Responses

THE llNIVE-itSITY O• i I THE UNIVERSITY OF CHICAGO PRESS JOUR ALS C H I C A G ·O Tl-IE LAW Cl-1001.

u JSTOR

The Laws of War, Common Conjectures, and Legal Systems in International Politics

Author(s): James D. Morrow

Source: The Journal of Legal Studies , Vol. 31, No. S1, Rational Choice and International LawA Conference Sponsored by the University of Chicago Law School (January 2002), pp. S41-S60

Published by: The University of Chicago Press for The University of Chicago Law School

Stable URL: https://www.jstor.org/stable/10.1086/340810

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at https://about.jstor.org/terms

The University of Chicago Press and The University of Chicago Law School are collaborating with JSTOR to digitize, preserve and extend access to The Journal of Legal Studies

This content downloaded from �������������131.94.186.70 on Mon, 25 Nov 2019 19:14:54 UTC�������������

All use subject to https://about.jstor.org/terms

THE LAWS OF WAR, COMMON CONJECTURES, AND LEGAL SYSTEMS IN INTERNATIONAL

POLITICS

JAMES D. MORROW*

Abstract

The laws of war have a mixed record of limiting violence during war. I explain these treaties as the codification of the common conjecture of an equilibrium of a game theory model of war. These laws succeed only when the parties comply on their own or reciprocal sanctions dissuade parties who are willing to violate the agreement. Treaties support reciprocity by specifying when reciprocal sanctions are appropriate, by controlling the response to violations outside the control of state parties, and by screening out states that are unwilling to abide by a treaty through their refusal to ratify it. I discuss the design of the laws of war: why they are multilateral and general rather than bilateral and specific to each war, why treaties are separated by subject matter, why they devolve responsibility for individual vi- olations, and the difficulties in determining how strict these laws should be.

The twentieth century saw the advent and explosion of the laws of war, formal treaties that attempt to regulate military conduct during wartime. These treaties have a mixed record of compliance; some, such as bans on the use of chemical weapons, have generally succeeded; others, such as the protection of civilians in civil wars, have generally failed; and others, such as the conventions on prisoners of war (POWs), have been observed in some cases and not in others. This paper presents an explanation of how these treaties could work and why they fail in some cases. It also explores the question of which agreements states are willing to make.

International law must be self-enforcing to be effective. Successful self- enforcement of law among states requires either that they must be willing to live up to their legal obligations regardless of what others do or that reciprocity must deter violations. The laws of war limit violence in wartime when states at war observe the treaties they have accepted through ratification.

* Department of Political Science, University of Michigan, Ann Arbor. Originally presented at the conference Rational Choice and International Law at the University of Chicago Law School, April 27–28, 2001. I would like to thank the participants at the conference; their comments were very helpful in revising the paper. I would particularly like to thank Saul Levmore and Eric Posner for their comments.

[Journal of Legal Studies, vol. XXXI (January 2002)] � 2002 by The University of Chicago. All rights reserved. 0047-2530/2002/3101-0011$01.50

S41

This content downloaded from 131.94.186.70 on Mon, 25 Nov 2019 19:14:54 UTC

All use subject to https://about.jstor.org/terms

S42 the journal of legal studies

The combination of reciprocal enforcement and state values is central to the success of these treaties and international law generally.

Reciprocal sanctions could fail for three reasons. First, the treaty might not be sufficiently specific about what actions are unacceptable, which leads to a spiral of mistaken retaliation. Second, violations outside the control of the warring parties could be interpreted as intentional violations of the treaty, which trigger mistaken retaliation. Third, the sanctions might not deter de- liberate violations by a state. A treaty system can reduce the consequences of each of these three failures for states that intend to comply with limitations on violence.

Treaties address the first problem by detailing what acts are unacceptable. The combination of lawful intent and reciprocity can work only when the parties share an understanding about what behavior is unacceptable. Other- wise, reciprocal punishments could be triggered against a state that sought to fulfill its legal obligations. The retaliating state might understand an act of the opposing side to be a violation when the opposing state believed its action was acceptable under their agreement. The latter state in turn might perceive the retaliatory acts of the former state as initial violations of their agreement. Specificity in a treaty ameliorates this problem by producing a common document that states accept on ratification.

Treaties address the second problem in two ways. The responsibility to enforce violations by individual combatants is devolved to ratifying states; a state can more easily investigate and punish its own soldiers than the states who suffer the crimes of those individuals. The control of individual vio- lations limits the possibility that such violations trigger mistaken retaliation, even if that control is erratic. Treaties also help to create common expectations about the level and type of violations outside the control of states. Often, warring states must judge their opponent’s compliance by inferring intentions from behavior on the battlefield. Shared expectations assist warring states in judging whether the pattern of violations they observe is the result of indi- vidual violations or state policy.

Treaties address the third problem by screening out states unwilling to comply with a treaty through ratification of that treaty. States that do not ratify a treaty generally do not comply with the standards of that treaty. Failure by a state to ratify a treaty indicates its likely intention to violate that treaty in the event of war. Other states then can anticipate that state’s likely violations and attempt to limit their effect.

I argue that treaties, and international law more generally, aid reciprocal enforcement of agreements. The dual roles of treaties—to create a shared understanding of unacceptable conduct and to screen out those who will not comply—are the central points of my argument. I begin with a general discussion of the role of shared understandings in game theory. I review a series of models that explain how the laws of war work during wartime. On the basis of this analysis, I discuss the strategic operation of the laws of war,

This content downloaded from �������������131.94.186.70 on Mon, 25 Nov 2019 19:14:54 UTC�������������

All use subject to https://about.jstor.org/terms

war and international politics S43

how a shared understanding of unacceptable conduct is central to compliance with the standards of those laws, and how ratification of a treaty helps to address uncertainty about other states’ intentions by screening out those states that do not accept the standard. I then discuss the practical political problems of the laws of war from the perspective of my argument. There are three problematic areas in treaty formation—differences among states about the appropriate standard of conduct, reciprocal enforcement when states cannot completely observe the causes of violations, and the need for agreements to operate at the individual level as well as the state level—which must be appropriately dealt with to successfully restrain violence during war. I explain two characteristics of the laws of war, multilateralism and fire walls—the division of different issues in the laws of war into separate treaties—by examining the incentives of states to make such agreements in advance of war. I close with the paradox in this approach to international law.

