4.0 (Sarkees and Wayman 2010). Following other scholars, I divide several long multiactor wars into separate military confrontations, which more accurately reflects actual fighting on the ground (Downes 2008, 43; Reiter and Stam 2002, 39).⁶ The final data set contains 79 wars in total.

Dependent Variable: Prisoner Abuse

I define prisoner abuse as a military strategy enacted by political and military authorities that involves the intentional killing or harming, either directly or indirectly, of enemy combatants who have laid down their arms and surrendered. While the motives of individual soldiers in committing violations are certainly worthy of further study, the most widespread cases of prisoner abuse, such as the treatment of Soviet prisoners by Nazi Germany, are almost always directed by higher authorities or require their tacit approval. This study thus focuses on government policy and practices rather than isolated acts by individual soldiers or small troop units.

The outcome of interest concerns the overall level of abuse inflicted by captor states against prisoners. The unit of analysis is the warring–directed dyad, meaning there are two observations for every pair of belligerents, where each state is considered a potential violator or victim, respectively. For instance, in the 1982 Falklands War, one observation examines British treatment of Argentinean prisoners, and the other looks at Argentinean treatment of British prisoners. In wars involving more than two parties, there are two corresponding observations for each pair of opposing states.

The analysis is then limited to states deemed “capable captors.”⁷ A state is considered a capable captor if it meets both of the following criteria. First, a state must have the capacity for an independent prisoner policy. This is not much of a concern in wars limited to two states, but in conflicts involving large coalitions, subordinate partners often have little leeway in adopting separate military practices from more powerful patrons. During the Korean War, allied forces nominally operated under a unified United Nations command, but the United States dominated all aspects of the planning and prosecution of the war, including prisoner policies (Springer 2010, 167-68). I do not treat states in such subordinate positions as separate observations, since their behavior is not independent but rather largely determined by their larger ally. Only when states have separate operational control over prisoner treatment are they considered to have an independent prisoner policy, as in the case of US and British forces during the Second World War.

Second, states must also have a realistic opportunity to mistreat prisoners by capturing enemy combatants during hostilities.⁸ Most wars involve both sides capturing substantial, though not necessarily equal, numbers of prisoners. In cases of particularly one-sided fighting, however, it is not uncommon for the losing side to capture few if any prisoners. For instance, during the Second World War, Belgium was quickly overrun by German forces and records indicate no Wehrmacht soldiers were captured (Warmbrunn 1993, 45-48).

Coding such cases as absent of prisoner abuse would equate states that never had a meaningful chance to harm enemy prisoners with those that actually had an opportunity but chose to treat their captives decently. The latter group makes up legitimate examples of good prisoner treatment because the state committed few violations. In contrast, the former group is better seen as irrelevant; prisoner abuse is impossible because no prisoners were captured in the first place (Mahoney and Goertz 2004, 654). I thus limit the analysis to states that had both the capacity and the opportunity to harm prisoners during war.

Developing a reliable measure for prisoner abuse presents several challenges. Belligerents have incentives to conceal their own crimes, while exaggerating those of their adversary to increase domestic or third-party support. Beyond intentional deception, the sheer mass of prisoners, often numbering in the thousands or sometimes millions, makes accurate data collection even more difficult. Despite these limitations, substantial variation is evident in the treatment of prisoners. Even if it were possible to amass detailed information on the number of prisoner deaths, this would only reveal one aspect of the broader phenomenon of prisoner abuse. Prisoners do not necessarily need to be killed in order to suffer greatly. Torture or lifelong ailments caused by poor nutrition or hard labor conditions also count as painful forms of prisoner abuse according to my definition.

The dependent variable, Prisoner Abuse, is a categorical measure where each observation exhibits either high, medium, or low levels of prisoner abuse. The strategy of relying on a categorical variable is consistent with many other studies of wartime conduct confronting similar data challenges.⁹ High-level cases involve state-sanctioned, widespread, and systematic abuse of prisoners. Although differences in the extent of violence may exist, cases included in this category generally involve captors showing little or no regard for the rights of prisoners. Examples include Japanese treatment of Allied prisoners during the Second World War, or French treatment of Chinese prisoners during the Boxer Rebellion.

At the other extreme are low-level cases where prisoners are generally treated well and abuse is rare. This does not necessarily mean violations never occur during the course of conflict. Warfare unfortunately, but almost inevitably, entails a harsh reality for most participants and the lot of prisoners is far from ideal even in the best circumstances. What distinguishes low-level cases is abuse remains the exception rather than the rule. Examples include US treatment of Iraqi prisoners during the 1991 Gulf War and British treatment of Argentinean prisoners in the 1982 Falklands War.

Between the two points lie medium-level cases, where abuses certainly occur to a much greater extent than in low-level instances, but where violations are overall less common or extensive compared to high-level abuse. This category may also include more frequent individual acts of abuse which are not explicitly supported by the state, but nor are firm steps taken by authorities to prevent further violations from taking place. Examples include US treatment of North Korean and Chinese prisoners during the Korean War.

In order to construct the overall high/medium/low indicator of prisoner abuse, I first coded prisoner treatment across several distinct types of violations.¹⁰ A wide range of acts technically contravene the laws of war. Under the Third Geneva Convention, denying prisoners access to scientific equipment, musical instruments, and sports outfits count as violations.¹¹ Although such violations certainly cause some hardship, I focus on particularly heinous acts deemed to be “grave breaches” of the Convention.¹² Following the grave breaches logic, I examine six separate violations of prisoner abuse: execution, torture, denial of legal rights, compulsory military conscription, hard labor, and poor housing and nutrition.

