Worldviews in Foreign Policy: Realism, Liberalism, and External Conflict PDF

Political Psychology, Vol. 24, No. 3, 2003 Worldviews in Foreign Policy: Realism, Liberalism, and External Conflict Thomas S. Mowle United States Air Force Academy International relations studies have been unable to determine whether realist or liberal theories better fit state behavior in various...

Political Psychology, Vol. 24, No. 3, 2003 Worldviews in Foreign Policy: Realism, Liberalism, and External Conflict Thomas S. Mowle United States Air Force Academy International relations studies have been unable to determine whether realist or liberal theories better fit state behavior in various situations, possibly because these studies have attributed motive and action to the states rather than to the decision-makers within them. This article develops a new, more direct approach to resolving this problem. Hypotheses were tested regarding conditions under which decision-makers are likely to articulate a problem representation consistent with liberal or realist elements of a worldview. This was done by content analysis of statements about 36 foreign conflicts by the governments of three “bystander” nations—the United States, Canada, and India—over a 16-year period. The findings indicate that systemic and situational factors are far more important than domestic factors. States tend to represent wars in congruence with liberalism primarily when their security is already assured by another power or when the conflict does not involve allies, rivals, or fellow democracies. Thus, most of the expectations of realism are supported at the psychological level. KEY WORDS: problem representation, foreign policy decision-making, worldview, content analy- sis, United States foreign policy, international relations theory Liberal and realist theories of international behavior present quite different visions of how states interact with one another. Although individual scholars differ in their presentation of these paradigms, those in each school tend to share some basic principles. For example, realists generally suggest that interstate coopera- tion is severely limited by each state’s need to guarantee its own security in a global condition of anarchy, whereas liberals suggest that cooperation can be made more tenable through formal or informal institutions. These positions have been debated over the years—from Carr (1939) through Grieco (1988, 1990, 1993) and Keohane (1993)—but political scientists have so far been unable to show that either of these understandings of the world better explains how states actually behave. 561 0162-895X © 2003 International Society of Political Psychology Published by Blackwell Publishing. Inc., 350 Main Street, Malden, MA 02148, USA, and 9600 Garsington Road, Oxford, OX4 2DQ 562 Mowle One reason it has been difficult to choose between the two explanations is that scholars differ when interpreting the evidence: Where one scholar sees liberal institutionalism, another sees self-interest. At the broadest level of policy, Layne and Schwarz (1993, p. 5) described U.S. foreign policy since the 1950s as “liberal internationalist,” whereas Mearsheimer (1995, p. 5) contrasted President Clinton’s “neo-Wilsonism” with Cold War “balance of power politics.” At a much narrower level of policy, Grieco (1993, p. 327) and Keohane (1993, p. 280) both examined the same states’ opposition to a steel anti-dumping agreement. While Grieco wrote that the opposition showed realist concern for relative gains, Keohane wrote that the opposition showed liberal concern for absolute gains. As Keohane admitted (1993, p. 279), the problem is that “two parties that are indifferent to one another’s welfare will behave, at the margin, as if they care about relative gains.” Keohane is right: Liberalism and realism are as much qualities of motive as they are of action. A comparison of the two approaches must somehow pry open the intent behind the action. In other words, it must shift the level of analysis from the state—which has neither intent nor independent action—to the individuals within the state who direct purposive action, as suggested by Byman and Pollack (2001). This article uses problem representations and worldviews to break through the ambiguity often found when one tries to interpret large-scale state actions. These questions are examined in the context of a common, important, and under- studied situation: A state is faced with an “external conflict”—a war or other mil- itary action to which it is not initially a party (but in which it may later intervene). This article’s findings are generally consistent with the systemic and situational factors suggested by Waltz (1979), Grieco (1988), and Maoz and Russett (1993). In particular, it appears that decision-makers tend to express a problem represen- tation consistent with a realist worldview more often in situations when the exter- nal conflict is objectively more important to that state. Worldviews and Problem Representations A state’s behavior is not reflexive; rather, it flows from the way its foreign policy decision-makers understand what is happening. For example, members of the Reagan administration in the mid-1980s believed that cooperation with the Soviets would produce Soviet assertiveness, not further cooperation (Jervis, 1988, p. 326). Others could (and did) define the situation and the appropriate U.S. strat- egy differently. These assumptions—which include images of other actors in the world, causal beliefs about how they interact with one another, and prescriptions about appropriate courses of action—constitute a “worldview” (Barber, 1993, p. 131; Cottam & McCoy, 1998, p. 117; Doyle, 1997, p. 17; Young, 1998, p. 215). The worldview influences the way individuals interact with reality: “Beliefs set up expectations, and when an event occurs, we are likely to interpret the event in relation to our expectations” (Voss & Dorsey, 1992, p. 11). Related research includes Holsti’s (1962) and Jervis’ (1976) work with perception, Brecher’s Worldviews in Foreign Policy 563 (1972) attitudinal prism, and Wendt’s (1992, p. 397) discussion of the meaning of anarchy. The worldview is not, unfortunately, directly observable. One approach to describing an individual worldview in detail is to follow in the footsteps of Axelrod’s (1977) cognitive mapping or George’s (1979) operational codes, as has been done by Dille (2000), Dille and Young (2000), Schafer and Crichlow (2000), and Walker (1995). This approach is similar to Hermann’s (1988) work on Hafez al-Assad. This method is difficult to apply, however, where many actors are rel- evant in a state. Furthermore, the results cannot easily be generalized to other indi- viduals, even (in Hermann’s case) to Assad’s fellow Baathist Saddam Hussein. Hence, this approach is not well suited to addressing the debate between realism and liberalism in international relations, although it is an essential part of under- standing and predicting the behavior of a specific state and its leader. The approach used