I. Treaties as Shared Understandings

Reciprocal enforcement depends on a shared understanding of what con- duct is unacceptable and what consequences follow from such conduct. Not all possible shared understandings are enforceable. The parties may not be willing to carry out reciprocal punishment, that punishment may be insuf- ficient to deter violations, or difficulties monitoring the standard could make effective reciprocity impossible. Laws of war can be effective only to the extent that the parties can enforce them against one another; they must possess both the ability and the willingness to make the treaty work.

When the parties can enforce a treaty among themselves, a wide range of possible treaties are enforceable. The specific agreement matters. In the ab- sence of a shared understanding of which agreement is in effect, actors may misinterpret one another’s actions or may respond inappropriately to others’ acts. International law seeks in part to create such a shared understanding among the parties. Treaties codify shared understandings of what behavior is unacceptable and what consequences may follow.1 Ratification of a treaty shows that a state publicly accepts that standard.

Effective law of war requires both a shared understanding and the ability of states to enforce the law on one another. Political institutions, of which international law is an example, can be thought of as an equilibrium in a

1 International law at times attempts to create novel agents who can assist with the enforce- ment of a shared understanding on conduct. Dispute panels are one example of such agents. See Paul R. Milgrom, Douglass C. North, & Barry R. Weingast, The Role of Institutions in the Revival of Trade: The Medieval Law Merchant, Private Judges, and the Champagne Fairs, 2 Econ. & Pol. 1 (1990), for a game-theoretic attempt at examining the consequences of the creation of an agent, the medieval law merchant, who collects and disseminates information to enhance reciprocal enforcement.

This content downloaded from �������������131.94.186.70 on Mon, 25 Nov 2019 19:14:54 UTC�������������

All use subject to https://about.jstor.org/terms

S44 the journal of legal studies

game.2 An equilibrium requires strategies that are mutual best replies and a common conjecture that all are playing their equilibrium strategies. The former guarantees that no player wishes to deviate from her equilibrium strategy; the latter that all share a common understanding of how they will play the game. International law can be thought of as the codification of the common conjecture of a specific equilibrium, and the players’ strategies on that equilibrium path describe the behavior under that law. We can then assess how well behavior complies with the standards embodied in a treaty.

This analysis shifts our attention from studying only the norms and stan- dards in a treaty to understanding the political problems of those norms and standards in practice. Legal systems, in the sense of how specific laws are understood, implemented, and enforced by parties, are central to the analysis. Does a shared understanding operate as the common conjecture of an equi- librium of military competition during wartime, and what behavior should result from that shared understanding?

II. Modeling the Enforcement of the Laws of War

If the laws of war codify the common conjecture underlying an equilibrium that describes conduct during wartime, we need to understand the logic of such an equilibrium and the range of possible equilibria. The former explains how and when an agreement limits violence during wartime; the latter details the set of enforceable agreements. This section discusses a progression of models of war, the laws of war, and their enforcement during war. The conclusions of these models of the laws of war illuminate why states might choose to create such agreements in addition to furthering our understanding of how they operate in practice. I sketch the logic of these models rather than present them formally.3

War can be thought of as a strategic competition over the stakes in dispute, such as territory. This competition can be represented with a war-of-attrition model where the sides fight over the military balance between themselves.4

In each round, each state chooses whether to quit or continue the war and a battle strategy if it chooses to continue the war. The war continues until one side concedes the stakes to the other by quitting.5 If both sides continue the war in a round, the combination of their battle strategies shifts the military

2 Andrew Schotter, The Economic Theory of Social Institutions (1982); and Randall L. Calvert, The Rational Choice Theory of Social Institutions: Cooperation, Coordination, and Communication, in Modern Political Economy 216 (Jeffrey S. Banks & Eric A. Hanushek eds. 1995).

3 See James D. Morrow, Strategy, Victory, and the Laws of War (unpublished manuscript, Univ. Michigan, Dep’t Pol. Sci. 2002), for formal presentations.

4 For a brief introduction to war-of-attrition models, see Drew Fudenberg & Jean Tirole, Game Theory 119–26 (1991).

5 If both sides choose to quit in the same round, we assume that neither receives the stakes.

This content downloaded from �������������131.94.186.70 on Mon, 25 Nov 2019 19:14:54 UTC�������������

All use subject to https://about.jstor.org/terms

war and international politics S45

balance from its current value and imposes costs on both sides.6 Each side’s costs of fighting in a round depend on the current military balance, with costs rising as balance shifts against a state, and the battle strategies of both sides. Each round can be thought of as one battle in the war, and the sides choose their battle strategies in the hope of shifting the balance in their favor.

A side will quit when it is losing badly enough on the battlefield, that is, when the military balance shifts far enough against it that it prefers to quit rather than continue to fight. A state’s break point is the value of the military balance at which it is indifferent between quitting and continuing to fight. A state will quit once the military balance is less favorable than its break point.

The model views battle outcomes as determined solely by the interaction of the sides’ battle strategies. Each side’s battle strategy specifies not only battle plans, which are conventionally thought of as military strategy, but also details of the employment of troops and weapons that could affect the result or costs of the battle. For instance, a battle plan executed using chemical weapons is a different battle strategy from the same battle plan executed without chemical weapons. Each player’s set of battle strategies is very large then because many details of doctrine, deployment, and conduct on the bat- tlefield determine the outcomes of battles. Although many of the battle strat- egies may differ slightly, they are separate strategies unless they produce identical effects against all the strategies of the other player.

The laws of war can be thought of as a prewar agreement by the sides to abstain from using certain battle strategies during the war. An agreement not to use chemical weapons, for example, merely means that the sides have agreed not to use any battle strategy that employs chemical weapons. Of course, the agreement does not prevent them from using those battle strategies in a war; it merely specifies which strategies are deemed unacceptable.