I code each component violation according to a similar categorical scheme.¹³ I then use these individual components to construct a measure for the overall level of violations based on the following decision rule—the summary indicator equals the highest level of abuse across the six component abuses. For instance, if a state engaged in high levels of abuse for execution, but low levels of abuse for labor conditions, it is still considered to have engaged in high levels of abuse overall. I chose this rule to hold states to the prevailing stringent legal standards, since any of these violations on their own is considered a war crime. However, the results do not change substantially when using alternative procedures for constructing the prisoner abuse summary indicator.¹⁴

I rely mainly on secondary resources supplemented by primary documents for coding each dimension of prisoner abuse. In order to minimize biases posed by relying on historical materials, I employ multiple sources for each war to the greatest extent possible, as well as rely on studies written well after the war ended, or by individuals not party to the conflict. I was unable to code 5 percent of the cases, but the current data set remains the most complete collection of information on prisoner treatment to date. Cases were also independently coded by research assistants using the same set of sources; intercoder reliability showed approximately a 70 percent agreement in coding decisions.

Independent Variables

The independent variables are drawn from a variety of sources. Democracy is a dichotomous variable based on the Polity IV index (Marshall and Jaggers 2007). A country is coded 1 as a democracy if it has a Polity score of 7 or higher before the war started, and 0 otherwise. In order to ensure results are not driven by the particular democracy measure, I also considered different cutoffs and alternative indicators; the results do not change substantially.

For hypotheses relating to the nature of the conflict, the severity of the fighting was measured using War of Attrition, which is a dichotomous variable based on work from Downes (2008, pp. 59-60) on civilian victimization.¹⁵ The variable equals 1 for conflicts that become bogged down in static or protracted fighting and 0 otherwise. The trench warfare typifying much of the Western Front during World War I is one of the most prominent examples of attritional warfare.¹⁶

War aims are represented using two different variables. Territorial Annexation is a dichotomous variable that equals 1 for observations where the violating state aims to conquer territory from the adversary and 0 otherwise (Downes 2008, 61). Examples include German ambitions in acquiring Soviet territory during World War II, and Italy’s conquest of Ethiopia during their war from 1935 to 1936. Expansive War Aims is a similar variable that takes a value of 1 if the state in question either sought regime change or the enemy’s unconditional surrender and 0 otherwise (Downes 2008, 60-61). Instances include the US aim to topple Saddam Hussein during the 2003 Iraq War, or US demands for the unconditional surrender of Japan during World War II.

Turning to testing some of the conventional explanations for prisoner treatment in the existing literature, the constraining effect of retaliatory expectations requires examining the captor status of each state in a directed dyad. If both states hold prisoners, then each may be deterred from committing abuse out of fear their opponent will respond in kind. Deterrence takes on a value of 1 if both sides are considered capable captors and 0 otherwise.

Measures for identity and images of the adversary have proven problematic in past research (Abdelal et al. 2006). Following other quantitative studies of armed conflict, I use Cultural Differences as a proxy, which measures the ex ante civilizational differences between warring parties (Henderson and Tucker 2001). While somewhat crude, this variable captures key cleavages between different societies: Western, Latin American, Hindu, Orthodox, Islamic, African, Sinic, Buddhist, Japanese, and a residual other category. Cultural Differences equals 1 if belligerents are from different civilizations, or both are coded as other, and 0 if they are from the same civilization. Although far from ideal, this remains the best available option for incorporating identity differences into large-N studies.

For international law, Treaty Ratification is a dichotomous variable measuring whether the state ratified the prevailing treaty respecting prisoner rights by the start of the war.¹⁷ Joint Treaty Ratification is a corresponding variable reflecting whether both states in each pair of directed dyads ratified the relevant treaty. Because Joint Treaty Ratification is essentially an interaction between the unilateral ratification decisions of both states, I add Treaty Ratification (Opponent) so that all constitutive terms are included in the model (Brambor, Clark, and Golder 2006, 66). In order to assess wider normative effects of the Geneva Conventions, Post-1949 Norms is a dichotomous variable that equals 1 for wars starting after 1949, and 0 for earlier wars.

I also include a number of control variables commonly thought to influence wartime conduct. As numerous historical episodes have demonstrated, states are often likely to retaliate against abuses committed by the other side (MacKenzie 1994, 491). Prisoner Abuse by Opponent is a corresponding categorical variable measuring the level of prisoner abuse committed by the adversary.¹⁸

States with larger material capabilities may have a greater ability to abuse enemy combatants. On the other hand, more powerful states may have less need to mistreat prisoners, given their military superiority. More pragmatically, stronger states may simply have greater resources at their disposal to adequately care for prisoners falling into their hands. By comparison, weaker states may be unable to properly care for prisoners irrespective of the nature of their intentions. Relative Capabilities is defined as the percentage of total capabilities of all belligerents in the war controlled by each state, which is measured using the COW composite index of national capabilities (Singer, Bremer, and Stuckey 1972).¹⁹

Results

Table 1 shows the results from an ordered logit analysis of the determinants of prisoner abuse. A positive coefficient indicates an increase in the variable is associated with a rise in a state’s level of prisoner abuse. Since prisoner abuse may not be independent within a given conflict, robust standard errors clustered by war are reported. Model 1 provides the main model estimating the causes of prisoner abuse and includes the primary explanatory and control variables.

