Rational Choice Theory and International Law: Insights and limitation Robert o. Keohane James b. duke professor of political science Duke University Author's note: The article appears in The Journal of Legal Studies, Volume XXXI(1), Pt. 2, January 2002 S307-S319. This volume is an attempt to explore the value of rational-choice analysis for the study of international law. As the editors suggest, the field of international law has been quite resistant to rational-choice analysis. It is important to recognize, however, that the enterprise of which this volume is an example is not new. Political scientists have used rational-choice analysis for two decades to understand international institutions, beginning notably with a special issue of international Organization in 1982. Some international legal scholars understood fairly quickly that"international institutions, "in the language of political science, were largely equivalent to "international law"as they understood it. Indeed, Anne-Marie Slaughter characterized institutionalist work on international regimes as"reinventing international law in rational-choice language Kenneth abbott was the pioneer among international legal scholars in introducing political scientists' arguments about institutions into the legal literature. More recently, another special issue of International Organization -this time an interdisciplinary venture involving international legal scholars -has been devoted to the subject of realization I These comments are highly indebted to the discussion at the conference on Rational Choice and International Law, University of Chicago Law School, April 27-29, 2001. I especially want to thank Lenneth Abbott, George Downs, John Ferejohn, Jack Goldsmith, Nannerl O Keohane, Eric Posner, Imke Risopp-Nickelson, and Duncan Snidal for comments on earlier drafts. 2 International Organization, volume 36, no. 2(Spring 1982), reprinted in Stephen DKrasner,ed International Regimes(Ithaca: Cornell University Press 1983). The papers in this volume that used rational-choice frameworks were Arthur A Stein, "Coordination and Collaboration: Regimes in an Anarchic World"(pp 115-140 in the Krasner volume)and Robert O Keohane, "The Demand for International Regimes"(pp. 141-171). The latter paper is the theoretical kernel of my 1984 book, After Hegemony: Cooperation and Discord in the world Political Economy(Princeton: Princeton University Press) Anne-Marie Slaughter Burley, "International Law and International Relations Theory: A Dual Agenda, " 87 American Journal of International Law 205(1993), p. 220. Kenneth w. Abbott, "Modern International Relations Theory: A Prospectus for Intermational Lawyer 14 Yale ournal of international Law 335 (1989) International Organization, volume 54, no. 3(summer 2000), reprinted as Judith L. Goldstein, Miles, Kahler, Robert O. Keohane, and Anne-Marie Slaughter, eds, Legalization and World Politics(Cambridge MIT Press, 2001). Kenneth w. Abbott and duncan Nidal rely heavily on rational-choice analysis in their ontribution to this volume; and the influence of the rationalist way of thinking pervades it. See Abbott and Snidal, "Hard and Soft Law in International Governance, "in Legalization, pp. 37-7
1 Rational Choice Theory and International Law: Insights and Limitations1 Robert O. Keohane James B. Duke Professor of Political Science Duke University Author’s note: The article appears in The Journal of Legal Studies, Volume XXXI (1), Pt. 2, January 2002: S307-S319. This volume is an attempt to explore the value of rational-choice analysis for the study of international law. As the editors suggest, the field of international law has been quite resistant to rational-choice analysis. It is important to recognize, however, that the enterprise of which this volume is an example is not new. Political scientists have used rational-choice analysis for two decades to understand international institutions, beginning notably with a special issue of International Organization in 1982.2 Some international legal scholars understood fairly quickly that “international institutions,” in the language of political science, were largely equivalent to “international law” as they understood it. Indeed, Anne-Marie Slaughter characterized institutionalist work on international regimes as “reinventing international law in rational-choice language.”3 Kenneth Abbott was the pioneer among international legal scholars in introducing political scientists’ arguments about institutions into the legal literature.4 More recently, another special issue of International Organization – this time an interdisciplinary venture involving international legal scholars -- has been devoted to the subject of legalization.5 1 These comments are highly indebted to the discussion at the conference on Rational Choice and International Law, University of Chicago Law School, April 27-29, 2001. I especially want to thank Kenneth Abbott, George Downs, John Ferejohn, Jack Goldsmith, Nannerl O. Keohane, Eric Posner, Imke Risopp-Nickelson, and Duncan Snidal for comments on earlier drafts. 2 International Organization, volume 36, no. 2 (Spring 1982), reprinted in Stephen D. Krasner, ed., International Regimes (Ithaca: Cornell University Press 1983). The papers in this volume that used rational-choice frameworks were Arthur A. Stein, “Coordination and Collaboration: Regimes in an Anarchic World” (pp. 115-140 in the Krasner volume) and Robert O. Keohane, “The Demand for International Regimes” (pp. 141-171). The latter paper is the theoretical kernel of my 1984 book, After Hegemony: Cooperation and Discord in the World Political Economy (Princeton: Princeton University Press). 3 Anne-Marie Slaughter Burley, “International Law and International Relations Theory: A Dual Agenda,” 87 American Journal of International Law 205 (1993), p. 220. 4 Kenneth W. Abbott, “Modern International Relations Theory: A Prospectus for International Lawyers,” 14 Yale Journal of International Law 335 (1989). 5 International Organization, volume 54, no. 3 (summer 2000), reprinted as Judith L. Goldstein, Miles, Kahler, Robert O. Keohane, and Anne-Marie Slaughter, eds., Legalization and World Politics (Cambridge: MIT Press, 2001). Kenneth W. Abbott and Duncan Snidal rely heavily on rational-choice analysis in their contribution to this volume; and the influence of the rationalist way of thinking pervades it. See Abbott and Snidal, “Hard and Soft Law in International Governance,” in Legalization, pp. 37-72
My own work has been part of this rationalist research program and I am highly committed to a modest form of rational-choice theory. Hence I wish to begin this commentary by indicating what that commitment means-and doesnt mean-to me Defining rationality involves very difficult issues of information. Rationality could be defined in such that the proposition of rational choice is irrefutable principle. If rationality simply means that a person does what he most prefers, a madman is perfectly rational. Perhaps he prefers to mutilate himself or to destroy those he loves On this basis, Hitler would have had to be judged rational-evil, but rational --when he ordered the Holocaust. But if all behavior is by definition rational, the concept of rationality"becomes meaningless. However, rationality need not be a meaningless concept, because a key condition for rationality involves the agents beliefs. Jon Elster offers a definition of what might be called"pure rationality, which emphasizes beliefs: An action, to be rational, must be the final result of three optimal decisions. First, it must be the best means of realizing an individual's desires, given his beliefs. Next, these beliefs must themselves be optimal given the information available to him. Finally, the person must collect an optimal amount of evidence -neither too much nor too little Of course, this definition of