Economic Interdependence and International Conflict: New Perspectives on an Enduring Debate

Citation preview

Page vii →

Acknowledgments The chapters in this book grew out of a conference held at the Ohio State University in September 2000. We are grateful to the Mershon Center for generously funding this conference and to the Christopher H. Browne Center for International Politics at the University of Pennsylvania for additional financial support. We are also grateful to the conference participants and two anonymous reviewers for helpful comments on the chapters. Finally, we would like to thank Patrick McDonald, Joseph McGarvey, and Wynn Kimble for their extraordinary efforts in organizing the conference and preparing this book. Page viii →

Page 1 →

Interdependence and Conflict: An Introduction Edward D. Mansfield and Brian M. Pollins Over the past few decades, there has been a surge of interest in the relationship between economic interdependence and political conflict. One view that has gained considerable popularity and empirical support is that heightened interdependence fosters cooperative political relations. Voiced with increasing regularity in both academic and policy circles, this claim has been used to help justify the formation of the European Economic Community, Richard Nixon’s opening to China, Willy Brandt’s Ostpolitik, and Henry Kissinger’s conception of détente with the Soviet Union. Nonetheless, critics of this argument have not been stilled. Some observers maintain that, rather than fostering cooperation, increased interdependence generates political discord. Even more widespread is the argument that economic exchange has no strong bearing on the high politics of national security. This debate is hardly new. For centuries, the nature and strength of the links between interdependence and conflict have been the subject of heated disagreement. Until lately, however, these links remained the subject of remarkably little systematic scrutiny. In recent years, a growing number of studies have attempted to fill this gap in the literature, but they have yet to resolve various crucial questions. Most important, how and to what extent does interdependence influence political antagonism? What are the causal mechanisms driving this relationship? And what is the most appropriate way to test this relationship? The purpose of this book is to shed new light on these key issues. In this introduction, we survey the recent literature on the effects of economic interdependence on political conflict, placing primary emphasis on these questions and on the theoretical and empirical developments needed to answer them. Our central arguments are threefold. First, existing research has Page 2 → focused too much on addressing whether there is a relationship between interdependence and conflict and too little on identifying the underlying micro-foundations of any such relationship. Particularly important in this regard is the need to specify more fully both the domestic and the international political bases of the nexus between interdependence and conflict. Second, the existing literature has placed far too little stress on the contingent nature of the relationship between interdependence and the use of armed force. Recent studies strongly suggest that the strength and direction of the relationship may change over time and across different types of conflict and that the relationship may depend on both domestic and international factors. Third, too little attention has been paid to the definition and measurement of both interdependence and conflict. More specifically, scholars have yet to resolve in what form and in what sense interdependence is expected to influence conflict of which type and at what level of intensity. The following chapters make considerable headway in resolving these issues. This is not to suggest that the authors are of a single mind. They approach this issue from different disciplinary perspectives and theoretical traditions; they focus on different causal mechanisms; and they employ a wide variety of methodological tools. However, these disagreements serve to create a dialogue that enriches our understanding of the relationship between interdependence and conflict. Equally, this volume helps resolve a broader set of debates over the merits of liberal and realist explanations of international relations, the causes of war, and the political economy of national security. It also sheds light on key policy issues, especially those regarding the instruments and application of economic statecraft.

The Conceptual Links between Interdependence and Conflict Central to much of the literature on interdependence and conflict is the longstanding claim that open international markets and heightened economic exchange inhibit interstate hostilities. Liberals have been the most forceful advocates of this thesis and have stressed a variety of different causal mechanisms in developing it.1 One argument—cast primarily at the level of the nation-state—is that economic exchange and military conquest are