Table 1.

Ordered Logit Analysis of the Determinants of Prisoner Abuse in Interstate Wars, 1898–2003

	Main model
	Main model	Democracy × Deterrence	Democracy × Reciprocity	Democracy × Attrition	Democracy × Initiator	Democracy × Treaty	Democracy × Norms
	(1)	(2)	(3)	(4)	(5)	(6)	(7)
Democracy	−1.09* (0.47)	−0.60 (1.53)	−0.71 (0.70)	−0.67 (0.53)	−1.36* (0.54)	−0.47 (0.72)	−1.03 (0.77)
War of attrition	1.31** (0.34)	1.30** (0.34)	1.30** (0.34)	1.53** (0.50)	1.30** (0.34)	1.35** (0.35)	1.31** (0.34)
Expansive war aims	0.16 (0.34)	0.18 (0.34)	0.17 (0.35)	0.30 (0.34)	0.24 (0.33)	0.13 (0.34)	0.15 (0.34)
Territorial annexation	2.04** (0.49)	2.05** (0.50)	2.04** (0.50)	2.09** (0.53)	2.08** (0.52)	1.98** (0.47)	2.04** (0.49)
Cultural differences	0.36 (0.23)	0.36 (0.23)	0.36 (0.23)	0.37 (0.22)	0.34 (0.23)	0.36 (0.23)	0.38 (0.25)
Prisoner abuse by opponent	1.41** (0.45)	1.41** (0.45)	1.50** (0.48)	1.41** (0.45)	1.44** (0.48)	1.40** (0.44)	1.41** (0.45)
Deterrence	−1.03 (0.63)	−0.91 (0.66)	−1.03 (0.63)	−0.97 (0.63)	−1.08^† (0.63)	−1.02 (0.63)	−1.04^† (0.62)
Relative capabilities	0.23 (0.52)	0.20 (0.50)	0.19 (0.51)	0.25 (0.54)	0.04 (0.56)	0.25 (0.51)	0.25 (0.51)
Treaty ratification	0.57 (0.69)	0.56 (0.69)	0.59 (0.69)	0.52 (0.72)	0.63 (0.66)	0.77 (0.70)	0.57 (0.71)
Treaty ratification (opponent)	0.73 (0.58)	0.70 (0.58)	0.73 (0.58)	0.59 (0.60)	0.76 (0.53)	0.60 (0.58)	0.71 (0.62)
Joint treaty ratification	−1.34^† (0.74)	−1.29^† (0.74)	−1.37^† (0.76)	−1.18 (0.73)	−1.42* (0.72)	−1.20 (0.73)	−1.31 (0.80)
Post-1949 norms	0.54* (0.26)	0.52* (0.26)	0.51^† (0.27)	0.50^† (0.27)	0.51^† (0.27)	0.56* (0.26)	0.57 (0.40)
Democracy × Deterrence		−0.54 (1.57)
Democracy × Prisoner abuse by opponent			−0.33 (0.43)
Democracy × Attrition				−1.01 (1.10)
Initiator					0.10 (0.49)
Democracy × Initiator					0.84 (0.74)
Democracy × Treaty						−0.87 (0.95)
Democracy × Norms							−0.14 (1.07)
First cutpoint (τ₁)	0.18 (0.61)	0.28 (0.64)	0.22 (0.61)	0.32 (0.65)	0.15 (0.61)	0.32 (0.62)	0.19 (0.62)
Second cutpoint (τ₂)	2.69** (0.64)	2.79** (0.66)	2.74** (0.64)	2.84** (0.70)	2.67** (0.66)	2.85** (0.64)	2.70** (0.64)
Observations	272	272	272	272	272	272	272
Log likelihood	−215.6	−215.4	−215.2	−214.3	−214.2	−214.7	−215.6
χ²	88.16	92.22	96.57	87.58	88.17	89.44	91.95
Pseudo R ²	.270	.270	.271	.274	.274	.272	.270

Note. Robust standard errors in parentheses (clustered by war).

** p < 0.01. * p < 0.05. ^† p < 0.1 (two-tailed test).

The findings from model 1 offer some modest support for several of the more common arguments for prisoner abuse. Deterrence is in the expected negative direction, meaning that belligerents who are each capable captors are more likely to exert restraint when deciding whether to abuse prisoners, though the coefficient narrowly misses being statistically significant at the 10 percent level. Nevertheless, the results suggest troops from states that fail to take captives of their own are in greater danger of prisoner abuse, as in the case of Kuwait in the opening phase of the Persian Gulf War at the hands of Iraq.

Cultural Differences is positive and in the expected direction, though it also fails to be statistically significant. This suggests barbaric images of the enemy may play some role in prisoner abuse, though the effect appears far from robust. It is indicative that conflicts between belligerents sharing a similar cultural background have sometimes proven to be extremely harsh both on and off the battlefield, as was evident on the Western Front during the First World War.²⁰

Despite being championed by many activists, international law and humanitarian norms appear to provide few protections for prisoners. The coefficient for Joint Treaty Ratification is negative and statistical significant at the 10 percent level. This suggests that international law may have a constraining effect but only when both sides have committed to international law. On the other hand, the weakly positive results for the lower-order ratification terms suggest that states feel less bound by legal constraints if their adversary failed to commit as well.