rationality turns it into an ideal type. No action is ever,in this definition, perfectly rational. But we can certainly make judgments about whether the behavior of particular agents approximates rationality or whether, like Hitler's, it flunks the informational tests. Hitler's beliefs about non-Aryans were wildly distorted, even given the information available to him. And he by no means searched analysis as inapplicable at least to some aspects of his behavior View rational-choice thoroughly for evidence. Hence we can judge him irrational and When I say that an actor behaves strategically, I do not mean that she behaves perfectly according to the rationalist ideal-type. To characterize behavior as strategic however, does have six implications: 1)that the agent makes choices in anticipation of the likely choices and reactions of others; 2)that she is concerned about the consequences of her actions; 3)that she orders her preferences in a broadly consistent fashion, so that her preferences are relatively stable across time and across issues; 4) that she selects means that she believes will achieve better rather than worse anticipated consequences in terms of her own ordered preferences; 5) that her beliefs are reasonable in view of available information; and 6) that she has searched for information that is relevant to the attainment of her objectives Strategic action is a matter of degree-we can act more or less strategically. Rationality is an ideal type, which is useful as a baseline for judging behavior as more or ss strategic. The assumption of rationality is also useful for creating formal models but is important to recognize that these are not models of actual behavior, but of ideal- typical behavior. Another way to express this point is to say that these models do not Jon Elster, Nuts and Bolts for the Social Sciences( Cambridge: 1989), p. 30
2 My own work has been part of this rationalist research program and I am highly committed to a modest form of rational-choice theory. Hence I wish to begin this commentary by indicating what that commitment means – and doesn’t mean – to me. Defining rationality involves very difficult issues of information. Rationality could be defined in such a way that the proposition of rational choice is irrefutable in principle. If rationality simply means that a person does what he most prefers, a madman is perfectly rational. Perhaps he prefers to mutiliate himself or to destroy those he loves. On this basis, Hitler would have had to be judged rational – evil, but rational -- when he ordered the Holocaust. But if all behavior is by definition rational, the concept of “rationality” becomes meaningless. However, rationality need not be a meaningless concept, because a key condition for rationality involves the agent’s beliefs. Jon Elster offers a definition of what might be called “pure rationality,” which emphasizes beliefs: “An action, to be rational, must be the final result of three optimal decisions. First, it must be the best means of realizing an individual’s desires, given his beliefs. Next, these beliefs must themselves be optimal, given the information available to him. Finally, the person must collect an optimal amount of evidence – neither too much nor too little.”6 Of course, this definition of rationality turns it into an ideal type. No action is ever, in this definition, perfectly rational. But we can certainly make judgments about whether the behavior of particular agents approximates rationality or whether, like Hitler’s, it flunks the informational tests. Hitler’s beliefs about non-Aryans were wildly distorted, even given the information available to him. And he by no means searched thoroughly for evidence. Hence we can judge him irrational and view rational-choice analysis as inapplicable at least to some aspects of his behavior. When I say that an actor behaves strategically, I do not mean that she behaves perfectly according to the rationalist ideal-type. To characterize behavior as strategic, however, does have six implications: 1) that the agent makes choices in anticipation of the likely choices and reactions of others; 2) that she is concerned about the consequences of her actions; 3) that she orders her preferences in a broadly consistent fashion, so that her preferences are relatively stable across time and across issues; 4) that she selects means that she believes will achieve better rather than worse anticipated consequences in terms of her own ordered preferences; 5) that her beliefs are reasonable in view of available information; and 6) that she has searched for information that is relevant to the attainment of her objectives. Strategic action is a matter of degree – we can act more or less strategically. Rationality is an ideal type, which is useful as a baseline for judging behavior as more or less strategic. The assumption of rationality is also useful for creating formal models – but is important to recognize that these are not models of actual behavior, but of idealtypical behavior. Another way to express this point is to say that these models do not 6 Jon Elster, Nuts and Bolts for the Social Sciences (Cambridge: 1989), p. 30
purport to be comprehensive, but rather focus on one central aspect of human behavior. The models may turn out to be good approximations of behavior under certain conditions, and therefore of great value. But they are not universally and unconditionally correc They generate hypotheses, not conclusions Furthermore, it is important not to equate rationality with materialistic self- interest, or to regard strategic action as motivated solely by materialistic self-interest. To do so would be to equate rational choice theory with a crude philosophical materialism. Yet there is no necessary connection at all. a rational altruist allocates her charitable donations so that the marginal value of the additional dollar, given to each of a variety of charities, is equalized. The rational environmentalist chooses strategies with the highest expected value of improving environmental outcomes. The rational anti-globalization demonstrator picks situations where the world 's media will be hungry for pictures of demonstrators, engulfed in waves of tear-gas, confronting troops in ugly gas masks. A Elster says, the rational-choice theorist"can say that the person acts so as to maximize utility, as long as we keep in mind that this is nothing but a convenient way of saying that option might be one that gives pleasure to others and none to himself ng act, his preferred he does what he most prefers. There is no implication of hedonism. In I believe that political actors behave strategically, according to what James March and Johan Olsen call a"logic of consequences, most of the time, perhaps almost all of the time. Hence rational-choice analysis is essential for studying political interactions and the institutions that both shape such interactions and are created by them. But members of the mass public often do not act strategically- and in the theory of rational choice, often should not. If one has negligible influence over an outcome, the investments in information and calculation necessary to determine one s optimal strategy will not themselves be optimal, so it may make more sense to operate on the basis of a rule of thumb. Voters, for instance, may be"rationally ignorant. " Likewise, citizens may decide to uphold broad norms, such as defending human rights, or protecting life, even if they do not find it worthwhile to think-through the consequences of doing so They could, for instance, have favored the arrest, in Great Britain, of former President inochet of Chile without being required, by reason or interest, to calculate whether that action would advance or retard the protection of human rights in developing countries Making such a calculation in a definitive way would be impossible even for specialists since it would depend not only on correctly modeling the various interlocked games involved, now and in the future, but upon the values of certain parameters of these games Given this understanding of rationality and strategic action, I ask: what do the papers in this volume tell us about the contributions that theories based on the assumption of rationality-"rational-choice theories"-make to the study of international law? I do 7I am indebted to duncan Nidal for this formulation Elster, Nuts and Bolts, pp. 23-24 James G. March and Johan P. Olsen, The Institutional Dynamics of International Political Orders. In Exploration and Contestation, pp 303-329 Shepsh ering work in this field was done in the late 1970s and the 1980s by Kenneth A Shepsle ce of "Institutional Equilibrium and Equilibrium Institutions, "in Political Science: The Scie Politics, ed. Herbert F. Weisberg(New York: Agathon), pp. 51-81