A law of war is enforceable when neither side is willing to play any of its proscribed battle strategies. Such compliance could occur because neither side believes that banned strategies are effective against lawful ways of fighting. If the sides believe that banned strategies have military value, they still might be able to enforce an agreement through reciprocity, which is understood as the abandonment of the agreement after a first use of a pro- scribed battle strategy. First use removes the agreed-upon restraints and leads to a war in which all battle strategies are available. A side contemplating first use of a banned strategy assesses any immediate value from both the first use and the long-run consequences of the abandonment of limits in the war. In addition, the leader of a state may face audience costs for his violation of the prewar agreement.7 Domestic or international audiences may take actions against a leader after first use of a proscribed strategy, reducing his

6 Formally, the range of battle strategies and outcomes is given by a zero-sum game. 7 See James D. Fearon, Domestic Political Audiences and the Escalation of International

Disputes, 88 Am. Polit. Sci. Rev. 577 (1994), for the concept of audience costs.

This content downloaded from �������������131.94.186.70 on Mon, 25 Nov 2019 19:14:54 UTC�������������

All use subject to https://about.jstor.org/terms

S46 the journal of legal studies

willingness to break an agreement. Reciprocity enforces a law of war when the combination of long-run consequences and audience costs exceeds the short-run gain of first use and so prevents violations.

The overall attraction of violating an agreement to limit violence can be assessed by comparing the warring parties’ break points with the constraints as opposed to without them. If both break points shift in favor of one side without the constraints, this side is more likely to win the war and so gains from violating the agreement. If both break points move in opposite direc- tions, so that both sides require a higher or lower level of defeat before surrendering, the effect of the agreement on the chances of each side winning is indeterminate. Compliance with such an agreement is more likely because it does not favor one side over the other. As mentioned above, advantages from first use could also lead to the breakdown of an agreement because a side could find the shift in the military balance from first use great enough to shift which side is likely to win. A law of war is more likely to be observed during wartime under the following conditions: 1. when neither side would use the proscribed strategies even if they were

available; 2. when the military benefits of first use are small; 3. when the long-run military effect of the proscribed strategies does not

favor one side over the other; and 4. when audience costs for violations are large.

The problems of enforcement loom in the desirability of any agreement and in the question of which agreements states should be willing to conclude. Agreements that cannot be enforced in any case have little reason to exist in the logic of the model.8 First and foremost, a formal agreement helps to specify which battle strategies are acceptable and which are unacceptable. Because the battle strategies in the model detail how the sides fight as well as their battle plans, there may only be small differences in conduct across many strategies. It could be very important to distinguish these small dif- ferences in creating a common conjecture about which strategies are unac- ceptable during wartime. For example, is the use of riot-control agents (namely, tear gas) illegal under the 1925 Geneva Protocol that forbids the use of chemical weapons? This level of detail is necessary to specify which battle strategies are unacceptable. Again, states could disagree on which precise actions are unacceptable even if they share general values on proper conduct during wartime. Legalization here aids the limitation of violence by reducing confusion about whether a side has violated an agreement.

Further, the large number of strategies with small differences across many of them suggests that a wide range of agreements that differ slightly could

8 I understand that others may see value in agreements that cannot be enforced if those agreements foster the coalescence of new political forces in favor of the values in those agreements. Such considerations are beyond this model and the discussion in this paper.

This content downloaded from �������������131.94.186.70 on Mon, 25 Nov 2019 19:14:54 UTC�������������

All use subject to https://about.jstor.org/terms

war and international politics S47

all be enforceable. If an agreement that battle strategies {1, … , n} are ac- ceptable and that battle strategies {n � 1, … , n � k} are unacceptable can be enforced, then an agreement that shifts strategy {n � 1} from unacceptable to acceptable can probably also be enforced. When there are multiple equi- libria, such as the multiple standards that are enforceable here, the specific common conjecture determines which equilibrium the players are playing. Formal treaties on laws of war specify the standard of proper behavior out of the set of enforceable standards. This specificity aids in compliance be- tween states that are willing to limit violence during wartime.

The sketch of the model above assumes that the war is fought under complete information—the players know one another’s values for the out- comes of the game. This common knowledge covers each side’s value of the prize—the common source of incomplete information in war-of-attrition models—and the costs and results of each pair of strategies.9 The sides do not differ in their judgments of the efficacy of battle strategies, and so they can predict when prewar agreements will be violated. In actuality, states differ in their judgments about the desirability of various laws of war and they do not know completely other states’ willingness to honor agreed-upon standards. In the area of treatment of POWs, for instance, some states act to protect POWs, with the effect of encouraging surrender by enemy troops, while other states choose to exploit POWs, whom they hold for information, labor, or to encourage retaliation on the battlefield and thus discourage their own soldiers from surrendering.10 Asymmetric information about the value of battle results, both in costs that each side suffers and in the shift in the military balance, could reflect these differences in willingness to comply with the laws of war.

States facing such uncertainty about their opponent’s willingness to comply must judge compliance from reports about conduct on the battlefield and behind the lines. These judgments are complicated by uncertainty about whether observed violations are the product of a decision by the opponent to violate the agreement or individual action contrary to a state policy. The former case calls for a reciprocal response; the latter does not. Indeed, reprisal in the latter case could lead to a collapse of the agreement when both sides

9 Readers should be aware that the following discussion is not based on a formal model, but rather on an understanding of the strategic dynamics that commonly result from incomplete information.

10 See Richard Overy, Russia’s War: A History of the Soviet War Effort, 1941–1945, at 297–98 (1997), on Soviet use of German prisoners as labor; A. J. Barker, Prisoners of War 99–100 (1975), on German use of Soviet prisoners as labor; Ikuhiko Hata, 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, at 253 (Bob Moore & Kent Fedorowich eds. 1996), on Japanese policies designed to make their own troops less willing to surrender; and Gerald F. Linderman, The World within War: America’s Combat Experience in World War II 128–29 (1997), on how relative leniency by U.S. forces encouraged German soldiers to surrender.