Post-1949 Norms is significant but in the opposite direction than expected, with wars taking place after 1949 associated with greater prisoner abuse. This does not necessarily imply growing support for humanitarian norms has directly led to worsening levels of prisoner abuse, since the variable may be capturing other developments also varying over time. Nevertheless, the result represents a sobering trend suggesting the conduct of war has not become more humane in recent years. International norms might be better viewed as responding to, rather than shaping, wartime conduct.

Turning to the other main hypotheses, attributes of the nature of the conflict appear to play a much more prominent role on the level of prisoner abuse. War of Attrition is positive and statistically significant, which supports the view that more severe wars are more likely to lead to higher levels of abuse. The static and drawn-out fighting characterizing the 1998–2000 war between Ethiopia and Eritrea led to attempts by both sides to punish their adversary through harming captured troops (Gilkes and Plaut 1999, 52). Similarly, during both World Wars, states often exploited prisoners ruthlessly to shore up their productive capacities for the war effort (Davis 1977).

Looking at the particular aims sought through war, Territorial Annexation is also positive and significant, indicating that countries seeking to conquer enemy territory are especially likely to abuse prisoners. The brutal treatment of prisoners by both Armenia and Azerbaijan in their war over the contested Nagorno–Karabakh region during the early 1990s is thus far from an exception. As expected, however, not all war aims are created equal; the coefficient for Expansive War Aims is slightly positive but fails to achieve statistically significance.²¹

Model 1 also provides support for the restraining role of regime type. Democracy is negative and statistically significant, meaning democracies are less likely to resort to higher levels of prisoner abuse compared to nondemocratic belligerents. While exceptions certainly exist, such as the conduct of the Western democratic powers during the Boxer Rebellion, on the whole, democracies appear to treat their captives in a more humane manner. The democracy finding is particularly intriguing, given recent work on the treatment of civilians has found regime type either has no effect, or that democracies are actually more likely to commit violations. Although by no means absolute, democratic norms and institutions appear to provide meaningful constraints against abusing prisoners.²²

The results are mixed for the remaining control variables. Material capabilities seem to have little bearing on prisoner abuse; the coefficient is slightly positive in most models but far from achieving statistically significance. Prisoner abuse thus appears to be neither a tool of the powerful nor a weapon of the weak to level the playing field. By contrast, reciprocity does seem to figure prominently in the decision making of belligerents. When their own soldiers are abused by the adversary, states are more likely to act in a similar manner.²³

The subsequent models in Table 1 delve more deeply into the impact of regime type by examining possible interactions between democracy and several other explanatory variables. These conditional hypotheses also provide a way to assess the relative role of normative and institutional mechanisms on the conduct of democracies. It should be noted, however, that specification tests reveal none of the conditional models significantly improve the goodness of fit over the baseline model 1.²⁴

Nevertheless, the remaining models provide some indications of the role of regime type across different contexts. Interpreting interactive hypotheses is problematic because neither the constitutive nor the interaction terms represent unconditional marginal effects. Since the relationship between regime type and the conditioning variable depends on value of both the lower order and the interaction terms, calculating marginal effects using substantively meaningful values is normally recommended (Brambor, Clark, and Golder 2006, 73-74). This is relatively straightforward here, given the relevant explanatory variables are all dichotomous. The interaction effects are evaluated using the change in the predicted probability of committing high levels of abuse, which is reported in parentheses below where relevant.

Looking at model 2, democracies do appear more concerned with risks of retaliation due to public pressures and institutional incentives. Democratic belligerents are less likely (−0.19) to commit high-level abuse when both sides hold prisoners, and the effect is statistically significant. The lower-order coefficient for democracy remains negative but is no longer significant, suggesting normative constraints exert less force in situations where democracies have fewer concerns over retaliation. However, even in cases where the adversary has committed prisoner abuse as in model 3, democracies are less likely to resort to abuse in kind (−0.18), perhaps partly out of fear of further escalation by their opponent.

There is less support, however, for other institutional elements concerning democratic distinctiveness in prisoner treatment. Despite purportedly greater incentives to win quickly, in model 4 democracies are actually less likely to commit prisoner abuse in wars of attrition (−0.34), though the effect is not statistically significant. The negative effects is nonetheless interesting in light of other findings that democracies are especially prone to targeting civilians in wars of attrition (Downes 2006, 176). This suggests that the incentives and wartime behavior of democracies may differ in important ways across different issue areas. Democracies also do not appear better at picking easier fights that should result in less need for prisoner abuse. The results of model 5 show democratic initiators are less likely to commit abuse (−0.11), though the effect is far from significant. If anything, the results indicate that democracies who are targeted are significantly less likely to resort to prisoner abuse (−0.23).