in this article is not to develop a full map of the world- view, but rather to address only those aspects of the worldview relevant to the representation of external conflicts. Further simplifying this approach, it assumes that this partial worldview could align more or less with realism or liberalism. This is a reasonable assumption if we consider where worldviews, especially the worldviews of foreign policy decision-makers, originate. The lens of academia ultimately frames their vision of international affairs. They are trained either directly in universities or by advisors with formal training. Their vision, however, will be somewhat blurred. Decision-makers are usually not theorists (Kissinger being an exception), because they must use judgment to make foreign policy deci- sions that affect their state (George, 1993). Therefore, what they understand of international political theory will be stripped down—a worldview lacking in the subtlety of the theories of Waltz (1979) or Russett (1993). Most commonly, national decision-makers will see the broad outlines, dimly traced to illuminating undergraduate lectures, of assertions as to what actions are generally appropriate. In other words, worldviews do not spring randomly from each individual’s unique experiences, but are learned through a combination of formal study and socialization with other policymakers. Doyle (1997, p. 36) sug- gested that at least in “Western” countries, these worldviews will be parallel to the major theoretical approaches. Tetlock (1993) observed that “policymakers have been found to rely heavily on theory-driven as opposed to data-driven pro- cessing of incoming evidence” (p. 323) and that “the influence tactics that policymakers adopt are profoundly colored by psychological and political assumptions they hold concerning (a) the most effective strategies for eliciting desired responses from other states and (b) the nature of other states and the prob- able responses of the states” (p. 326). Theories of international politics contain assertions analogous to those of a worldview about how international actors behave. A leader may naïvely believe that a multilateral institution can lead to lasting cooperation between states; the astute theorist may conclude that the presence of orderly regimes alters the basic causal variables that would otherwise promote conflict between states (Krasner, 564 Mowle 1983, p. 361). Another leader may simply be certain that human nature drives other leaders to lust for power; the sophisticated political scientist may deduce that states must be power-seeking, not security-seeking, because otherwise the security dilemma would have fallen apart years ago to produce harmony (Mearsheimer, 2001; Schweller, 1996, pp. 117–118). Keohane (1986), referring to the importance of developing theory from assumptions, has stated, “No one can cope with the complexities of world politics without the aid either of a theory or of implicit assumptions and propositions that substitute, however poorly, for theory. Reality has to be ordered into categories, and relationships drawn between events” (p. 4). This method evaluates problem representations to infer a worldview that can be compared with realism and liberalism. A problem representation is a mental model of goals, constraints, preferred solutions, and expectations about the effectiveness of various tactics (Beasley, 1998, pp. 81–83; Newell & Simon, 1972; Voss, 1998, pp. 9–13). This is similar to Snyder, Bruck, and Sapin’s (1962, pp. 64–65) “definition of a situation” (Voss & Dorsey, 1992, p. 6). Problem represen- tations help decision-makers to solve complex problems by breaking them down into smaller, more manageable problems. These representations are inherently tied to elements of their belief structure, as discussed by Simon (1969, pp. 68–72), Sylvan, Majeski, and Milliken (1991, p. 328), and Sylvan and Voss (1998). The problem representation can influence future courses of action. Allison (1971, p. 1) classically asked, “Why did the Soviet Union place strategic offen- sive missiles in Cuba?” By so structuring the situation, he specified that the problem facing U.S. decision-makers was “How should we respond to the strate- gic offensive missiles in Cuba?” Further research has shown that once ExComm interpreted the raw intelligence information as “Soviet strategic offensive mis- siles” (as opposed to, perhaps, “deterrent enhancements to the Cuban right to defend itself”), its members became less likely to choose the less forceful options (Sylvan & Thorson, 1992; Thorson, 1984). Problem representations can be found in official statements issued about the state’s policy and purpose. This use of public statements raises an important question about the validity of this approach, which must be addressed before proceeding. One must define foreign policy in terms of observable behavior, not unobservable “goals.” Public statements, an observable behavior, can be aggre- gated to discover trends in policy stances (Hermann, 1978). Much of the best international relations research (Grieco, 1990, pp. 182–183; Hellman & Wolf, 1993; Mastanduno, 1991) uses policy statements to seek out liberalism or realism. This holds as well for much of the research done on problem representations (e.g., Breuning, 1998; Tetlock, 1985; Young, 1998). This article follows the approach of Cottam and McCoy (1998, pp. 129–139), who emphasized that they were not seeking private images, but rather those that represented the collective Carter administration. Because “foreign policy is a public enterprise... one can mean- ingfully refer to publicly-expressed problem representations” (Sylvan, 1998, Worldviews in Foreign Policy 565 p. 340). To a certain extent, one has little choice: Internal documents are rarely available for comparison, and they will not provide very many cases to build upon. Private memoirs are open to the author’s spin, and even alleged transcripts may not faithfully represent the proceedings (Gaenslen, 1992, pp. 172–176). Analyz- ing public statements to discern the underlying worldviews moves toward Riker’s (1977, pp. 28–29) ideal of scientific progress through studying patterns in small, repeated events rather than trying to “generalize about huge events which turn out to belong to classes with very few members.” With a large N, trial balloons or statements that do not accurately reflect the official problem representation can be expected to have limited impact on the overall study, and such inaccurate problem representations would tend to make the analytical results appear to be less significant than they would be if we were able to exclude such cases. Further reducing the concern about the validity of public statements, this article is not concerned with individuals and their personal worldviews. Instead, the focus is on observable “official” problem representations that result from the policymaking process. In order to respond to an external conflict (or any other situation), the decision-makers must agree upon, or agree to accept, a “group” problem representation. Hoffman (1993) and Janis (1982) both provided exam- ples of decision-making groups that come to accept a common representation of the problem despite their own privately held views. Billings and Hermann (1998) concluded, “Among group members there may be different problem representa- tions initially, but it is most unlikely that the group can move to closure on a deci- sion without tacit acceptance, if not total agreement, among most members on the problem representation” (p. 56). Rubino-Hallman (1998) and Sylvan and Haddad (1998) have both demonstrated the formation of group representations in their empirical work. Voss and Dorsey (1992), citing the work of Cottam (1977), con- curred with the approach used in this article: “From the policy adopted by a given state, one can infer a prevailing worldview approximating that held by a hypo- thetical decision maker” (p. 16). If different worldviews would produce different problem representations, then we can infer elements of the underlying collective worldview by beginning with the problem representation. This approach requires us to be able to isolate elements of problem repre- sentations, from which one infers a worldview to compare with realism and lib- eralism. We can begin by treating realism and liberalism as coherent worldviews. We can then deduce what sort of problem representation would be developed by a hypothetical person whose worldview corresponded precisely to realism or lib- eralism. We can identify the elements of those ideal problem representations of an external conflict that would be distinctly realist or liberal. Five such indicators are described below. We can then search for those indicators among problem rep- resentations found in official statements about external conflicts. The presence of these indicators can be used to infer the extent to which the actual worldview that shaped a particular problem representation corresponds to a purely realist or liberal worldview. 566 Mowle Indicators of Realist and Liberal Worldviews Before identifying characteristic indicators of these ideal worldviews, and then moving on to develop hypotheses, we must establish the nature of realism and liberalism. This is more difficult than one might anticipate, for the terms have been used somewhat casually over time. On the other hand, this article is not the place to recreate the entire history of realism and liberalism, for which there are many existing sources in addition to those described here.1 Doyle (1997, pp. 93–136) followed the standard division in describing the classical realism typified by Morgenthau (1985, pp. 4–17) and the structural realism typified by Waltz (1979, pp. 105–107). Legro and Moravcsik (1999, pp. 12–17) asserted that a core assumption of all realisms is that the relevant actors are “rational unitary political units in anarchy”—in other words, states. Moreover, state goals are fundamentally conflictual, although they may be “deterred or dis- suaded” from pursuing their preferences by superior power wielded by other states. Conflicts between states are resolved on the basis of material capabilities. The resulting pattern of behavior, as described by Mearsheimer (1995, pp. 11–12), is mutual fear and suspicion between states, an emphasis on short-term interests, and the pursuit of relative advantage over other states. Mearsheimer (2001) has further elaborated the view that states seek power, which has been called offen- sive realism, following Schweller’s (1996, pp. 117–118) observation that if all states only seek security, then the basis of conflict would have dwindled away over time. This is the realism that forms the basis of both the criteria below and the later hypotheses. Liberalism is, if anything, even less unified. Nye (1988, p. 246) described commercial, democratic, sociological, and institutional variants; these set aside the Wilsonian “liberal idealism,” which is more akin to the utopianism critiqued by Carr (1939). These liberalisms tend to overlap, as described by Doyle (1997, pp. 230–300), in that each type of liberalism has at its root a phenomenon described by Krasner (1983, p. 361): An institution, broadly defined, either mod- erates state preferences or constrains state actions. Rather than relying directly on power to achieve narrow self-interests, “actors invoke norms as shortcuts to their decision problems” (Kratochwil, 1993, p. 471). The following criteria and the later 1 For additional descriptions of the differences and commonalities between liberalism and realism, see Baldwin (1993, pp. 4–8), Kegley (1995), Mearsheimer (1995), Niou and Ordeshook (1991, p. 484), Powell (1994, pp. 340–343), and Wendt (1992, pp. 391–393). Realism is also described by Claude (1962, pp. 90–92), Grieco (1990), Hellman and Wolf (1993), and Kaplan (1957, pp. 23–29). Liberalism is described by Deutsch et al. (1957), Doyle (1986), Goldstein and Keohane (1993), Keohane (1984), Keohane and Nye (1977), Lake (1992), Maoz and Russett (1993), Martin (1992), Mitrany (1966), and Rosenau (1990). Collective security is advocated by Bennett (1988, p. 135), Cusack and Stoll (1992, p. 5), and Riggs and Plano (1994, pp. 100–101) and is seen as unworkable by Betts (1992), Claude (1962, pp. 97–110), Kupchan and Kupchan (1991), Mearsheimer (1995), and Morgenthau (1985, p. 452). Worldviews in Foreign Policy 567 hypotheses follow this general understanding of liberalism, with some emphasis on the most relevant variants—those of democratic peace, derived from the research agenda pioneered by Russett (1993), and neoliberal institutionalism, best described by Keohane (1993). From these idealized forms of realism and liberalism, the following five cri- teria can be deduced as being most likely to distinguish a problem representation of an external conflict that flows from a realist worldview from one that flows from a liberal worldview: 1. Concern for relative versus absolute gains accruing to different actors as a result of the external conflict. This indicator is central to the literature (see, e.g., Keohane, 1993; Niou & Ordeshook, 1991; Waltz, 1979). A realist problem rep- resentation of an external conflict would include the question “Can my state suffer a relative loss based on the outcome of this conflict?” If some other state can achieve a relative gain, then some action (external or internal) should be taken to counter it. Likewise, people with a realist worldview would be interested in exploiting any potential gains their own state could accrue from the conflict, whether it is interstate or intrastate. If a coalition exists, realists would expect to see dissension among the partners over distributing the spoils of war (see Krasner, 1991). Those with a liberal worldview, on the other hand, take positions “based on their assessments of their own welfare, not that of others” (Keohane, 1984, p. 66) and are “indifferent to the gains achieved by others” (Grieco, 1988, p. 487). A liberal problem representation would address whether one’s own state can gain or lose power, without explicit concern for the differential impact on other states. 