3 purport to be comprehensive, but rather focus on one central aspect of human behavior.7 The models may turn out to be good approximations of behavior under certain conditions, and therefore of great value. But they are not universally and unconditionally correct. They generate hypotheses, not conclusions. Furthermore, it is important not to equate rationality with materialistic selfinterest, or to regard strategic action as motivated solely by materialistic self-interest. To do so would be to equate rational choice theory with a crude philosophical materialism. Yet there is no necessary connection at all. A rational altruist allocates her charitable donations so that the marginal value of the additional dollar, given to each of a variety of charities, is equalized. The rational environmentalist chooses strategies with the highest expected value of improving environmental outcomes. The rational anti-globalization demonstrator picks situations where the world’s media will be hungry for pictures of demonstrators, engulfed in waves of tear-gas, confronting troops in ugly gas masks. As Elster says, the rational-choice theorist “can say that the person acts so as to maximize utility, as long as we keep in mind that this is nothing but a convenient way of saying that he does what he most prefers. There is no implication of hedonism. In fact, his preferred option might be one that gives pleasure to others and none to himself.”8 I believe that political actors behave strategically, according to what James March and Johan Olsen call a “logic of consequences,” most of the time, perhaps almost all of the time.9 Hence rational-choice analysis is essential for studying political interactions and the institutions that both shape such interactions and are created by them.10 But members of the mass public often do not act strategically – and in the theory of rational choice, often should not. If one has negligible influence over an outcome, the investments in information and calculation necessary to determine one’s optimal strategy will not themselves be optimal, so it may make more sense to operate on the basis of a “rule of thumb.” Voters, for instance, may be “rationally ignorant.” Likewise, citizens may decide to uphold broad norms, such as defending human rights, or protecting life, even if they do not find it worthwhile to think-through the consequences of doing so. They could, for instance, have favored the arrest, in Great Britain, of former President Pinochet of Chile without being required, by reason or interest, to calculate whether that action would advance or retard the protection of human rights in developing countries. Making such a calculation in a definitive way would be impossible even for specialists, since it would depend not only on correctly modeling the various interlocked games involved, now and in the future, but upon the values of certain parameters of these games. Given this understanding of rationality and strategic action, I ask: what do the papers in this volume tell us about the contributions that theories based on the assumption of rationality – “rational-choice theories” – make to the study of international law? I do 7 I am indebted to Duncan Snidal for this formulation. 8 Elster, Nuts and Bolts, pp. 23-24. 9 James G. March and Johan P. Olsen, “The Institutional Dynamics of International Political Orders.” In Exploration and Contestation, pp. 303-329. 10 Pioneering work in this field was done in the late 1970s and the 1980s by Kenneth A. Shepsle. See Shepsle, “Institutional Equilibrium and Equilibrium Institutions,” in Political Science: The Science of Politics, ed. Herbert F. Weisberg (New York: Agathon), pp. 51-81
not attempt a systematic evaluation of these highly varied papers, but rather use some of them to illustrate the points I want to make. In discussing the value of rational-choice theory, I will also refer to its limitations, of which it is particularly important for its practitioners to be aware My argument is that rational-choice theory is valuable in at least five ways: 1 )as a basis for skeptical interpretations of talk and action; 2)as a menu of causal mechanisms; 3) as an explanatory device for helping to resolve specific puzzles; 4 )as part-but only part-of an explanation of legal and political phenomena; and 5)as the basis for generating further puzzles for research. Different methods are associated with these different tasks Rational-Choice Skepticism ultimate skepticism, renouncing the possibility of stable knowledge. Rationalism can a In an age of doubt, skepticism is easy. In a sense, postmodernism constitutes th also be relentlessly skeptical, but in a way that is ultimately more constructive. By exposing the failures of superficial or inconsistent explanations, it can clear the way for positive theory, empirical testing, and a better understanding of social reality The paper in this volume by Jack Goldsmith and Eric Posner exemplifies the skepticism of rational-choice theory. Goldsmith and Posner quite elegantly demonstrate that moralistic and legalistic talk is not necessarily indicative of moralist and legalistic motivations by the speakers. Their method, quite appropriately, is conceptual rather than mpirical. Signalling theory indicates that cynically self-interested statesmen should sound as moralistic as truly moralistic statesmen in environments that reward such talk What Goldsmith and posner do not claim to establish is what the motivations of these statesmen in fact are. Since there are typically pooling equilibria, what the theory tells us is that we can't tell. The theory that leaders of some states are influenced by moral norms may under some conditions imply the same behavior as the theory that all leaders act on the basis of narrow self-interest Rational-choice theory is quite good at telling us what we don' t know. Another case in point is principal-agent theory, which makes the important methodological point that we should not attribute power to agents merely on the basis that they undertake important actions, without being contradicted by their principals. The theory of agent autonomy and that of perfect dominance by the principal are observationally equivalent It is important to know what we don't know. But we should remind ourselves that being aware of our ignorance is not equivalent to having made a valid inference about social reality II Jack Goldsmith and Eric A Posner. "Moral and Legal Rhetoric in International Relations: A Rational Choice Perspective, this volume, pp --- Mark A Pollack, "Learning from the Americanists(Again): Theory and method in the Study of Delegation, "West European Politics, January 20