Turning to the role of international law and norms, in model 6, democratic ratifiers of the prevailing laws of war are significantly less likely to abuse prisoners (−0.24), which is consistent with other research on regime type and compliance with international law. This finding, however, could be consistent with either normative mechanisms emphasizing domestic cultural values for the rule of law, or institutional mechanisms focusing on the greater pressures to uphold commitments due to domestic audiences (Morrow 2007, 561). Model 7 further suggests that normative factors may have greater pull for certain regimes in recent years. Democracies are less likely to commit prisoner abuse in wars after the advent of the 1949 Geneva Conventions (−0.23) though the effect fails to be statistically significant. In light of the positive coefficient for the post-1949 period in the baseline model 1, this tentatively suggests greater divergence between democracies and autocracies in prisoner treatment over time. Taken together, the findings from the additional models indicate several conditional effects for regime type and offer some support for both institutional and normative factors in shaping democratic conduct. Nonetheless, the overall results suggest the baseline model 1, which focuses on the direct effect of democracy, provides a useful model for understanding overall patterns in prisoner abuse.

Based on model 1, I assess the substantive significance of the results by estimating the predicted probability a state will commit high levels of prisoner abuse, while changing one independent variable at a time. Table 2 reports the predicted probability of high-level prisoner abuse at both the lower and the upper values for each variable, as well as the absolute and percentage change. The results show that several variables have a profound impact on resorting to prisoner abuse. Democracies are more than 50 percent less likely to abuse prisoners at the highest levels compared to their autocratic counterparts. In contrast, wars devolving into attritional fighting, or involving territorial annexation, are, respectively, two and three times more likely to result in high-level abuse. States whose prisoners are targeted with high levels of abuse by their opponent are similarly more than twice as likely to inflict equivalent levels of violence. Deterrence has a comparable moderating impact to regime type, though fails to be statistically significant. Wars taking place after 1949 are 44 percent more likely to result in high-level abuse, which questions any inherent humanizing trends in recent warfare. The remaining variables from the analysis have varying effects on the predicted probability of high-level prisoner abuse but have much wider confidence intervals that include zero, which suggests little apparent substantive effects.

Table 2.

Substantive Effects of Independent Variables on the Probability of High Levels of Prisoner Abuse

	Initial probability	Probability after change in variable	Absolute change	Percent change
Democracy	.37	.17	−.20 (−.34, −.03)	−54
War of attrition	.20	.49	.28 (.15, .43)	138
Expansive war aims	.30	.33	.03 (−.11, .18)	10
Territorial annexation	.24	.70	.46 (.25, .63)	193
Cultural differences	.26	.33	.07 (−.02, .17)	28
Prisoner abuse by opponent	.29	.62	.33 (.13, .53)	113
Deterrence	.53	.29	−.24 (−.51, .04)	−46
Relative capabilities	.29	.33	.04 (−.13, .21)	13
Treaty ratification	.23	.34	.11 (−.18, .33)	47
Treaty ratification (opponent)	.21	.35	.14 (−.09, .31)	64
Joint treaty ratification	.52	.22	−.30 (−.57, .04)	−57
Post-1949 norms	.26	.38	.12 (.01, .22)	44

Note. 95% confidence intervals for absolute change are reported in parentheses. Baseline probability of high level of prisoner abuse equals 0.31 where all variables are held at their means. For the predicted probabilities reported above, all independent variables are held constant at their means except for the variable of interest. Dichotomous variables are changed from 0 to 1, continuous variables from their 10th to 90th percentile values, and Prisoner abuse by opponent from medium to high values. All estimates were calculated using Clarify.

Although not reported here, I also conducted a number of robustness checks to ensure that the findings are not driven by country-, regime-, or war-specific effects. The results for the main variables of the analysis do not change substantially when incorporating any of these additional concerns.²⁵ Given the analysis focused on states deemed to be capable captors, I also reran the analysis incorporating noncapable states, but the results again remain largely the same.²⁶ Taken together, the combined set of tests, while far from exhaustive, suggests the findings regarding regime type, the nature of the war aims, and the severity of the fighting are key factors for understanding the determinants of prisoner abuse.

Conclusion

War often involves captors mercilessly victimizing enemy soldiers unlucky enough to fall into their grasp, while in other instances prisoners are humanely treated. I sought to provide an explanation for these differences by examining a wide range of factors thought to influence the conduct of states during war. Democracies were on average less likely to commit abuse against prisoners than autocratic belligerents. The role of democracy was also conditional in certain respects, especially in situations of deterrence, treaty ratification, and during the post-1949 period. In contrast, certain attributes of the conflict itself, especially attritional warfare and the desire for territorial conquest, provide strong incentives for states to commit violence against prisoners. Reciprocity also plays an important role in decision making, either in the form of retaliation against violations committed by the adversary or more, modestly by deterring abuses through expectations of retribution.

Other factors commonly believed to affect state conduct, however, received less support. Despite often deep-seated animosities, belligerents from different cultures were not especially likely to abuse their prisoners, though this factor deserves further scrutiny. Normative and legal appeals appear to have gained little traction despite laudable efforts; commitment to the laws of war, or the spread of humanitarian norms, offered few constraints against abuse except in the case where both sides had ratified the prevailing laws of war. Material capabilities also provided little leverage for understanding prisoner abuse, since both weak and strong powers equally prey upon prisoners.

Several findings stand out in relation to the larger literature on wartime conduct. The enabling effects of attrition and territorial annexation track closely with research on civilian targeting. The democracy finding, on the other hand, is particularly surprising in light of recent studies on civilian targeting, which find regime type does not matter or that democracies are more likely to commit violations (Downes 2007; Valentino, Huth, and Croco 2006).