This content downloaded from �������������131.94.186.70 on Mon, 25 Nov 2019 19:14:54 UTC�������������

All use subject to https://about.jstor.org/terms

S48 the journal of legal studies

sought to comply. This noise varies with the specific subject of the agreement. It is unlikely that individual commanders can use chemical weapons without higher authorization, while common soldiers hold the power to summarily execute soldiers who are attempting to surrender on the battlefield.

The models discussed above assume that states have perfect control of their militaries, which is contrary to the general discussion in this section. Atrocities by individual soldiers can violate state policy to comply with existing treaties, which creates both noise and the two-level problem. This section draws on models of the problems of reciprocity under noise and a two-level problem to discuss how the legal system of the laws of war ad- dresses these problems.

Noise complicates the problem of determining when a state should take reprisals11 against perceived violations by the other side. Were those atrocities a result of state policy or individual violations? A rational response requires ignoring low-level violations while responding to more significant viola- tions.12 In practice, states create “bright lines” that trigger retaliation when crossed. Because observed atrocities could combine state and individual vi- olations and the other side is likely to respond in turn to reprisals, responding to all violations could unravel the agreement through escalating reprisals. When the line is crossed, however, retaliation should be disproportionate to the triggering offense. Disproportionate responses restore the deterrent effect when states do not retaliate against every violation. How tolerant of minor violations the bright line is depends on the magnitude of noise present on the issue. Issues that face little possibility of noise, such as chemical weapons, have a quick trigger for retaliatory response; issues with a great deal of noise, such as POWs, should allow a substantial level of individual atrocities before retaliation occurs.

These uncertainties about compliance, both anticipated before a war and actual during a war, complicate enforcement. Uncertainty about the oppo- nent’s willingness to live up to an agreement is compounded by uncertainty about its compliance during the war. As Section III discusses, decreasing both of these uncertainties is one of the main reasons for formal treaties on the laws of war.

III. Ratification as Screening

Inferring other states’ intentions to live up to a standard in advance of war is a deeper problem. Here agreement on a single international standard could

11 I use “reprisal” in its original meaning of a legal retaliatory response, as opposed to its current meaning of a retaliatory atrocity.

12 Edward J. Green & Robert H. Porter, Noncooperative Collusion under Imperfect Price Information, 52 Econometrica 87 (1984); George W. Downs & David M. Rocke, Tacit Bar- gaining, Arms Races, and Arms Control (1990); and George W. Downs & David M. Rocke, Optimal Imperfection? Domestic Uncertainty and Institutions in International Relations (1995).

This content downloaded from �������������131.94.186.70 on Mon, 25 Nov 2019 19:14:54 UTC�������������

All use subject to https://about.jstor.org/terms

war and international politics S49

help states separate themselves into those that accept the standard and those that do not. The process of ratification signifies acceptance by the ratifying state of the specific standards in a treaty. Of course, states sometimes do violate treaties that they have ratified, but failure to ratify is still a clear signal of a state’s refusal to accept the standards of a treaty. During World War II, the failure to ratify the 1929 Geneva Convention on POWs by at least one member of a warring dyad corresponds well with death rates of POWs held by either member of that dyad and with historians’ descriptions of the brutality of combat in those dyads.13 Joint ratification then correlates with stronger compliance with the standards for treatment of POWs.

Given that ratification of a treaty does not prevent a state from violating that treaty in future wars, why do some states refuse to ratify treaties on the laws of war? There are audience costs to ratifying a treaty cynically, with the intention of ignoring its conditions. Other states may be reluctant to make agreements with states that treat treaty obligations cavalierly, which creates an international audience. Governments also answer to domestic audiences. Between the world wars, Japan sought to instill in its soldiers a code of conduct that equated surrender on the battlefield with dishonor greater than death. The Japanese army believed that the fighting spirit of its soldiers was the one advantage it had over possible opponents and so instructed recruits in this code to induce them to fight to the death.14 Ratifying the 1929 Geneva Convention would have sent conflicting signals to Japanese soldiers about the dishonor of surrendering to the enemy and would have possibly under- mined the training regimen. Refusal to ratify can then be in the interest of a state that has no intention of following the standards in the treaty.

Ratification then can operate as a screen on the intentions of states to observe the standards of a treaty. Those who do not ratify a treaty screen themselves out as being unwilling to live up to the treaty. The process is screening as opposed to signaling because states either accept or reject the treaty instead of stating publicly their future conduct during wartime. The latter form of signaling would not establish a common conjecture that the state would follow its stated standard, nor would such declaration necessarily provide the level of detail generally provided in a treaty to specify which battle strategies were unacceptable. State declaration of intent also creates the problem of determining what standard holds when warring parties have declared different standards in advance. The ability to state reservations upon ratification of a treaty does allow states to signal some of their intentions about their future conduct. They can specify possible opponents against whom they are not bound by the treaty; they can also specify particular provisions of the treaty that they find unacceptable. Because reservations to a treaty are

13 See James D. Morrow, The Institutional Features of the Prisoners of War Treaties, 55 Int’l Org. 971 (2001).

14 Barker, supra note 10, at 122; Hata, supra note 10.

This content downloaded from �������������131.94.186.70 on Mon, 25 Nov 2019 19:14:54 UTC�������������

All use subject to https://about.jstor.org/terms

S50 the journal of legal studies

specific and legally binding, they provide a stronger signal of state intentions than a simple statement would.

For the states that intend to comply with the treaty, screening out likely violators allows them to have a better idea of whether an opponent will violate the standards when a war begins. They may be able to prepare them- selves to respond appropriately to reduce possible effects of first use. Their creation of a single standard through treaty negotiations establishes a screen that could separate out some states that would not live up to the standard. Multiple levels of commitment to a treaty would enhance screening by sep- arating states into further classes of likely levels of compliance in the future. However, such multiple levels of commitment create the problem of what standard exists between warring states that accept different levels. Reser- vations provide some flexibility to adjust a ratifying state’s precise com- mitment without creating such difficulties.