4 not attempt a systematic evaluation of these highly varied papers, but rather use some of them to illustrate the points I want to make. In discussing the value of rational-choice theory, I will also refer to its limitations, of which it is particularly important for its practitioners to be aware. My argument is that rational-choice theory is valuable in at least five ways: 1) as a basis for skeptical interpretations of talk and action; 2) as a menu of causal mechanisms; 3) as an explanatory device for helping to resolve specific puzzles; 4) as part – but only part – of an explanation of legal and political phenomena; and 5) as the basis for generating further puzzles for research. Different methods are associated with these different tasks. Rational-Choice Skepticism In an age of doubt, skepticism is easy. In a sense, postmodernism constitutes the ultimate skepticism, renouncing the possibility of stable knowledge. Rationalism can also be relentlessly skeptical, but in a way that is ultimately more constructive. By exposing the failures of superficial or inconsistent explanations, it can clear the way for positive theory, empirical testing, and a better understanding of social reality. The paper in this volume by Jack Goldsmith and Eric Posner exemplifies the skepticism of rational-choice theory.11 Goldsmith and Posner quite elegantly demonstrate that moralistic and legalistic talk is not necessarily indicative of moralist and legalistic motivations by the speakers. Their method, quite appropriately, is conceptual rather than empirical. Signalling theory indicates that cynically self-interested statesmen should sound as moralistic as truly moralistic statesmen in environments that reward such talk. What Goldsmith and Posner do not claim to establish is what the motivations of these statesmen in fact are. Since there are typically pooling equilibria, what the theory tells us is that we can’t tell. The theory that leaders of some states are influenced by moral norms may under some conditions imply the same behavior as the theory that all leaders act on the basis of narrow self-interest. Rational-choice theory is quite good at telling us what we don’t know. Another case in point is principal-agent theory, which makes the important methodological point that we should not attribute power to agents merely on the basis that they undertake important actions, without being contradicted by their principals. The theory of agent autonomy and that of perfect dominance by the principal are observationally equivalent.12 It is important to know what we don’t know. But we should remind ourselves that being aware of our ignorance is not equivalent to having made a valid inference about social reality. 11 Jack Goldsmith and Eric A. Posner, “Moral and Legal Rhetoric in International Relations: A Rational Choice Perspective,” this volume, pp --- . 12 Mark A. Pollack, “Learning from the Americanists (Again): Theory and Method in the Study of Delegation,” West European Politics, January 2002
Rational-Choice Theory and Causal Mechanisms Any coherent social science explanation requires a causal mechanism. That i there must not merely be a correlation, but also an account of how one set of actions led to another. Causal mechanisms are embedded in ordinary narratives-we ask "why" all the time. When we ask, "why did someone do that? "we usually have in mind a potential set of explanations, each of which either focuses on the person's preferences and his choices of means to attain them, or on the ideas of appropriate behavior that he held Historians seek to infer causality in specific cases by showing that they have identified the crucial causal mechanisms at work in the situation, whose operation they can trace through time. Social scientists typically seek to find a number of instances of a phenomenon, or good tests of a causal mechanism, in order to arrive at generalizations, nd to decrease the uncertainty of their inferences Before causal mechanisms can be tested, they must be articulated and specified. Rational-choice theory is particularly good at this task, since interesting causal mechanisms can be developed on the basis of casual observation, followed by the methods of logical analysis and modeling. The problem is not to understand empirically how frequently patterns of actual behavior approximate the causal mechanism, or the conditions under which the mechanism operates, but simply how the mechanism The paper in this volume by George Downs and Michael Jones is a fine example of how rational-choice theory can be used to explore a causal mechanism, in this case that of reputation. Downs and Jones do not focus on a particular puzzle, but on the mechanism of reputation itself. The hypothetical mechanism under scrutiny is that commitments will be kept, even when it is costlier to abide by a particular provision than to break it, because of the potential reneger's concern about the consequences of renegin for its reputation. Downs and Jones take this mechanism -which is suggested by rational-choice theory -seriously. But they also point out the fallacy of believing that concern about reputation would guarantee compliance with all commitments In the Downs and Jones model, the costs and benefits of compliance with contractual obligations fluctuate stochastically, and it is inefficient or ineffective for actors to impose penalties for noncompliance above a certain threshold. Given these conditions, compliance with treaties is expected to be imperfect. There is also little ason to believe, they argue, that states will have equally strong incentives to comply with all of their agreements a states different agreements will be of different value to it, and the state may find that reneging on an agreement in one issue-area does not have a major impact on its reputation for compliance on other issues I3 On causal mechanisms see Elster, Nuts and Bolts, cited above. On problems of inference in social science, see Gary King, Robert O. Keohane, and Sidney Verba, Designing Social Inquiry: Scientific Inference in Qualitative Research(Princeton: Princeton University Press, 1994) 14 George W. Downs and Michael A. Jones, "Reputation, Compliance and International Law. "This
5 Rational-Choice Theory and Causal Mechanisms Any coherent social science explanation requires a causal mechanism. That is, there must not merely be a correlation, but also an account of how one set of actions led to another. Causal mechanisms are embedded in ordinary narratives – we ask “why” all the time. When we ask, “why did someone do that?” we usually have in mind a potential set of explanations, each of which either focuses on the person’s preferences and his choices of means to attain them, or on the ideas of appropriate behavior that he held. Historians seek to infer causality in specific cases by showing that they have identified the crucial causal mechanisms at work in the situation, whose operation they can trace through time. Social scientists typically seek to find a number of instances of a phenomenon, or good tests of a causal mechanism, in order to arrive at generalizations, and to decrease the uncertainty of their inferences.13 Before causal mechanisms can be tested, they must be articulated and specified. Rational-choice theory is particularly good at this task, since interesting causal mechanisms can be developed on the basis of casual observation, followed by the methods of logical analysis and modeling. The problem is not to understand empirically how frequently patterns of actual behavior approximate the causal mechanism, or the conditions under which the mechanism operates, but simply how the mechanism operates. The paper in this volume by George Downs and Michael Jones is a fine example of how rational-choice theory can be used to explore a causal mechanism, in this case that of reputation.14 Downs and Jones do not focus on a particular puzzle, but on the mechanism of reputation itself. The hypothetical mechanism under scrutiny is that commitments will be kept, even when it is costlier to abide by a particular provision than to break it, because of the potential reneger’s concern about the consequences of reneging for its reputation. Downs and Jones take this mechanism – which is suggested by rational-choice theory – seriously. But they also point out the fallacy of believing that concern about reputation would guarantee compliance with all commitments. In the Downs and Jones model, the costs and benefits of compliance with contractual obligations fluctuate stochastically, and it is inefficient or ineffective for actors to impose penalties for noncompliance above a certain threshold. Given these conditions, compliance with treaties is expected to be imperfect. There is also little reason to believe, they argue, that states will have equally strong incentives to comply with all of their agreements. A state’s different agreements will be of different value to it, and the state may find that reneging on an agreement in one issue-area does not have a major impact on its reputation for compliance on other issues. 