2. A primary concern with one’s own interests versus collective norms and interests. Neither side of the debate excludes the relevance of either norms or interests, but the emphasis differs. Some realists assert that interests, beyond the basic interest of survival, are assigned by the relative distribution of national capa- bilities (Powell, 1994, pp. 317–318). The more precise interests, however—those likely to be found in policy statements—would be endogenous and unique to the state. Liberals, on the other hand, believe institutional norms can change state leaders’ conceptions of their own interests (Keohane, 1993, p. 271). A realist problem representation would consider the effects of a conflict on the material, short-term and security-related, interests of the state. A liberal problem represen- tation would derive longer-term “interests” from the norms of an international community, often enshrined in a formal institution. One example of this is U.S. National Security Advisor Anthony Lake’s 1993 remark that “to the extent democ- racy and market economics hold sway in other nations, our own nation will be more secure, prosperous and influential” (cited by Layne, 1994, p. 46). Lake’s remarks, and later UN involvement in several internal conflicts, suggest that these norms have evolved to cover internal as well as international conflicts. A collec- tive security system would be the ultimate expression of a liberalism in which interests are defined in terms of a community norm of peaceful resolution of dis- 568 Mowle putes. In Betts’ (1992) words, “community of power replaces balance of power” (p. 203). 3. Emphasis on protection of one’s own state versus protection of others in an international community. The classic arguments for balance of power (e.g., Claude, 1962, p. 127; Waltz, 1979, p. 118) rely on the ultimate fear that others who become more powerful may eventually threaten one’s own state. In contrast, the liberal response is motivated more by outrage than by fear. The illegitimate use of force constitutes “cheating” against an institution of stability, which must be punished lest the institution be brought down (Axelrod & Keohane, 1993, pp. 94–105; Powell, 1991, p. 1308). This view can be traced back to Kant and espe- cially to Grotius. In a Kantian worldview, one shares a common interest with many others against an opposing set of “oppressors” against whom any action is justi- fied, even obligatory. The Grotian worldview emphasizes that states must follow rules so as to maintain a World International Society (Bull, 1985, pp. 30–33). Liberal problem representations would focus on “justice” and identifying the “aggressor,” whereas realist problem representations would focus on “danger” and “threat.” This indicator has a more narrowly military focus than the previous one. 4. Viewing the conflict’s ramifications in terms of the combatants only, versus in terms of the lessons that might be taken by other states. Both realists and lib- erals concern themselves with the aftermath of a conflict, but realism once again casts these concerns more narrowly. Claude (1962, p. 127) saw states in a balance system as most concerned with the participants in the conflict. Morgenthau (1985, p. 67) saw victory in war as often leading to imperialist policies. Waltz (1979) would also focus on the potential relative gain by one of the combatants if it wins. For example, a realist in the early 1990s might have argued that if Serbia were not prevented from conquering Bosnia, it would next try to attack Macedonia or recover Croatia. There would be no reason to presume, however, that any other state would observe Serbian success and conclude that it also could use military force to resolve border grievances. The liberal worldview, on the other hand, con- siders the impact on other countries’ behavior: “Aggression” must be punished, or “genocide” stopped, because the successful use of force could encourage other countries to follow suit. The history of the 1930s, in which all three Axis powers seem to have been encouraged by the lack of effective response to each other’s military actions, provides a solid example. The policy of appeasement was not an expression of liberalism—it was a combination of “liberal” idealism as critiqued heavily by Carr (1939) and a realism of trying to divert war toward other targets (Christensen & Snyder, 1990). Returning to Serbia in the 1990s, the liberal would have been concerned that Serbian victory might undermine international norms against overrunning sovereign members of the United Nations, committing genocide against religious minorities, and using armed force to settle disputes. This might lead other countries (Russia, for example) to follow the same strategy. Layne and Schwarz (1993) cited President Clinton along these lines. Worldviews in Foreign Policy 569 5. Viewing alliances and multilateral institutions as a tool of the state versus as the goal of the state. For realists, coalitions are useful tools for improving one’s chances of favorably influencing the future distribution of national capabilities—“temporary marriages of convenience, where today’s alliance partner might be tomorrow’s enemy” (Mearsheimer, 1995, p. 11). Any coalition- building effort, however, would be independent of any initial actions taken, because the threat, if there is one, must be addressed with or without assistance. The states would also tend not to allow the alliance to direct their actions, because states have a strong interest in retaining their independence (Grieco, 1993, p. 315; see also Hellman & Wolf, 1993, pp. 9–10; Morgenthau, 1985, pp. 201–202; Walt, 1987). For liberals, on the other hand, the international community has defined the norms being violated. Institutions, which need not be organizations (Stein, 1993, p. 46), encourage states to work multilaterally even where interests do not converge. Because maintaining the institution is an important value in itself, indi- vidual states defer to the group decision on actions to be taken, even if that deci- sion is not the state’s ideal policy choice (Stein, 1993). Thus, the position taken with respect to the external conflict will follow the lead and guidance of other states in the institution. In principle, of course, a problem representation could include both sides of some of these criteria. For example, the state could be concerned with both the immediate combatants and the lesson others would take from it. Likewise, both national and community interests could be affected by a war. If both are present, then one looks to other indicators to decide whether the problem representation seems to better fit realism or liberalism; if these are inconclusive, then one eval- uates the problem representation as mixed. Nothing in this method assumes that all problem representations can be categorized, nor that all the worldviews one infers from them will be close to an ideal form of liberalism or realism. Hypotheses This article examines states’ responses to external conflicts—“foreign” wars and military engagements in which they were not an initial participant. Such con- flicts may be remote from any power interest (as in Somalia), or they may involve a military rival (as in the Soviet invasion of Afghanistan) or threaten an impor- tant economic resource (as in the Iraqi invasion of Kuwait). In some of these conflicts, the initial bystander nation eventually intervenes with diplomatic, eco- nomic, or military power. This topic is suitable methodologically, yielding a large number of cases, because the leaders of states must consider how their state should react to an external conflict much more often than to one in which their state is an initial pro- tagonist. Substantive reasons also guide this choice of issue area. As Levy (1989, p. 216) has noted, the field of international relations has paid too little attention 570 Mowle to the issue of interventions into ongoing conflicts.2 This seems curious, consid- ering that all major wars begin as a conflict between two states. For example, the decisions of initial “bystanders” were rather important in the Austro-Serbian war of 1914, the German-Polish war of 1939, the Korean civil war of 1950, and the Iraq-Kuwait war of 1990. This article lays a foundation for examining the condi- tions when intervention into an ongoing conflict is likely, and the expected nature of that intervention. “External conflict” is defined using the Stockholm International Peace Research Institute’s definition of war (Sollenberg, 1995, p. 20): “a contested incompatibility which concerns government and/or territory where armed force is used between two parties, of which at least one is the government of a state”— and for this study, neither party is the government or citizenry of the state observ- ing the war. This does not place a lower limit on the conflict, as in the common definition used by Small and Singer (1982), in part for the sake of thoroughness and in part because the leaders of the initial bystander state under study cannot be sure of the final extent of casualties. This study does not distinguish between intrastate and interstate conflicts because that question may be wrapped up in the problem representation, as in Bosnia. Furthermore, many of these conflicts have both intrastate and interstate elements. One difficulty with this choice of subject is that the literature on international politics is relatively sparse in terms of specific hypotheses concerning what a realist or a liberal might focus on with respect to an external conflict. Blainey (1988, pp. 57–67) noted that the combatants’ expectations about bystanders’ behavior are important, but did not offer a way to predict that behavior. Liberals tend to be less concerned about external conflict than realists. Morgenthau (1985) suggested that “the standard for judging... involvement and for determining the response of... policymakers was whether an important shift in the distribution of power was taking place and hence of the status quo” (p. 57). This raises the question of why the conflict was “important.” The following hypotheses are thus somewhat eclectic in their sources. They are grouped into three levels of analysis. The two systemic hypotheses, based on a state’s position in the international system, must be derived primarily from realism, because liberalism does not have a fully systemic version. Situational hypotheses, based on attributes of the conflicts themselves, come from a mixture of realist and liberal sources. Three of these are derived from Grieco’s (1988) assertion that “the coefficient for a state’s sensitivity to gaps in payoffs... will be greater if a state’s partner is a long-term adversary rather than a long-term ally; 2 This question was a major focus of Blainey (1988) and has also been examined by Alteld and Bueno de Mesquita (1979), Brecher and Wilkenfeld (1982), Bueno de Mesquita (1981), Butterworth (1978), Cusack and Eberwein (1982), Eberwein (1982), Eckhardt and Azar (1978), Gochman and Long (1983), Gochman and Maoz (1984), Haas (1983), Kaw (1990), Kegley and Raymond (1986), Kim (1991), Pearson (1974a, 1974b), Raymond and Kegley (1987), and Siverson and King (1980). Alker and Christenson (1972) and Alker and Greenberg (1977) examined circumstances under which the United Nations intervenes in wars. Worldviews in Foreign Policy 571 if the issue involves security rather than economic well-being; if the state’s rela- tive power has been on the decline rather than on the rise” (p. 501). To these realist hypotheses, we add a liberal one based on whether a democracy is in the conflict. Finally, we have three domestic hypotheses. The one based on economic per- formance is enough of a state attribute to be derived from realism, whereas the other two must be derived primarily from liberalism, because realists do not look within the state. Note that this article does not hypothesize that any individual worldview is changing so readily from liberalism to idealism. Although realist and liberal the- ories can be viewed as distinct “ideal” worldviews, this is not to say that any indi- vidual, or any group of individuals, holds such a “pure” liberal or realist worldview (Doyle, 1997, pp. 17–37). Real worldviews would almost certainly be more complex, along the lines suggested by Russett (1993, pp. 31–34) in his nor- mative explanation of the democracy-peace puzzle. Decision-makers in a market democracy may tend to have one set of beliefs about how other market democ- racies will behave or react, and a very different set of beliefs about how the leaders of less developed states will behave or react, yet these sets of beliefs could easily coexist within a single coherent worldview. In this case, the hypotheses will suggest scripts that are likely to trigger the more realist or liberal dimensions of the complex actual worldview (Larson, 1994; Shank & Abelson, 1977). Further- more, this article does not define any individual worldview, but infers a “group worldview” that has been expressed in the group’s problem representation. Changes in the problem representation may reflect competition among members of the decision group, as seen in the Carter administration (Campbell, 2001; Lebow & Stein, 1993) or in debate within the first Bush administration over whether the war in Bosnia was an internal or international conflict (Baker, 1995, p. 637; Bert, 1997, p. 97; Woodward, 2000, pp. 217–231). This article does not address the process by which a single problem representation is developed from each decision-maker’s individual representation.3 Systemic Hypotheses Hypothesis 1: If a state has its security guaranteed by another state, then it will be more likely to express a liberal problem representation of external con- 3 Axelrod (1977) argued that interests in complex situations are discovered through the decision process, not separately from it. Shapiro and Bonham (1982) argued that the power and interests of the individuals in the group strongly influence the final problem representation. The best recent work on the subject is by Sylvan and Voss (1998). Within that text, Beasley (1998) found approaches that emphasize “competing preferences” inadequate. Voss (1998) indicated that once individuals form their problem representations, they only change them with reluctance: They tend to tinker with the solution or search for scapegoats before redeveloping a representation. He believed that change is more likely in a group context. Sylvan and Haddad (1998), in a manner similar to the story model of jury deliberation set forth by Pennington and Hastie (1987), suggested that groups select from among competing problem representations on the basis of which one creates the more compelling narrative. 