Screening, when successful, divides states into those who intend not to comply with the treaty and those who probably will. This division suggests another role of common conjectures in the laws of war, in-group versus out- group enforcement. If states can separate themselves publicly into these two groups, then members of the in-group can act differently toward members of the out-group, which makes their shared standard stronger. For example, the Chemical Weapons Convention separates states by ratification and re- stricts trade in chemical agents and precursors from ratifying states to non- ratifying states. This division provides a positive incentive to ratify the treaty and accept snap inspections of possible production sites for chemical weap- ons. Because almost all industrialized states have ratified the treaty, states unable to produce these chemicals themselves have a strong incentive to join the in-group to import those chemicals. This mechanism requires both a shared understanding of which states are in which group and what chemicals are restricted.

IV. The Practical Politics of War

In practice, the laws of war must address three main issues: variation in state motivations concerning proper conduct, violations by individuals as opposed to state violations, and the difficulty of monitoring compliance. The importance of each of these issues varies with the specific issue addressed in a treaty. The way a specific agreement works in practice (or fails to work) depends on the magnitude of the problem that each of these issues poses for that agreement. The practical problems of regulating the use of chemical weapons differ from those that exist in the protection of POWs.

States vary greatly in their interest in supporting restrictions on violence during wartime. For instance, the United States has generally tried to uphold those treaties that it has ratified but will not consider legal restrictions on aerial bombardment. State policy on specific restrictions on war varies with

This content downloaded from �������������131.94.186.70 on Mon, 25 Nov 2019 19:14:54 UTC�������������

All use subject to https://about.jstor.org/terms

war and international politics S51

the strategic effects of the limitation, a desire to curry favor with public opinion domestically or internationally, and the moral stance of the state. These variations drive both a problem of distribution about which standards to adopt—different states prefer different standards—and a problem of en- forcement—some states will not live up to an agreed-upon standard.

The strategic effects of limiting violence cover a wide range of reasons why states support or oppose such agreement in both negotiations and prac- tice. Militarily, an agreement may have an asymmetric effect if it is successful, such as the restrictions on submarine warfare that advantaged Great Britain relative to Germany in the world wars. The side disadvantaged by that agree- ment is unlikely to comply. Such asymmetric effects need not take place on the battlefield to have military consequences; states can gain an advantage by exploiting POWs as labor, as the Japanese and Nazis did during World War II.15 Precisely because such asymmetries are likely to lead states to ignore commitments, they pose a particular problem for enforcement of a candidate agreement. Even agreements that would be symmetric in their effects may be difficult to enforce because neither side wishes to abandon its freedom to employ certain acts or weapons. The protection of civilians in civil wars is often difficult because both the government and rebels use violence against civilians to intimidate them. Some weapons that cause sub- stantial collateral damage, such as high-powered artillery, are seen as too valuable militarily to be restricted through legal agreement.

States may choose to comply with standards to seek support elsewhere. For example, Japan treated Russian prisoners well during the Russo-Japanese War in order to attract support from Western nations.16 Often, the target audience is domestic. Attempts by the U.S. military to portray recent bom- bardments as precise, controlled, and aimed solely at military targets seek to prevent a loss of public support for such military action. Similarly, states may comply with codes of conduct because their leaders believe that humane treatment is moral and appropriate. In some situations, government leaders must resist pressure from their own civilians to treat the enemy more harshly.17

Some states may be unwilling to engage in reciprocal enforcement of the laws of war because they believe that they are better off upholding the standard no matter what the other side does. For instance, the United States and Great Britain were unwilling to respond in kind to Japanese mistreatment of POWs. To the extent that there was a reciprocal response, it came on the battlefield in the unusual ferocity of combat on Pacific islands.18

15 Barker, supra note 10, at 99–112. 16 Jonathan F. Vance, Objects of Concern: Canadian Prisoners of War through the Twentieth

Century 21–22 (1994). 17 See John W. Dower, War without Mercy 53–57 (1986), on attitudes of U.S. civilians

toward the Japanese during World War II. 18 See Linderman, supra note 10, ch. 4.

This content downloaded from �������������131.94.186.70 on Mon, 25 Nov 2019 19:14:54 UTC�������������

All use subject to https://about.jstor.org/terms

S52 the journal of legal studies

These variations in preferred policy create problems for setting a single standard and then for enforcing that standard. A treaty requires resolving the distributional problem inherent in selecting a standard. Further, because no treaty fully specifies all possible contingencies, states cannot know the full consequences of a treaty when it is adopted. This combination creates a conflict between distribution and information at the heart of coordination on how to cooperate.19 Even if a single standard is codified in a treaty, some states may believe they will be better off ignoring the standard. Such vio- lations could be opportunistic or premeditated. Japan neither signed nor rat- ified the 1929 Geneva Convention on POWs because it already had policies to discourage Japanese soldiers from surrendering and viewed the Geneva standard with scorn.20 States may be able to deter opportunistic defections through reciprocal sanctions, but states committing premeditated violations prefer the breakdown of a treaty to compliance.

The laws of war attempt to control both individual and state behavior. For many areas of concern, such as the protection of civilians, controlling the behavior of individual soldiers on the battlefield is necessary for effective limitation of violence. Individual soldiers commit violations on their own even if they are contrary to state policy. For other issues, such as chemical weapons, use is controlled centrally, reducing the possibility of soldiers vi- olating a standard on their own initiative. As pointed out above, behavior at the state and individual levels is linked: state violations often lead to unusual violence on the battlefield, and large-scale individual violations can trigger retaliation by states.

This two-level problem—that treaties must operate on both the individual and state levels—creates two issues. First, there will always be some vio- lations when the policing of individuals is critical for an issue. Soldiers commit crimes against prisoners and civilians even in the best-disciplined armies. Perfect compliance cannot be expected. The management of indi- vidual violations is generally left to the militaries of those violators. A vi- olator’s own military justice system is more likely to be able to collect the information to determine what happened and bring the violator to trial. That system of devolved responsibility requires active state participation in dis- ciplining its own soldiers. State policy can encourage individual violations in two ways: active policy to promote atrocities and neglect of discipline. The training and discipline of soldiers in the laws of war are essential for the limitation of individual violations.