13 On causal mechanisms see Elster, Nuts and Bolts, cited above. On problems of inference in social science, see Gary King, Robert O. Keohane, and Sidney Verba, Designing Social Inquiry: Scientific Inference in Qualitative Research (Princeton: Princeton University Press, 1994). 14 George W. Downs and Michael A. Jones, “Reputation, Compliance and International Law.” This volume, pp. ---
Intuitions based on rational-choice theory help Downs and Jones to illumi complexity of reputational mechanisms. In particular, their analysis questions o ate the simplifying assumption that state actors are homogeneous, which was used for the sake of convenience in some earlier rational-choice models. But it also suggests the hypothesis that, contrary to speculation in much of the literature, reputational constraints will favor stronger rather than weaker actors The analysis of Downs and Jones indicates the value of articulating causal mechanisms. Rational-choice deductions that assume actor homogeneity are shown to be problematic-therefore, to the empirical investigator, probably not worth testing. One should get the causal mechanism right, in principle, before one sees whether it applies in the real world. The analysis that they offer suggests hypotheses about the segmentation of reputation and certain advantages of the strong that may have been insufficiently recognized. Several other causal mechanisms, suggested by rational-choice theory, are prominent in the literature, in this volume and elsewhere. Reciprocity in an impersonal and extensive environment such as that of world politics rests heavily on rational calculations of long-term mutual advantage. Signalling, cheap-talk and other means of screening good from bad types of players, may help to explain certain forms of rhetoric, ction, and institutions in world politics, including legalized talk, legalized forms of behavior,and legalized institutions, as Goldsmith and Posner explain in their paper. 6 Institutions, including legal institutions, may be invented in order to solve collective action problems by reducing transaction costs, providing information, and increasing the credibility of commitments. Legal institutions in particular may be valuable in dealing with problems of incomplete contracting. The menu of causal mechanisms identified by rational-choice theory is rich and tasty As we will see below, however, identifying mechanisms is not equivalent to successful explanation. It remains to theorize the conditions under which different mechanisms come into play. This is the really difficult task, which has not been solved. At present, we can more readily describe these mechanisms, and how they work, after the fact than we can specify the conditions under which one or more mechanism will operate. ex ante IS Robert O. Keohane, "Reciprocity in International Relations. "International Organization, volume 40, no. I(winter 1986): 1-27. 16 Goldsmith and Posner, this volume, pp 17 Keohane, After Hegemony, cited, chs. 5 and 6; Paul Milgrom, Robert North and Barry Weingast,"Th Role of Institutions in the Revival of Trade: the Medieval Law Merchant, Private Judges, and Champagne Fairs. Economics and Politics 1: 1-23. See also Frieder Roessler, Warren F. Schwartz and Alan O Sykes "The Economic Structure of Renegotiation and Dispute Resolution in the WTO/GATT System, "and Andrew Guzman and Beth A Simmons, "To Settle or Empanel? A Transaction Cost Approach to the WTO's Dispute Settlement Process, " both in this volume. ee abbott and Snidal. "Hard and Soft Law in International Governance. " "cited above. fn 5. and Roessler, Schwartz and Sykes, this volume
6 Intuitions based on rational-choice theory help Downs and Jones to illuminate the complexity of reputational mechanisms. In particular, their analysis questions the simplifying assumption that state actors are homogeneous, which was used for the sake of convenience in some earlier rational-choice models. But it also suggests the hypothesis that, contrary to speculation in much of the literature, reputational constraints will favor stronger rather than weaker actors. The analysis of Downs and Jones indicates the value of articulating causal mechanisms. Rational-choice deductions that assume actor homogeneity are shown to be problematic – therefore, to the empirical investigator, probably not worth testing. One should get the causal mechanism right, in principle, before one sees whether it applies in the real world. The analysis that they offer suggests hypotheses about the segmentation of reputation and certain advantages of the strong that may have been insufficiently recognized. Several other causal mechanisms, suggested by rational-choice theory, are prominent in the literature, in this volume and elsewhere. Reciprocity in an impersonal and extensive environment such as that of world politics rests heavily on rational calculations of long-term mutual advantage.15 Signalling, cheap-talk and other means of screening good from bad types of players, may help to explain certain forms of rhetoric, action, and institutions in world politics, including legalized talk, legalized forms of behavior, and legalized institutions, as Goldsmith and Posner explain in their paper.16 Institutions, including legal institutions, may be invented in order to solve collective action problems by reducing transaction costs, providing information, and increasing the credibility of commitments.17 Legal institutions in particular may be valuable in dealing with problems of incomplete contracting.18 The menu of causal mechanisms identified by rational-choice theory is rich and tasty. As we will see below, however, identifying mechanisms is not equivalent to successful explanation. It remains to theorize the conditions under which different mechanisms come into play. This is the really difficult task, which has not been solved. At present, we can more readily describe these mechanisms, and how they work, after the fact than we can specify the conditions under which one or more mechanism will operate, ex ante. 15 Robert O. Keohane, “Reciprocity in International Relations.” International Organization, volume 40, no. 1 (winter 1986): 1-27. 16 Goldsmith and Posner, this volume, pp. ---. 17 Keohane, After Hegemony, cited, chs. 5 and 6; Paul Milgrom, Robert North and Barry Weingast, “Th Role of Institutions in the Revival of Trade: the Medieval Law Merchant, Private Judges, and Champagne Fairs.” Economics and Politics 1: 1-23. See also Frieder Roessler, Warren F.Schwartz and Alan O.Sykes, “The Economic Structure of Renegotiation and Dispute Resolution in the WTO/GATT System,” and Andrew Guzman and Beth A. Simmons, “To Settle or Empanel? A Transaction Cost Approach to the WTO’s Dispute Settlement Process,” both in this volume. 18 See Abbott and Snidal, “Hard and Soft Law in International Governance,” cited above, fn. 5, and Roessler, Schwartz and Sykes, this volume.