572 Mowle flicts than states that must provide their own security. Waltz (1979) suggested that realism may be less common among states whose “preservation were provided by others” (pp. 70–71). This protection reduces their concern over relative gains and losses. (See Table 1 for the operationalization of this and all other independ- ent variables.) Liberal theory would not expect this to be a significant factor. Hypothesis 2: If a state has its security guaranteed by another state, then it will be more likely to express a liberal problem representation of external con- flicts when the distribution of capabilities in the international system is bipolar than when it is not bipolar. Both Waltz (1964; 1979, pp. 161–193) and James and Brecher (1988), following realist theory, have asserted that a multipolar system is less stable, in the sense of preventing war, than a bipolar distribution of capa- bilities. A “protected” state should be less ready to pursue liberal policies when the international system is not bipolar, because war and shifting alliances could change its protected status. Although today’s system might be unipolar as opposed to multipolar, Layne (1993, p. 7) and Mearsheimer (2001) illustrated why realists would expect other states in a unipolar system to behave as if the system were multipolar. Because states must balance to get security, they will balance against even an apparently benevolent unipole. Kegley and Raymond (1994), working from liberal theory, suggested the opposite—that a future multipolar world could be peaceful and stable. Situational Hypotheses Hypothesis 3: If no long-term adversary is involved in the external conflict, either as an initial participant or as a bystander that has chosen to intervene, then an initially bystanding state will be more likely to express a liberal problem representation than it would in conflicts where such an adversary is involved. Grieco (1988, p. 501), writing from realist theory, suggested that the presence of an adversary is one of the situations that would increase “a state’s sensitivity to gaps in payoffs.” Hypothesis 4: If no formal military allies of the observing state were initial parties to the conflict, then the bystander state would be more likely to express a liberal problem representation than it would toward conflicts in which such allies were initial parties. This hypothesis also flows from realist theory, seen in Grieco (1988, p. 501), and has been supported by Reiter (1994, p. 496) and Moul (1988). Although the mere current involvement of a rival is enough to trigger the preceding hypothesis, for this one the ally must be an initial party, not a later participant. Hypothesis 5: If one of the initial combatants is perceived to be a procedural democracy, then bystanding procedural democracies are more likely to express a realist problem representation of the conflict than they would toward conflicts in which no procedural democracy is an initial combatant. Realist theorists would not expect the regime type to affect problem representations, but liberal theorists Worldviews in Foreign Policy 573 Table 1. Variable Operationalization Hypothesis Variable Value 1 Guaranteed = 1 if the state has its security guaranteed by a pole, = 0 if the security state takes responsibility for its own security. 2 Bipolar = 1 if the security guarantee is from a bipole (pre–31 December security 1991), = 0 otherwise. 3 Current rival = 1 if one of the military participants in the conflict is a current rival of the observing state, as defined by Diehl (1985); = 0 otherwise. 3 Long-term = 1 if at least one of the military participants in the conflict is a rival long-term rival of the observing state, as defined by Geller (1993, pp. 180–181), based in turn on Wayman and Jones (1991, pp. 5– 6); = 0 otherwise. 4 Ally = 1 if at least one of the initial participants in the conflict is a formal military ally of the observing state, as defined by Kegley and Raymond (1990, p. 52) and Sorokin (1994, p. 425) [Walt’s (1987, p. 12) definition of alliance is too vague and Reiter’s (1994, p. 495) is too restrictive]; = 0 otherwise. 5 Democracy = 1 if one or more of the principal parties in the conflict receives a score of 7 or greater on democracy in the POLITY III index (Jaggers & Gurr, 1995), which updates POLITY II (Gurr, 1990) [Lake (1992, p. 35) and Maoz & Russett (1993, pp. 628–629) also use this definition, which is similar to the standard definitions used in comparative politics, such as Linz (1975, pp. 182–183)]; = 0 otherwise. 6 Distance = 1 if the conflict is outside the sphere of interests, = 0 if it is within it. 7 Absolute = 1 if economic growth is negative for two consecutive quarters, recession = 0 otherwise. 7 Relative = 1 if the economy is growing at a slower rate than the G-7 [using recession the OECD as an alternative comparison group would produce the same results; no comparison group was found for India], = 0 otherwise. 8 Popularity = the percentage saying they “approve” of the American president’s “job performance” (in The Gallup Poll Monthly and its predecessors, The Gallup Report and the Gallup Opinion Index) or who would vote for the prime minister’s party in Canada (as reported in Hastings & Hastings, 1978–1994). Poll answers are assumed to be valid for 1 month or until the next poll is taken, whichever is shorter. Insufficient data were found to include this variable for India. A small amount of data are missing for Canada and the United States when the gap between consecutive polls exceeded 1 month. 9 Approaching = 1 if an election will occur in the next 3 months (Meernik, 1994, election p. 131), = 0 otherwise. 