Second, individual violations compound the problem of determining whether the other side is complying with a treaty. The protection of sailors on the high seas requires that individual ship captains protect the survivors

19 See James D. Morrow, Modelling the Forms of International Cooperation: Distribution versus Information, 48 Int’l Org. 387 (1994).

20 Hata, supra note 10.

This content downloaded from �������������131.94.186.70 on Mon, 25 Nov 2019 19:14:54 UTC�������������

All use subject to https://about.jstor.org/terms

war and international politics S53

of ships that they sink. However, atrocities on the high seas are difficult for the state of the victims to discover simply because there may be no survivors. In the absence of public state declarations of policy and intent, other states must infer those policies from observed violations. Because some violations occur even when state policy supports a treaty and other violations are hidden from observation because they occur behind enemy lines, knowing when to respond to a pattern of observed violations is difficult.

Individual violations then are a source of the third practical problem in reciprocal enforcement: the difficulty of determining whether the other side has complied. We can think of this problem as the problem of inferring a signal in the presence of noise—that is, inferring state policy when unintended violations may occur. Noise creates the possibilities of failing to retaliate when you should and of retaliating when you should not. Both types of error undermine a standard, the former by reducing deterrent effect, the latter by creating reciprocal spirals of atrocities. Because reciprocal sanctions are often actions that would be unacceptable under the treaty if they were not in response, the shared perception that an act is a violation or acceptable re- sponse to a violation is critical to avoiding spirals. Noise undermines such shared perceptions of acts.

Noise also arises from acts that are hard to observe. For instance, POW camps are behind enemy lines and so not open to direct observation by the home state of the prisoners. Further, the detaining power probably does not wish to allow observers from the home state to inspect camps out of the fear of contact with the prisoners. One solution to such problems is neutral ob- servers, and the Red Cross plays that role with respect to POWs by requiring states to collect and submit lists of POWs they hold, inspecting camps, and delivering mail and packages to POWs. Often the first sign that a state is violating the POW standards is its unwillingness to allow the Red Cross to visit its camps.21 Other actions cannot be observed so easily by neutral parties; observing acts on the battlefield imposes the risk of death on observers, thus militaries are the most effective observers of the conduct of their own soldiers.

Inadvertent acts also create noise. During the Italian campaign of World War II, a German bombardment hit an American supply dump of chemical weapons and released a cloud of poison gas that drifted toward the German lines.22 This cloud could easily have been interpreted as an illegal chemical weapons attack by the Germans. The vagaries of battle can produce outcomes that can be interpreted as violations of the laws of war.

Practically, a legal system to limit violence during wartime must reflect

21 During World War II, the United States and Canada inferred that Japan was not living up to the 1929 Geneva Convention on POWs on the basis of Japanese interference with Red Cross inspections and reports on their POW camps. See Vance, supra note 16, at 186–88.

22 Jeffrey W. Legro, Cooperation under Fire: Anglo-German Restraint during World War II 200 (1995).

This content downloaded from �������������131.94.186.70 on Mon, 25 Nov 2019 19:14:54 UTC�������������

All use subject to https://about.jstor.org/terms

S54 the journal of legal studies

and accept the consequences of these three aspects of the laws of war. Ef- fective limitation on violence in the face of variation in state preferences in standards (or the absence of standards) requires that a standard be acknowl- edged in advance and that an answer to state deviations from that standard exists. This shared understanding must address what behavior is unacceptable and how states may respond to it to produce expectations for a successful legal system. The practical politics that arise from that shared understanding must deal with both the two-level problem and the difficulty of enforcing agreements in the face of noise. To understand these practical issues, we must understand how shared understandings and actors’ incentives reinforce or undermine one another. In short, we need to understand when a shared understanding embodies the common conjecture of an equilibrium.

V. The Design of the Laws of War

The arguments above explain how and why laws of war can operate successfully to limit violence. I now turn to the parallel question of which agreements can be formed in advance of war. I focus on four elements of the system of laws of war: their multilateral nature, their separation by issues into different treaties, the devolution of responsibility to each state for pun- ishing individual violations by its soldiers, and the strength of restrictions on violence. For each of these issues of design, I ask about the advantages or disadvantages of these features compared with alternative arrangements. My arguments about how these agreements operate in practice allow an analysis of the consequences of these alternative arrangements. Each of these alternative arrangements has attractive features that are outweighed by their disadvantages only when we consider how these arrangements would work in practice.

A. Multilateral Treaties

The laws of war are multilateral treaties that are formed far in advance of war rather than bilateral agreements that are formed at the beginning of a war. Bilateral agreements could be more efficient in the sense that warring states could reach a specific agreement that fit their own views of acceptable conduct more closely than a multilateral treaty negotiated before war could. Bilateral agreements, however, face concerns about how the laws of war could shift each side’s chance of winning if that pair of states were to find themselves at war. As stated in Section II, limiting the set of available battle strategies can shift the break points of both sides in favor of one side. Understandably, the side disadvantaged by a bilateral agreement might refuse to accept such an agreement, even if both sides would enforce the agreement once it was in place. The multilateral agreement has the advantage of being negotiated under a veil of ignorance of which wars will be fought in the

This content downloaded from �������������131.94.186.70 on Mon, 25 Nov 2019 19:14:54 UTC�������������

All use subject to https://about.jstor.org/terms

war and international politics S55

future.23 States might agree to a general standard in advance of war because they anticipate that they are advantaged in some possible wars even if they are disadvantaged in others. The multilateral standard also reduces negotiation costs compared with a full set of bilateral treaties negotiated in advance of war, even given the difficulties of multilateral negotiations.