Rational-Choice Theory as a Way of solving Specific Puzzles Not all of the papers in this volume express skepticism about certain interpretations of a general class of phenomena(such as moralistic talk), or sketch out causal mechanisms whose implications could apply to a wide variety of actions or institutions. Instead, some of the papers focus on very specific empirical puzzles, and explain specific institutional arrangements in the World Trade Organization. These o seek to use rational-choice theory to solve them. Roessler, Schwartz and Sykes seek to arrangements enable states to suffer limited retaliation rather than to change their laws to comply with rulings of the Dispute Settlement System; furthermore, they delay the imposition of sanctions and limit their permissible extent. Guided by transaction cost theory, Guzman and Simmons explore the conditions under which states settle disputes without recourse to a dispute settlement panel Solving specific puzzles requires the analyst to engage in some empirical work either anecdotally, as in the paper by roessler and his colleagues, or with systematic data, variables in the causal mechanism and to assess, in one way or another, the . explanatory as in the Guzman -Simmons paper. The research design is to identify the ke impact on the chosen dependent variable of these explanatory variables Often, as in the paper by roessler et al., such analyses rely heavily on general descriptions of patterns of behavior and of institutions, and no alternative theories are considered. Underlying the logic of such an analysis is a functional theory, in which(in the rational-choice version )anticipated effects explain causes. Such a theory is not in principle invalid, but in the absence of independent confirming evidence, it is subject to the post hoc ergo propter hoc fallacy. Another way of stating this problem is to emphasize the Folk Theorem of game theory: that strict game-theoretic analysis, without restrictive assumptions, typically predicts multiple equilibria, sometimes involving cyclin between different states of the world vang. The point is that, ex post, behavior can be interpreted, and modeled, in a large assumptions about common knowledge and constraints introduced to produce unique p of ways. The very plasticity of rational choice modeling-dependent, as it is, on equilibria-facilitates a variety of possible interpretations. So the demonstration that observed behavior could be a strategic equilibrium of an extensive-form game does not demonstrate that it is. I am reminded of Tweedledee's statement in Through the Looking Glass "lf it was so, it might be: and if it were so, it would be; but as it isnt. it aint. That's logic G. A Cohen, Karl Marx's Theory of History: A Defense(Princeton: Princeton University Press, 1978) For a fuller discussion of the virtues and pitfalls of functional theory, see Keohane, After Hegemony (supra, fn. 2) Lewis Carroll, Through the Looking Glass, chapter 4 7
7 Rational-Choice Theory as a Way of Solving Specific Puzzles Not all of the papers in this volume express skepticism about certain interpretations of a general class of phenomena (such as moralistic talk), or sketch out causal mechanisms whose implications could apply to a wide variety of actions or institutions. Instead, some of the papers focus on very specific empirical puzzles, and seek to use rational-choice theory to solve them. Roessler, Schwartz and Sykes seek to explain specific institutional arrangements in the World Trade Organization. These arrangements enable states to suffer limited retaliation rather than to change their laws to comply with rulings of the Dispute Settlement System; furthermore, they delay the imposition of sanctions and limit their permissible extent. Guided by transaction cost theory, Guzman and Simmons explore the conditions under which states settle disputes without recourse to a dispute settlement panel. Solving specific puzzles requires the analyst to engage in some empirical work, either anecdotally, as in the paper by Roessler and his colleagues, or with systematic data, as in the Guzman-Simmons paper. The research design is to identify the key explanatory variables in the causal mechanism and to assess, in one way or another, the impact on the chosen dependent variable of these explanatory variables. Often, as in the paper by Roessler et al., such analyses rely heavily on general descriptions of patterns of behavior and of institutions, and no alternative theories are considered. Underlying the logic of such an analysis is a functional theory, in which (in the rational-choice version) anticipated effects explain causes.19 Such a theory is not in principle invalid, but in the absence of independent confirming evidence, it is subject to the post hoc ergo propter hoc fallacy.20 Another way of stating this problem is to emphasize the Folk Theorem of game theory: that strict game-theoretic analysis, without restrictive assumptions, typically predicts multiple equilibria, sometimes involving cycling between different states of the world. The point is that, ex post, behavior can be interpreted, and modeled, in a large variety of ways. The very plasticity of rational choice modeling – dependent, as it is, on assumptions about common knowledge and constraints introduced to produce unique equilibria – facilitates a variety of possible interpretations. So the demonstration that observed behavior could be a strategic equilibrium of an extensive-form game does not demonstrate that it is. I am reminded of Tweedledee’s statement in Through the Looking Glass: “If it was so, it might be; and if it were so, it would be; but as it isn’t, it ain’t. That’s logic.”21 19 G. A. Cohen, Karl Marx’s Theory of History: A Defense (Princeton: Princeton University Press, 1978): 278. 20 For a fuller discussion of the virtues and pitfalls of functional theory, see Keohane, After Hegemony (supra, fn. 2). 21 Lewis Carroll, Through the Looking Glass, chapter 4
To avoid telling what Jon Elster calls"just-so stories, the investigator needs to specify the observable implications of her functional theory, not just for already-observed outcomes but for actions that are part of the process that leads to such outcomes. She should also indicate what alternative theories would predict bout both outcomes and processes, and assess them against her functional rational-choice theory. Rational-choice theory can be an enlightening and sometimes essential part of such an explanation; but if an empirical puzzle is being addressed, rational-choice theory is not a substitute for a good empirical research design Rational-Choice Theory as a Component of an explanation Following these remarks, it will come as no surprise to the reader that I prefer rational-choice theories that are embedded in a rich contextual and historical analysis, and which take into account both strategic calculations and other relevant conditions. Recent work on"analytic narratives"has attempted to combine historical and contextual understanding with rational-choice analysis, in ways that are certainly subject to criticism but that point, in my view, in the right direction. Rational-choice theory can provide insights into events that are too complex fully to model and too idiosyncratic to subject to statistical analysis. It can therefore suggest why certain institutional arrangements may have been considered, and how changes in external conditions could induce people to alter