574 Mowle would. This hypothesis may seem counterintuitive, but note that all interstate external conflicts (so far) involve at least one non-democracy. Thus, they either are between non-democracies or between democracies and non-democracies. Lake (1992) has shown that democracies tend to form coalitions in support of other democracies under attack. This immediate mutual support among democra- cies should apply even if the democracy “starts” the war, in which case the initial assumption would be that the war is being fought for a legitimate reason (Doyle, 1986, pp. 1160–1162). Although lasting alliances are features of liberalism, such tactical coalitions against a current or imminent threat are more consistent with realism. As with the prior two situational hypotheses, the value of this variable can change during the course of the conflict: Democratic nations may cease to be such, or non-democratic ones may adopt a democratic regime. Hypothesis 6: As the distance from a conflict to the bystander state increases, a liberal problem representation becomes more likely to be expressed. Walt (1987) used geographic proximity as one of the leading reasons why a state will perceive a threat from another. Realist theory would expect the representation of an exter- nal conflict to be based in part on the extent of a military threat, so conflicts that occur close to a state’s borders would be viewed through a realist lens. Such con- flicts are more likely to spill over into one’s own state, and they are also more likely than distant conflicts to result in an increased military threat. More distant conflicts, on the other hand, would be more likely than nearby ones to be repre- sented in terms of liberalism. Morgenthau (1985) used the Monroe Doctrine as an example of a state’s heightened concern over conflicts in its immediate neigh- borhood (pp. 55–57). Blainey (1988, pp. 228–242) saw proximity as a leading cause of wars widening to include other powers; his assessment fits what realists would find important. Proximity means more than sharing a border. Bremer (1993, p. 236) defined “contiguous” to include states less than 150 miles distant. A some- what less arbitrary method measures proximity on the basis of a broader geo- graphic region, similar to the old realist notion of “spheres of influence” or the American phrase “in our own backyard.” For the United States, this region includes the Caribbean basin: the Caribbean Islands, Central America, Panama, Colombia, and Venezuela. Nearly all of the U.S. interventions in Latin America occurred in this region; it has special trade and strategic relevance to the United States (see Pastor, 1992, pp. 22–25). President Reagan’s development scheme for Latin America, the Caribbean Basin Initiative, limited itself to this region. Domestic Hypotheses Hypothesis 7: If a state’s economy is in decline, then it is more likely to express a realist problem representation than when its economy is growing. A number of scholars involved in this debate have suggested that links should be sought between the domestic level and the prevalence of realism or liberalism Worldviews in Foreign Policy 575 (Axelrod & Keohane, 1993, pp. 101–102; Baldwin, 1993, p. 23; Grieco, 1993, p. 328; Keohane, 1993, p. 294). Grieco (1988, p. 501) argued from realism that states experiencing economic decline are more sensitive to relative gains than states whose power is increasing—that is, they will be more realist. This could be seen as a corollary of Wolfers’ (1962) “security rich” idea: A state in a reces- sion or depression will adopt a more self-interested view of the world because it can see more potential threats. Although the state might also seek to enhance its economy by adhering to economic institutions, such liberalism in that issue area would not tend to be mirrored by liberalism on security issues.4 Hypothesis 8: As a leader’s popularity among his or her constituency increases, so does the probability that his or her government will express a liberal perspective. The final two hypotheses derive from liberal interest in domestic pol- itics. Hagan (1993) argued that opposition can make it more difficult for leaders to maintain a foreign policy commitment unless broad national interests are involved. It would follow that leaders who perceive a greater threat to their own hold on power would be more likely to represent problems in terms of realist indi- cators. One measure of this in a democracy is the popularity of the head of gov- ernment. Another measure would need to be found if this study were to include non-democracies. Hypothesis 9: If no election in which a leader could lose office is scheduled or required to be held within the following 3 months, then the state is more likely to express a liberal perspective than at times when such an election is approach- ing. Regardless of current popularity, a leader would be more sensitive to losing office as an election approaches (in procedural democracies), as discussed by Meernik (1994). To some extent, acceptance of either this hypothesis or the pre- ceding one would tend to negate the relevance of worldviews as a strongly held belief system. One could argue, however, that the decision-makers’ full world- view includes a value on retaining personal power that overrides the value they place on the national interest. Methodology This study focuses on the period 1978–1994. Reasonable people can always argue for an extension of any time period under study. There is no obvious tem- poral boundary for this study. The goal is to select a period that features a variety of conditions in the international system, to avoid limiting the findings’ general- izability while keeping the research manageable. During these years, relations among the major powers moved from the end of détente to a renewed Cold War, 4 The “diversionary theory of war” (Blainey, 1988, pp. 72–86) does not directly apply because we are looking at bystanders, not initial combatants. Furthermore, Blainey found little evidence to support the idea that states start wars to divert their public from their own internal problems. 576 Mowle followed by the end of the Cold War and the end of bipolarity itself. Moving much earlier would enmesh the research in the Vietnam war, adding difficult questions about coding that conflict without a compensatory gain in explanatory power. Moving much later in time takes the study further into the post–Cold War era, during which one finds a remarkable lack of variance in the independent vari- ables. Once again, the extension would not add immediate substance. Later research, however, should certainly expand the time period under study. Using Sollenberg (1995) and the earlier SIPRI Yearbooks, as well as Tillema (forthcoming) and Sivard (1993), one finds 140 military conflicts during that time. Eight of these 140 conflicts experienced a change in an independent variable during the conflict, generally because of a change to or from democracy or the insertion or withdrawal of a rival’s military forces, so for purposes of this study one could consider there to be 148 conflicts from which to choose. Table 2 lists the 36 conflicts selected from the original 140 (about 26% of the total). Three of the 36 conflicts experienced a change in an independent variable, so one could argue that 39 of 148 conflicts (still about 26%) were selected; these latter numbers are reflected in Table 2. The conflicts were selected in order to achieve a repre- sentative distribution of the independent variables. For example, 12 (26%) of the 47 conflicts that involved a democracy were included in the sample, and 27 (27%) of the 101 conflicts that did not involve a democracy were included in the sample. Table 2 describes the distribution of