B. Fire Walls

The laws of war have been negotiated through a number of separate treaties that address different issues in the limitation of violence rather than a single overarching agreement covering all issues.24 Compliance with an overarching agreement might be more likely because violation on one issue, such as POWs, could lead to retaliation on another issue, such as chemical weapons. The state suffering first use would choose the form of retaliation that would produce the greatest strategic benefit to itself. The linkage between all limits on war could make reciprocity more effective than separation into individual treaties.25 However, the laws of war do not rely on such linkage across issues. Instead, the laws of war are negotiated in separate treaties, creating fire walls between the issues. A state violation in one area does not allow the opposing side to violate some other treaty, and such generalized retaliation is rarely threatened.26 The most general sense of a fire wall is the idea that even aggressors as defined by the United Nations still receive the protection of the laws of war if they are ratifying powers. In the model, fire walls can be thought of as separate agreements on sets of banned battle strategies. A violation of one agreement opens up only the strategies banned in that agree- ment, not those banned by other agreements. Given that some limits will not be enforceable in all wars, fire walls allow some limits to persist even when others fail. So, for instance, the use of V-weapons to bombard London by

23 Not surprisingly, potential agreements pushed by some countries are rarely accepted by the particular opponents at whom those agreements appear to be targeted. For example, France and Germany were unwilling to accept British proposals to limit submarine warfare against merchant shipping during the 1920s. Legro, supra note 21, at 36–37.

24 It need not be that all issues would have to be settled in one treaty for a universal agreement on the laws of war. New treaties could be added to that agreement upon their coming into force.

25 Michael D. McGinnis, Issue Linkage and the Evolution of International Cooperation, 30 J. Conflict Resol. 141 (1986).

26 The closest example of an explicit breach of a fire wall was U.S. Secretary of State James Baker’s threat before the Gulf War to Iraq’s Foreign Minister Tariq Aziz that the United States would expand its war aims to include the overthrow of the regime in Baghdad if Iraq used chemical or biological weapons. See Lawrence Freedman & Efraim Karsh, The Gulf Conflict, 1990–1991, at 257 (1993). Violations on the battlefield of one of the protections of soldiers trying to surrender, of medical personnel and the wounded, and of flags of truce often trigger violations of the others, which effectively breach the fire walls between these protections. See Linderman, supra note 10, at 135, 137, & 141 for examples.

This content downloaded from �������������131.94.186.70 on Mon, 25 Nov 2019 19:14:54 UTC�������������

All use subject to https://about.jstor.org/terms

S56 the journal of legal studies

the Nazis did not lead to the use of chemical weapons by Great Britain, despite arguments by Churchill to use gas against German civilians.27

Noise suggests another reason for creating fire walls between treaties that cover different issues in the laws of war. Because issues vary in the level of noise present, separating enforcement of the treaties from one another de- creases the chance that a breakdown on one issue will spread to another. Issues that present a high level of noise are more likely to break down in practice even if only from misunderstandings. Even if the system fails on those issues, other treaties that cover issues with less noise could still be sustained. Because noise increases the chance that agreements will fail in practice, fire walls that are created by the separation of treaties help to preserve some order on the battlefield.

C. Devolution of Responsibility for Individual Violations

The laws of war typically devolve responsibility for individual violations to ratifying states to enforce on their own soldiers. Alternatively, other states or international bodies could enforce the laws of war against individual violators. The International Criminal Court is a first step in the direction of creating such an international body.

The two-level problem of policing individual violations of the treaties resembles the prevention of the escalation of ethnic violence that James D. Fearon and David Laitin model.28 They compare two systems of response to violent acts across groups: in the first system, members of each group respond to violent acts by a member of the other group by retaliating against any member of the other group whom they meet; in the second system, members of each group discipline their own violators, and retaliation across groups is not allowed. The former system has less deterrent effect than the latter be- cause the violators are unlikely to face personal retaliation. It also has an increased chance of spiral escalation because the individuals who suffer retaliation for the acts of other members of their group are likely to perceive a reprisal as a violation. Fearon and Laitin show further that devolution of discipline to groups is less likely to collapse in the face of difficulties in determining who is responsible for a violation than general cross-group retaliation.

General retaliation across armies on the battlefield happens sometimes, particularly when it becomes clear that one side is unwilling to live by an agreement or police its own soldiers to live up to that agreement. Combat during World War II on the eastern front in Europe and in the island war in the Pacific was unusually brutal because Russian and American soldiers

27 Legro, supra note 22, at 164–67. 28 James D. Fearon & David D. Laitin, Explaining Interethnic Cooperation, 90 Am. Polit.

Sci. Rev. 715 (1996).

This content downloaded from �������������131.94.186.70 on Mon, 25 Nov 2019 19:14:54 UTC�������������

All use subject to https://about.jstor.org/terms

war and international politics S57

learned about the Nazi and Japanese treatment of POWs and men who were trying to surrender and responded with no quarter.29 These general collapses of the laws of war are more common in cases such as the treatment of POWs and civilians, where noise from individual violations is more common.

As in the case of ethnic conflict, devolution to states of punishment of individual violations by their own soldiers can prevent spirals of retaliation even in the face of noise. An army is more likely than its opponent to be able to determine which of its soldiers have committed atrocities and then apprehend and punish the violators. After all, the survivors of an atrocity are often those who committed the act and their comrades in arms rather than the victims. Even when a treaty allows a state to punish individual violators from a state at war with it, such as the POW treaty, the opportunity is almost never pursued during wartime. The United States, for instance, did not carry out capital sentences on German POWs whom it tried and convicted of murdering fellow German POWs inside camps in the United States during World War II.30 The fear was that the Nazis would retaliate against an ex- ecution even of a prisoner justly convicted of a capital offense while in captivity. States at war cannot demonstrate to one another that such trials are fair, and so the possibility of states trying individual violators from other states during wartime is likely to lead to reciprocal spirals between states rather than deterrence of individual violations. The devolution of responsi- bility for individual violations to states has two advantages: militaries are better able to collect information on violations by their own personnel, and they have a greater ability to arrest, try, and punish their own individual violators than the opposing side or an international body. The International Criminal Court recognizes these advantages in its respect for a state’s policing of its own citizens when this is effective.