these arrangements over time Kenneth Abbott and Duncan Snidal. Without passing inidgmlile l neir ol In this volume, the paper that approximates this ideal best is the one terpretation of their particular case study (about which I have little knowledge) I express admiration of the way in which they show that values and interests were intertwined. " Value activists, such as those of Transparency International, were active on these issues. They created coalitions urging tougher state policies against corruption-coalitions that generated substantial public support, as well as allies within governments and international organizations. These coalitions created legislation, such as the US Foreign Corrupt Practices Act, adopted in 1977. Under the Clinton Administration considerations both of values and interest led officials in the State Department to give priority to corruption issues The existence of value-oriented publics in Western Europe helped them to circumvent resistance by European governments and the interests they represented. Jon Elster, "Rational Choice History: A Case of Excessive Ambition, " American Political Science 2Robert H. Bates, Avner Greif, Margaret Levi, Jean-Laurent Rosenthal, and Barry R.Weingast,Analytic Narratives. Princeton: Princeton University Press, 1998. For a harsh but telling set of criticisms, see Jan Elster, "Rational Choice History, cited above 24 Kenneth W. Abbott and Duncan Snidal "Values and Interests: International Legalization in the fight against Corruption. "This volum
8 To avoid telling what Jon Elster calls “just-so stories,”22 the investigator needs to specify the observable implications of her functional theory, not just for already-observed outcomes but for actions that are part of the process that leads to such outcomes. She should also indicate what alternative theories would predict about both outcomes and processes, and assess them against her functional rational-choice theory. Rational-choice theory can be an enlightening and sometimes essential part of such an explanation; but if an empirical puzzle is being addressed, rational-choice theory is not a substitute for a good empirical research design. Rational-Choice Theory as a Component of an Explanation Following these remarks, it will come as no surprise to the reader that I prefer rational-choice theories that are embedded in a rich contextual and historical analysis, and which take into account both strategic calculations and other relevant conditions. Recent work on “analytic narratives” has attempted to combine historical and contextual understanding with rational-choice analysis, in ways that are certainly subject to criticism but that point, in my view, in the right direction. Rational-choice theory can provide insights into events that are too complex fully to model and too idiosyncratic to subject to statistical analysis. It can therefore suggest why certain institutional arrangements may have been considered, and how changes in external conditions could induce people to alter these arrangements over time.23 In this volume, the paper that approximates this ideal best is the one by Kenneth Abbott and Duncan Snidal.24 Without passing judgment on their interpretation of their particular case study (about which I have little knowledge), I express admiration of the way in which they show that values and interests were intertwined. “Value activists,” such as those of Transparency International, were active on these issues. They created coalitions urging tougher state policies against corruption – coalitions that generated substantial public support, as well as allies within governments and international organizations. These coalitions created legislation, such as the US Foreign Corrupt Practices Act, adopted in 1977. Under the Clinton Administration considerations both of values and interest led officials in the State Department to give priority to corruption issues. The existence of value-oriented publics in Western Europe helped them to circumvent resistance by European governments and the interests they represented. 22 Jon Elster, “Rational Choice History: A Case of Excessive Ambition,” American Political Science Review 94-3 (September 2000): 685-695. 23 Robert H. Bates, Avner Greif, Margaret Levi, Jean-Laurent Rosenthal, and Barry R. Weingast, Analytic Narratives. Princeton: Princeton University Press, 1998. For a harsh but telling set of criticisms, see Jan Elster, “Rational Choice History,” cited above. 24 Kenneth W. Abbott and Duncan Snidal, “Values and Interests: International Legalization in the Fight against Corruption.” This volume
The behavior of the value activists cannot be explained well on the basis of conventional economic conceptions of self-interest, such as income maximization Of course, it would be surprising if self-interest, perhaps in terms of desire for reputation or even notoriety, were not involved. Yet a complete explanation of their behavior on the basis of self-interest would require a broadening of the concept of self-interest to the point at which distinctions among motivations would be lost A scholar committed to rational-choice analysis, however, would have a response: that in moving too quickly from interests to values the investigator may overlook interest-based explanations, and take actors self-presentations too much at face value. Although the insistence of some scholars on seeking to explain al behavior as motivated by narrowly-defined self-interest overlooks whole dimensions of human thought and action it does have the advantage of forcing the investigator to think hard -and often ingeniously -about how interests could explain puzzling behavior. 2 For a discussion of "availability cascades, "often involving a combination of moralism, search for information, and concern about reputation, see Timur Kuran and Cass R. Sunstein, "Availability Cascades and Risk Regulation, "51 Stanford Law Review 387(1999). I am indebted to George Downs for this I am indebted to Eric Posner, in comments on an earlier version of this paper, for this point
9 The behavior of the value activists cannot be explained well on the basis of conventional economic conceptions of self-interest, such as income maximization. Of course, it would be surprising if self-interest, perhaps in terms of desire for reputation or even notoriety, were not involved.25 Yet a complete explanation of their behavior on the basis of self-interest would require a broadening of the concept of self-interest to the point at which distinctions among motivations would be lost. A scholar committed to rational-choice analysis, however, would have a response: that in moving too quickly from interests to values the investigator may overlook interest-based explanations, and take actors’ self-presentations too much at face value.26 Although the insistence of some scholars on seeking to explain all behavior as motivated by narrowly-defined self-interest overlooks whole dimensions of human thought and action, it does have the advantage of forcing the investigator to think hard – and often ingeniously – about how interests could explain puzzling behavior. 25 For a discussion of “availability cascades,” often involving a combination of moralism, search for information, and concern about reputation, see Timur Kuran and Cass R. Sunstein, “Availability Cascades and Risk Regulation,” 51 Stanford Law Review 387 (1999). I am indebted to George Downs for this reference. 26 I am indebted to Eric Posner, in comments on an earlier version of this paper, for this point