conflicts in the sample versus the full set of conflicts. To ensure variation at the system level, this study uses the United States as a great power, Canada as a lesser power ally of a great power, and India as a lesser power that is not an ally of a great power. Among allies of the United States, the choice of Canada best minimizes the effect of geography, historic memories of conflict, and extreme cultural differences. India is a regional power with aspira- tions to international leadership; choosing a less ambitious non-aligned state would reduce the frequency of problem representations being stated. Other “inde- pendent” medium or great powers, such as China, the Soviet Union/Russia, Iran, Nigeria, or South Africa, are poorer selections for initial study, because they either adopt a predominantly internal focus or are more likely to have leaders with worldviews culturally alien to realism and liberalism. India’s leadership has links to the British tradition of politics, its own historical experience with balancing among kingdoms, and its more recent promotion of international institutions as a means of promoting peace. Coding Problem Representations This study used a nearly complete set of statements on the 36 conflicts by the United States, Canada, and India.5 The Canadian Ministry of External Affairs and 5 Most Indian texts for 1993 and 1994 were missing. Worldviews in Foreign Policy 577 Table 2. Conflicts Selected for This Study Algeria intrastate (1992–94) Kagara war (1978–79) South Africa intrastate (1983–92) Argentine intrastate (1978–79) Liberia intrastate I (1985–88) South Atlantic war (1982) Burma intrastate (1978–94) Liberia intrastate II (1989–94) Sri Lanka intrastate (1983–94) Contra-Sandinista war (1980–90) Mexico intrastate (1994) Sudan intrastate (1983–94)* Former Yugoslavia (1991–94) Osirak reactor raid (1981) Third Indochina war (1978–89) Ghana intrastate (1981) Pakistan intrastate (1978–87) U.S.-Libya clashes (1981–86) Granada occupation (1983) Panama occupation (1989–90) Uganda intrastate (1981–91) India-Pakistan (1982–92) Peru intrastate (1981–94)* Western Saharan war (1978–91) Intifada (1987–94) Philippines intrastate (1978–94)* Yemen intrastate (1986–87) Iran-Iraq war (1980–88) Russo-Afghan war (1979–89) Yemen intrastate (1994) Iranian hostage crisis (1979–81) Shaba crisis (1978–79) North-South Yemen war (1979) Iraq-Kuwait war (1990–91) Somalia intrastate (1991–94) Zaire-Zambia (1982) Conflicts Total Sampled Percentage Involving a long-term rival 9 3 33% Involving a current rival 15 3 20% Not involving a rival 124 33 27% Involving a democracy 47 12 26% Not involving a democracy 101 27 27% Involving an ally 40 12 30% Not involving an ally 108 27 25% Within a sphere of interest 25 6 24% Outside a sphere of interest 123 33 27% *Peru is coded as a democracy through 1991. The Philippines are coded as a democracy beginning in 1987. Sudan is coded as a democracy from 1986 to 1988. Thus, 36 conflicts are listed above, but the numbers below represent 39 conflicts selected. The five unselected conflicts during this period for which an independent variable changed were Rhodesia/Zimbabwe intrastate, coded as a democracy in 1978 only; El Salvador intrastate, coded as a democracy after 1983; Bangladesh intrastate, coded as a democracy after 1990; the Ogaden war, which included forces of a U.S. rival until August 1979; and Angola intrastate, which included forces of a U.S. rival until August 1988. International Trade publishes speeches and press releases in several collections. These speeches, and some issued by the Canadian Mission to the United Nations and the Office of the Prime Minister, are indexed in Barrett (1982, 1987, 1994). The United States publishes relevant documents in two collections. Weekly Com- pilation of Presidential Documents includes all statements issued by the president or his office. Department of State Bulletin and its successor, U.S. Department of State Dispatch, include selected statements on foreign policy, primarily by offi- cials in the State Department. India’s Ministry of External Affairs publishes a monthly collection of statements, Foreign Affairs Record, which includes ministry press releases and statements of government officials. A search of these sources, and a supplemental Lexis/Nexis search, identified 2,249 codable cases, 578 Mowle where a “case” is a representation of a particular conflict in a text.6 Each of these texts forms a sampling unit (Krippendorff, 1980, pp. 57–61). If the text referred to more than one of the selected conflicts, then it was evaluated repeatedly, once for each of the conflicts addressed in it. The texts were evaluated with a thematic content analysis similar to that of Winter and Stewart (1977). Each discussion of the conflict in each text was coded twice with respect to the five indicators of realist and liberal worldviews discussed above: once for evidence of each of the liberal indicators, and again for evidence of each of the realist indicators. Each indicator was scored in the range 0–2, where 0 indicates the trait was not observed, 2 indicates clear evidence for the trait, and 1 indicates weaker or unclear evidence. Thus, each case has both a liberal and a realist aggregate score, each ranging in value from 0 to 10. Neither these problem representations nor the worldviews inferred from them should be expected to correspond perfectly to either realism or liberalism. Pure realism and pure liberalism are located within a large N-dimensional space within which possible worldviews can be located. A line segment drawn between those two points can be an axis within that space. Like all such line segments, this realist-liberal axis can be bisected by an (n - 1)-dimensional surface; the inter- section of that surface with the line segment can be assigned the value 0. Evalu- ating worldviews as if they were on a continuum between realism and liberalism is in effect to project the observed problem representation (from which one infers the worldview) onto that line and measure its distance from the neutral point. When we hypothesize that a state would be more likely to express a realist problem representation of the conflict, we do not mean that in this situation it shifts from fully liberal to fully realist. Rather, the impact of the variable is to move the projection of the problem representation along that line segment in the direction of realism, perhaps because now indicators of realism are mixed in with indicators of liberalism. To project the problem representation onto that axis, one combines the aggre- gate coding results into a single value: y = ln[(1 + S Liberal Indicators) / (1 + S Realist Indicators)] The ratio is used to locate the inferred worldview relative to a “pure” realist or liberal worldview. The “1” is added to avoid division by zero. The natural log 6 The initial set of texts included an additional 217 cases (0 (indi- cating a tendency toward liberalism) or 0.208.0000.265.0000 * * * * * * 2 Bipolar security >0 -.149.0000 -.153.0002 * * -.249.0024 * * 3 Current rival