D. Strength of Restrictions

The strength of restrictions in a treaty poses a dilemma for both screening and in-group/out-group logic that I have argued are central to the successful operation of the laws of war. Both require that ratification not be universal, that states divide themselves into different groups through ratification. A treaty ratified by all states communicates no advance warning of which states intend to violate that treaty. In-group/out-group logic requires a clear cost to joining the in-group in order to prevent all states from joining. Attempts to make law acceptable to all states by lowering standards undermine both screening and in-group/out-group logic. Stronger standards lead to fewer states that are willing to accept a treaty, which enhances the screening effect

29 See Linderman, supra note 10; and Stephen G. Fritz, Frontsoldaten: The German Soldier in World War II (1995).

30 Arnold Krammer, Nazi Prisoners of War in America 169–73 (1979).

This content downloaded from �������������131.94.186.70 on Mon, 25 Nov 2019 19:14:54 UTC�������������

All use subject to https://about.jstor.org/terms

S58 the journal of legal studies

and produces a greater limitation on violence among the ratifying states. Stronger standards also produce a larger consequence for joining the in-group through ratification, which reduces the number of states that would ratify cynically to gain the benefits of in-group membership. On the other hand, stronger standards discourage some states from joining the system by ratifying a treaty, which raises the chance of wars where at least one side will not have accepted limits on its behavior during the war. Where the exact balance between the benefits and costs of stronger conditions falls is not easy to determine, hence the dilemma of the strength of restrictions.

VI. Conclusion: A Paradox

I have argued in this paper that the laws of war operate as the codification of a common conjecture within an equilibrium and explored the consequences of that argument. I close by noting how this view of international law poses a paradox. In game theory, a common conjecture must be present to use equilibrium analysis, regardless of the source of that shared belief. A common conjecture specifies all possibilities whether they can occur in equilibrium or not (up to uncertainties and private information in the game). The paradox is this: why is law in the form of treaties necessary if the parties already share a complete understanding of possible responses to all contingencies?

This question challenges the role of customary international law. Although custom and existing practice often contribute to the development of a common conjecture, law has a specific role to play in specifying how those customs should operate. Custom as a guide often lacks sufficient detail to allow warring parties to share an understanding of what conduct is acceptable and what consequences follow from unacceptable conduct. Formal negotiation can sharpen the understanding among states about custom; formal negotiation creates common knowledge about which states have accepted the formal standard and thus creates an opportunity for screening. Custom leaves the answers to these questions unclear.

Shared understandings about limiting violence during wartime have worked in the absence of formal international law. For example, the violence of sieges from the Middle Ages into the modern era was regulated by an informal agreement on the rights and responsibilities of both parties.31 Once the fortifications of a city were breached, the besiegers could call on the defenders to surrender the city. If they did, their lives and those of the residents of the city would be protected. If the defenders chose to continue resistance, the attackers were free to do as they wished with them and the city if they captured the city by storm. This shared understanding sufficed even in the absence of a formal treaty. This practice did lead to several

31 See Geoffrey Parker, Early Modern Europe, in The Laws of War: Constraints on Warfare in the Western World 40, 48–51 (Michael Howard, George J. Andreopoulus, & Mark R. Shulman eds. 1994), for a discussion of this practice in early modern Europe.

This content downloaded from �������������131.94.186.70 on Mon, 25 Nov 2019 19:14:54 UTC�������������

All use subject to https://about.jstor.org/terms

war and international politics S59

horrific sacks of cities, notably Magdeburg in 1631 and Drogheda in 1649, when the garrison chose to resist. It also reduced the chance of unnecessary resistance by reassuring the garrison that quarter would be extended if it surrendered after a breach was made. Understandably, questions arose as to how large a breach in the walls must be to justify a call for the garrison to surrender. Refinement and codification of informal shared understandings of appropriate limits on violence are central issues in treaties on the laws of war.

The concept of equilibrium in game theory provides no guidance about the paradox because it assumes a common conjecture among the players. There is nothing for a treaty to clarify about how the standard operates; it is simply a matter of choosing a standard. The screening argument developed in the paper embodies the game-theoretic idea of common conjecture under the conditions of uncertainty about others’ intentions. Each state has private information about how it will behave during wartime, and the treaty operates to screen out types of states that are unwilling to live with the common standard. A treaty exists solely as a public act of establishing the standard, and ratification is the public act of accepting that standard.

The screening argument here misses an important element of the laws of war in practice. Standards develop over time in response to wartime expe- rience. If a common conjecture exists among states, they would not have to develop treaty standards over time because they would already share expec- tations for all contingencies. There are two separate issues here: one concerns unanticipated events and consequences, and the other contradictory under- standings of appropriate conduct. Both result from incomplete understand- ings.32 These issues could be explored by relaxing the assumption of common conjecture to allow for discordant beliefs about what should happen off the equilibrium path. Such beliefs could exist in equilibrium if the anticipations of both sides about what would happen off the equilibrium path support equilibrium behavior. In plain English, neither side believes that acting out- side the range of anticipations is in its interest. Development of a standard through treaties then could help create common knowledge of that standard and its application.

This argument implies that greater legalization is always preferable. It helps to eliminate unanticipated situations and reactions. Despite this argu- ment, greater legalization may not be beneficial in all cases. When inter- national law seeks to establish neutral actors to judge individual cases, general principles of law could be superior to highly detailed treaties. The arbitrators of a dispute would then apply those principles as is common in domestic law in the interest of helping the parties arrive at a common understanding of responsibility and appropriate response in the dispute. The laws of war, on the other hand, seek to coordinate parties at war when deliberation on

32 Would these incomplete understandings be “uncommon conjectures”?

This content downloaded from �������������131.94.186.70 on Mon, 25 Nov 2019 19:14:54 UTC�������������

All use subject to https://about.jstor.org/terms

S60 the journal of legal studies

the appropriate response to violations is probably difficult at best. Greater legalization in the laws of war is not attractive when the candidate treaty tries to address issues for which an agreement is probably not enforceable. That stronger treaty could undermine a weaker system that succeeds in a more limited way. Successful shared understandings to limit violence must serve as the common conjecture of some equilibrium.

This content downloaded from �������������131.94.186.70 on Mon, 25 Nov 2019 19:14:54 UTC�������������

All use subject to https://about.jstor.org/terms