When both strategy and beliefs are involved, the interaction between them can be explored, as in James Morrows paper. Morrow(p. 24)discusses the following paradox: "Why is law in the form of treaties necessary if the parties already share complete understanding of all possible responses to all contingencies? He points out that"the concept of equilibrium in game theory provides no guidance about the paradox because it assumes a common conjecture among the players. " As expressed by Thomas Schelling, the issue is how"focal points"are arrived at around which expectations converge; in the language of David Lewis, how do "conventions"come about? In th language of social construction, game theory does not tell us how understandings of reality are socially constructed -for example, why certain engraved pieces of green paper can be exchanged for valuable goods and services, while others cannot. The common knowledge produced by social processes provides a point of complementarity between constructivist thinking, which seeks to explain common knowledge or common conjectures, and game theory, which relies on it to solve games of strategic interaction Rational-Choice Theory as Puzzle-Generator At this point, the reader may recall the perhaps-apocryphal story told about the writer Gertrude Stein, on her deathbed. She asked her long-time friend, Alice B. Toklas what is the answer? ""Receiving no reply, Stein then said, "Well, then, what is the question? "In my judgment, rational-choice theory is unexcelled at asking questions. Its relentless skepticism is valuable, undermining conventional answers, critiquing analytical complacency, and demanding deepening of explanations, asking that exogenous factor causal mechanisms generates many research questions, not about the existence any be made endogenous, consistent with the premises of the theory. Its identification precise character of these mechanisms but about the conditions under which they operate and to what effect. Often opposite mechanisms are at play in a given situation, generating different expectations about behavior. Hence in the end, many of these seems worthwhile to list a few of them, even if answers remain elusive, uestions. it o questions will only have empirical answers, to be arrived at through well-designed ar painstaking research. However, since this volume is rich in promisin 27 James D. Morrow, "The Laws of War, Common Conjectures, and Legal Systems in International Politics, "this volume, pp. Thomas C Schelling, The Strategy of Conflict( Cambridge: Harvard University Press, 1960); David K Lewis, Comvenmtion: A Philosophical Study( Cambridge: Harvard University Press, 1969) The example is John R Searles. The Construction of Social Reality(New York: Free Press, 1995). On social constructivism and international relations, see Alexander Wendt, Social Theory of international Politics( Cambridge: Cambridge University Press, 1999). On this point see Peter J Katzenstein, RobertO. Keohane and Stephen Krasner, "International Organization and the Study of World Politics, "in Katzenstein et al., Exploration and Contestation, cited, fn.7,pp.40-42. sI The most recent special issue of International Organization(volume 55, no 4; Autumn 2001)is on"The Rational Design of International Institutions"(edited by Barbara Koremenos, Charles Lipson, and Duncan nidal). It raises a number of interesting questions, generated by rational-choice theory, about institutional design. Particularly novel is its attention to flexibility, defined (p. 773)as how institutional rules accommodate new circumstances; that institutional design helps actors deal with noise in the system(p. 1061)
10 When both strategy and beliefs are involved, the interaction between them can be explored, as in James Morrow’s paper.27 Morrow (p. 24) discusses the following paradox: “Why is law in the form of treaties necessary if the parties already share a complete understanding of all possible responses to all contingencies?” He points out that “the concept of equilibrium in game theory provides no guidance about the paradox because it assumes a common conjecture among the players.” As expressed by Thomas Schelling, the issue is how “focal points” are arrived at around which expectations converge; in the language of David Lewis, how do “conventions” come about?28 In the language of social construction, game theory does not tell us how understandings of reality are socially constructed – for example, why certain engraved pieces of green paper can be exchanged for valuable goods and services, while others cannot.29 The common knowledge produced by social processes provides a point of complementarity between constructivist thinking, which seeks to explain common knowledge or common conjectures, and game theory, which relies on it to solve games of strategic interaction.30 Rational-Choice Theory as Puzzle-Generator At this point, the reader may recall the perhaps-apocryphal story told about the writer Gertrude Stein, on her deathbed. She asked her long-time friend, Alice B. Toklas, “what is the answer?” Receiving no reply, Stein then said, “Well, then, what is the question?” In my judgment, rational-choice theory is unexcelled at asking questions. Its relentless skepticism is valuable, undermining conventional answers, critiquing analytical complacency, and demanding deepening of explanations, asking that exogenous factors be made endogenous, consistent with the premises of the theory. Its identification of causal mechanisms generates many research questions, not about the existence and precise character of these mechanisms but about the conditions under which they operate, and to what effect. Often opposite mechanisms are at play in a given situation, generating different expectations about behavior. Hence in the end, many of these questions will only have empirical answers, to be arrived at through well-designed and painstaking research. However, since this volume is rich in promising questions, it seems worthwhile to list a few of them, even if answers remain elusive.31 27 James D. Morrow, “The Laws of War, Common Conjectures, and Legal Systems in International Politics,” this volume, pp. -- . 28 Thomas C. Schelling, The Strategy of Conflict (Cambridge: Harvard University Press, 1960); David K. Lewis, Convention: A Philosophical Study (Cambridge: Harvard University Press, 1969). 29 The example is John R. Searle’s. The Construction of Social Reality (New York: Free Press, 1995). On social constructivism and international relations, see Alexander Wendt, Social Theory of International Politics (Cambridge: Cambridge University Press, 1999). 30 On this point see Peter J. Katzenstein, Robert O. Keohane and Stephen Krasner, “International Organization and the Study of World Politics,” in Katzenstein et al., Exploration and Contestation, cited, fn. 7, pp. 40-42. 31 The most recent special issue of International Organization (volume 55, no. 4; Autumn 2001) is on “The Rational Design of International Institutions” (edited by Barbara Koremenos, Charles Lipson, and Duncan Snidal). It raises a number of interesting questions, generated by rational-choice theory, about institutional design. Particularly novel is its attention to flexibility, defined (p. 773) as how institutional rules accommodate new circumstances; that is, how institutional design helps actors deal with noise in the system (p. 1061)