The results present a puzzle for why regime type appears to have varying effects on the conduct of states across separate wartime issue areas. Two potential differences between the treatment of prisoners and civilians offer possible explanations for these contrary findings. First, the dangers of reciprocity might be more difficult to avoid in the case of prisoner abuse compared to civilian violence. Recent research shows democracies tend to fight more wars on foreign soil (Valentino, Huth, and Croco 2010, 538), which makes retaliating against their own civilians difficult. On the other hand, ground troops remain a necessity for prosecuting most wars, and this inevitably puts soldiers from democracies in harm’s way. Democracies may, therefore, be more sensitive to the prospects of retaliation from committing prisoner abuse compared to targeting civilians. It would be also interesting, however, to examine the degree to which violations in one area may spillover into others. Second, the perceived benefits of civilian violence remain an attractive option for many states despite evidence to the contrary (Pape 1996, 10). Prisoner abuse, in contrast, imposes several especially unattractive costs, such as likely retaliation or spawning a more resolute opponent. The ongoing debate over the usefulness of torture, however, indicates prisoner abuse could eventually follow a similar trend to civilian victimization should the benefits be perceived to outweigh the costs (Greenberg 2006). Although only probative, this discussion highlights the need to think more concertedly about how regime type may operate differently depending on the particular issue at stake. The diverging results for democracy caution against pooling observations across various aspects of the laws of war without first taking into account possible sources of variation.

Returning to the specific question of regime type and prisoner abuse, the quantitative results point to a consistent restraining effect for democracy. The existing data, however, are less able to discern the impact of institutional versus normative factors in shaping the wartime behavior of democracies. Further research could be devoted to more in-depth case studies that can better examine the processes through which democratic norms or institutions shape prisoner treatment.

The results suggest several additional avenues for future research. The analysis focused on the determinants for aggregate levels of prisoner abuse, but it is unclear whether these relationships would be the same for specific types of violations. For instance, some scholars have argued that under many conditions torture is often viewed as acceptable, or even necessary, in democratic societies (Rejali 2007, 22-23). Finally, I limited the analysis to wars between states, but it remains an open question whether similar results would hold across other types of conflict, such as civil wars and insurgencies. Examining differences in the conduct of state and nonstate actors toward prisoners provides a potentially valuable opportunity to develop a more general understanding of the conduct of belligerents across a wider range of conflicts.

Footnotes

Acknowledgments

For valuable comments, I would like to thank the editor, two anonymous reviewers, Matthew Evangelista, Peter Katzenstein, Jonathan Kirshner, Michael Koch, Mark Peffley, Sophia Wallace, Christopher Way, and Jessica Weeks. Sarah Kachovec and Laura Shahan provided excellent research assistance. All errors remain his own. Replication materials are available from the author’s website and at http://jcr.sagepub.com/.

Author’s Note

Earlier versions of this article were presented at the 2008 and 2009 annual meetings of the American Political Science Association. The author affiliation will be changed from the existing to Rutgers University on July 1, 2012.

Declaration of Conflicting Interests

The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author received no financial support for the research, authorship, and/or publication of this article.

Notes

References

Abdelal

Rawi

Herrera

Yoshiko M.

Johnston

Alastair Iain

McDermott

Rose

. 2006. “Identity as a Variable.” Perspectives on Politics 4 (4): 695–711.

Axelrod

Robert

. 1984. The Evolution of Cooperation. New York: Basic Books.

Bartov

Omer

. 2001. The Eastern Front, 1941-1945, German Troops and the Barbarisation of Warfare. 2nd ed. Oxford, UK: Palgrave.

Baudendistel

Rainer

. 2006. Between Bombs and Good Intentions: The Red Cross and the Italo-Ethiopian War, 1935-1936. New York: Berghahn Books.

Beaumont

Joan

. 1996. “Protecting Prisoners of War, 1939-1995.” In Prisoners of War and their Captors in World War II, 281-297, edited by Moore

Fedorowich

. Oxford, UK: Berg.

Best

Geoffrey

. 1994. War and Law Since 1945. Oxford, UK: Clarendon Press.

Brambor

Thomas

Clark

William R.

Golder

Matt

. 2006. “Understanding Interaction Models: Improving Empirical Analyses.” Political Analysis 14 (1): 63–82.

Bueno de Mesquita

Bruce

. 1981. The War Trap. New Haven, CT: Yale University Press.

Bueno de Mesquita

Bruce

Morrow

James D.

Siverson

Randolph M.

Smith

Alastair

. 1999. “An Institutional Explanation of the Democratic Peace.” American Political Science Review 93 (4): 791–801.

10.

Bueno de Mesquita

Bruce

Siverson

Randolph M.

. 1995. “War and the Survival of Political Leaders: A Comparative Study of Regime Types and Political Accountability.” American Political Science Review 89 (4): 791–808.

11.

Cienciala

Anna M.

Lebedeva

Natalia S.

Materski

Wojciech

, eds. 2007. Katyn: A Crime Without Punishment. New Haven, CT: Yale University Press.

12.

Davis

Gerald H.

1977. “Prisoners of War in Twentieth-Century War Economies.” Journal of Contemporary History 12 (4): 623–634.

13.

Dower

John W.

1986. War Without Mercy: Race and Power in the Pacific War. New York: Pantheon Books.

14.

Downes

Alexander B.

2006. “Desperate Times, Desperate Measures: The Causes of Civilian Victimization in War.” International Security 30 (4): 152–195.

15.

Downes

Alexander B.

2007. “Restraint or Propellant? Democracy and Civilian Fatalities in Interstate Wars.” Journal of Conflict Resolution 51 (6): 872–904.

16.

Downes

Alexander B.

2008. Targeting Civilians in War. Ithaca, NY: Cornell University Press.

17.

Downs

George W.

Rocke

David M.

Barsoom

Peter N.

. 1996. “Is the Good News about Compliance Good News about Cooperation?” International Organization 50 (3): 379–406.

18.

Doyle

Michael W.

1983. “Kant, Liberal Legacies, and Foreign Affairs, Part 2.” Philosophy and Public Affairs 12 (4): 323–53.

19.

Elliott

H. Wayne

. 1995. “Hostages or Prisoners of War: War Crimes at Dinner.” Military Law Review 149 (3): 241–74.

20.

Engelhardt

Michael J.

1992. “Democracies, Dictatorships and Counterinsurgency: Does Regime Type Really Matter?” Conflict Quarterly 12 (3): 52–63.

21.

Ferguson

Niall

. 1999. The Pity of War. New York, NY: Basic Books.

22.

Ferguson

Niall

. 2004. “Prisoner Taking and Prisoner Killing in the Age of Total War: Towards a Political Economy of Military Defeat.” War in History 11 (2): 148–92.

23.

Finnemore

Martha

Sikkink

Kathryn

. 1998. “International Norm Dynamics and Political Change.” International Organization 52 (4): 887–917.

24.

Gilkes

Patrick

Plaut

Martin

. 1999. War in the Horn: The Conflict between Eritrea and Ethiopia. London: Royal Institute of International Affairs.

25.

Goemans

Hein E.

2000. War and Punishment: The Causes of War Termination and the First World War. Princeton, NJ: Princeton University Press.

26.

Greenberg

Karen J.

, ed. 2006. The Torture Debate in America. New York: Cambridge University Press.

27.

Gross

Michael L.

2010. Moral Dilemmas of Modern War: Torture, Assassination, and Blackmail in an Age of Asymmetric Conflict. Cambridge: Cambridge University Press.

28.

Gutman

Roy

Rieff

David

, eds. 1999. Crimes of War: What the Public Should Know. New York: W.W. Norton.

29.

Guzman

Andrew T.

2008. How International Law Works: A Rational Choice Theory. Oxford, UK: Oxford University Press.

30.

Harff

Barbara

. 2003. “No Lessons Learned from the Holocaust? Assessing Risks of Genocide and Mass Murder since 1955.” American Political Science Review 97 (1): 57–73.

31.

Hata

Ikuhiko

. 1996. “From Consideration to Contempt: The Changing Nature of Japanese Military and Popular Perceptions of Prisoners of War Through the Ages.” In Prisoners of War and their Captors in World War II, edited by Moore

Fedorowich

, 253–54. Oxford, UK: Berg.

32.

Henderson

Errol A.

Tucker

Richard

. 2001. “Clear and Present Dangers: The Clash of Civilizations and International Conflict.” International Studies Quarterly 45 (2): 317–38.

33.

Holsti

Kalevi J.

1991. Peace and War: Armed Conflicts and International Order, 1648-1989. Cambridge: Cambridge University Press.

34.

Horne

Alastair

. 2006. A Savage War of Peace: Algeria, 1954-1962. New York: New York Review of Books.

35.

Huntington

Samuel P.

1993. “The Clash of Civilizations?” Foreign Affairs 72 (3): 22–49.

36.

Liddell Hart

B. H.

1946. The Revolution in Warfare. London: Faber and Faber.

37.

Lipson

Charles

. 2003. Reliable Partners: How Democracies Have Made a Separate Peace. Princeton, NJ: Princeton University Press.

38.

Locke

John

. (1690)1980. Second Treatise of Government. Indianapolis, IN: Hackett.

39.

MacKenzie

S. P.

1994. “The Treatment of Prisoners of War in World War II.” Journal of Modern History 66 (3): 487–520.

40.

MacKenzie

S. P.

1995. “The Shackling Crisis: A Case-Study in the Dynamics of Prisoner-of-War Diplomacy in the Second World War.” International History Review 17 (1): 79–98.

41.

Mahoney

James

Goertz

Gary

. 2004. “The Possibility Principle: Choosing Negative Cases in Comparative Research.” American Political Science Review 98 (4): 653–69.

42.

Marshall

Monty G.

Jaggers

Keith

. 2007. Polity IV Project: Political Regime Characteristics and Transitions, 1800-2006. Severn, MD: Center for Global Policy, School of Public Policy, George Mason University, and Center for Systemic Peace.

43.

Mearsheimer

John

. 1994-1995. “The False Promise of International Institutions.” International Security 19 (3): 5–49.

44.

Merom

Gil

. 2003. 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.

45.

Moore

Bob

. 2000. “Unruly Allies: British Problems with the French Treatment of Axis Prisoners of War, 1943–1945.” War in History 7 (2):180–98.

46.

Morrow

James D.

2007. “When Do States Follow the Laws of War?” American Political Science Review 101 (3):559–72.

47.

Morrow

James D.

Hyeran

. 2006. “Compliance with the Laws of War: Dataset and Coding Rules.” Conflict Management and Peace Science 23 (1):91–113.

48.

Mueller

John E.

1973. War, Presidents, and Public Opinion. New York: Wiley.

49.

O’Donnell

Guillermo

Schmitter

Philippe C.

. 1986. Transitions from Authoritarian Rule: Prospects for Democracy. Baltimore: Johns Hopkins University Press.

50.

Pape

Robert A.

1996. Bombing to Win: Air Power and Coercion in War. Ithaca, NY: Cornell University Press.

51.

Rabbie

Jacob M.

1989. “Group Processes as Stimulants of Aggression.” In Aggression and War, 141-156, edited by Groebel

Hinde

. Cambridge: Cambridge University Press.

52.

Rachamimov

Alon

. 2002. POWs and the Great War: Captivity on the Eastern Front. Oxford, UK: Berg.

53.

Rees

Laurence

. 1999. War of the Century: When Hitler Fought Stalin. New York: New Press.

54.

Reiter

Dan

Stam

Allan C.

. 2002. Democracies at War. Princeton, NJ: Princeton University Press.

55.

Rejali

Darius

. 2007. Torture and Democracy. Princeton, NJ: Princeton University Press.

56.

Reus-Smit

Christian

. 2004. “The Politics of International Law.” In The Politics of International Law, edited by Reus-Smit

. 14–44 Cambridge: Cambridge University Press.

57.

Roberts

Adam

. 1994. “Land Warfare: From Hague to Nuremberg.” In The Laws of War: Constraints on Warfare in the Western World, edited by Howard

Andreopoulos

G. J.

Shulman, 116-139

M. R.

. New Haven, CT: Yale University Press.

58.

Roland

Charles G.

1996. “Human Vivisection: The Intoxication of Limitless Power in Wartime.” In Prisoners of War and their Captors in World War II, edited by Moore

Fedorowich

, 149–79. Fedorowich. Oxford, UK: Berg.

59.

Rossino

Alexander B.

2003. Hitler Strikes Poland: Blitzkrieg, Ideology, and Atrocity. Lawrence: University Press of Kansas.

60.

Rummel

R. J.

1995. “Democracy, Power, Genocide, and Mass Murder.” Journal of Conflict Resolution 39 (1): 3–26.

61.

Russett

Bruce

. 1993. Grasping the Democratic Peace. Princeton, NJ: Princeton University Press.

62.

Sarkees

Meredith

Wayman

Frank W.

. 2010. Resort to War: 1816-2007. Washington, DC: CQ Press.

63.

Simmons

Beth

. 2010. “Treaty Compliance and Violation.” Annual Review of Political Science 13: 272–96.

64.

Singer

J. David

Bremer

Stuart

Stuckey

John

. 1972. “Capability Distribution, Uncertainty, and Major Power War, 1820-1965.” In Peace, War, and Numbers, edited by Russett

B. M.

, 19–48. Beverly Hills, CA: Sage.

65.

Slaughter

Anne-Marie

. 1995. “International Law in a World of Liberal States.” European Journal of International Law 6 (1): 505–38.

66.

Speed

Richard B.

1990. Prisoners, Diplomats and the Great War: A Study in the Diplomacy of Captivity. New York: Greenwood Press.

67.

Springer

Paul J.

2010. America’s Captives: Treatment of POWs from the Revolutionary War to the War on Terror. Lawrence: University Press of Kansas.

68.

Stachura

Peter D.

2004. Poland, 1918-1945: An Interpretive and Documentary History of the Second Republic. New York, NY: Routledge.

69.

Streit

Christian

. 1986. “The German Army and the Policies of Genocide.” In The Policies of Genocide: Jews and Soviet Prisoners of War in Nazi Germany, edited by Hirschfeld

. 1–14 London: Allen and Unwin.

70.

Tsouras

Peter

. 2000. Greenhill Dictionary of Military Quotations. Mechanicsburg, PA: Stackpole Books.

71.

Valentino

Benjamin A.

Huth

Paul K.

Balch-Lindsay

Dylan

. 2004. ““Draining the Sea”: Mass Killing and Guerrilla Warfare.” International Organization 58 (2): 375–407.

72.

Valentino

Benjamin A.

Huth

Paul K.

Croco

Sarah

. 2006. “Covenants Without the Sword: International Law and the Protection of Civilians in Times of War.” World Politics 58 (3): 339–77.

73.

Valentino

Benjamin A.

Huth

Paul K.

Croco

Sarah E.

. 2010. “Bear Any Burden? How Democracies Minimize the Costs of War.” The Journal of Politics 72 (02): 528–44.

74.

Vance

Jonathan F.

2006. Encyclopedia of Prisoners of War and Internment. 2nd ed. Millerton, NY: Grey House.

75.

Walzer

Michael

. 2000. Just and Unjust Wars. 3rd ed. New York: Basic Books.

76.

Warmbrunn

Werner

. 1993. The German Occupation of Belgium, 1940-1944. New York: Peter Lang.

77.

Weart

Spencer R.

1998. Never at War: Why Democracies Will Not Fight One Another. New Haven, CT: Yale University Press.

78.

Zook

David H.

1960. The Conduct of the Chaco War. New York: Bookman Associates.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.04 MB

0.00 MB