substitute means of acquiring the resources needed to promote political security and economic growth (e.g., Staley 1939). As trade and foreign investment increase, there are fewer incentives to meet these needs through territorial expansion, imperialism, and foreign conquest (Rosecrance 1986). Conversely, barriers to Page 3 → international economic activity stimulate conflicts of interest that can contribute to political-military discord (Viner 1951, 259). Another liberal argument—cast largely at the level of the country-pair, or dyad—is that economic intercourse increases contact and promotes communication between private actors in different countries, as well as between governments. Rising contact and communication, in turn, are expected to foster cooperative political relations (Doyle 1997, chap. 8; Hirschman 1977, 61; Stein 1993; Viner 1951, 261). Still another theme stressed by many liberals is that commercial openness generates efficiency gains that, in turn, render private traders and consumers dependent on foreign markets. Because political antagonism risks disrupting economic relations among participants and jeopardizing the gains from trade, these actors have reason to press public officials to avoid military conflicts. For their part, public officials—who rely on societal actors for political support and have an interest in bolstering their country’s economic performance—have reason to attend to such demands. This argument, which is addressed at length in the following chapters, has been a centerpiece of liberal views on war for centuries. Montesquieu, for example, claimed that “the natural effect of commerce is to lead to peace. Two nations that trade together become mutually dependent: if one has an interest in buying, the other has an interest in selling; and all unions are based on mutual needs” (quoted in Hirschman 1977, 80). Whereas Montesquieu’s claim centers on bilateral relations, the argument that heightened economic dependence inhibits belligerence has also been cast at the systemic level of analysis. As Barry Buzan (1984, 598) mentions, a core element of the liberal position is that “a liberal economic order makes a substantial and positive contribution to the maintenance of international security.” However, the liberal view has been criticized by mercantilists and many realists who insist that unfettered economic exchange can undermine national security. Albert O. Hirschman ([1945] 1980), for example, has pointed out that the gains from trade often do not accrue to states proportionately and that the distribution of these gains can affect interstate power relations.2 Shifting power relations, in turn, are widely regarded as a potent source of military conflict (Gilpin 1981; Levy 1989; Mearsheimer 1990). In the same vein, the extent to which trade partners depend on their commercial relationship often varies substantially among the constituent states. If one partner depends on a trading relationship much more heavily than another partner, the costs associated with attenuating or severing the relationship are far lower for the latter than the former state. Under these circumstances, trade may do little to inhibit the less dependent state from initiating hostilities. Page 4 → Another challenge to the liberal thesis emphasizes that states have political reasons to minimize their dependence on foreign commerce and that military expansion offers one way to achieve this end. Hence, as trade flows and the extent of interdependence increase, so do the incentives for states to take military actions to reduce their economic vulnerability (Gilpin 1981, 140–41; Liberman 1996). Consistent with such arguments, Alexander Hamilton asserted in 1796 that protecting the industrial sector from foreign competition would enhance the United States’ “security from external danger” and give rise to “less frequent interruption of their peace with foreign nations” than open trade policies (quoted in Earle 1986, 235). Furthermore, as commerce rises, so does the range of economic issues over which disputes can emerge. In this vein, Kenneth Waltz (1970, 205, 222) maintains that since “close interdependence means closeness of contact and raises the prospect of at least occasional conflict. . . the [liberal] myth of interdependence. . . asserts a false belief about the conditions that may promote peace.” As such, heightened interdependence may actually stimulate belligerence. Finally, a wide variety of studies conclude that international economic relations have no systematic bearing on political conflict (Buzan 1984; Gilpin 1987; Ripsman and Blanchard 1996–97). Many of them hold that hostilities stem largely from variations in the distribution of political-military capabilities and that power relations underlie any apparent effect of economic exchange on military antagonism. That economic ties among the major powers were significant prior to World War I but far less extensive prior to World War II is frequently presented as evidence that such ties have little systematic impact on armed conflict when core national interests are at stake.

Despite enduring and heated debates about the relationship between interdependence and conflict, rigorous empirical analyses of this issue have only emerged recently.3 These studies have improved our understanding of the links between economic interactions and political discord, but they have not generated any consensus. Although many of them find that heightened interdependence inhibits conflict, others find that rising interdependence either has no deterrent effect on war or stimulates antagonism. The reasons for these divergent conclusions, we believe, can be traced to conceptual and methodological differences within this literature that have remained largely unexplored. Particularly important in this regard are central theoretical issues that this research community has yet to address and differences among researchers over how to define and measure both interdependence and conflict. In the remainder of the introduction, we develop these points and discuss how the contributors to this volume address them. Page 5 →

Theoretical Issues for Further Exploration Most of the burgeoning literature on interdependence and conflict consists of empirical efforts to determine whether these factors are related. Though this line of inquiry has yielded a set of important findings, inadequate attention has been paid to determining the causal mechanisms underlying key results. A closer examination of the microfoundations undergirding the links between interdependence and hostilities is sorely needed. Microfoundations Even a casual review of the myriad arguments regarding interdependence and conflict indicates that a wide variety of causal mechanisms have been stressed. From this welter of plausible hypotheses, we draw one simple conclusion: as long as scholars focus primarily on establishing whether these factors are systematically related, there will be various theories to fit the results of almost any empirical study. Greater attention, therefore, needs to be focused on specifying and testing the observable implications of particular causal mechanisms advanced in theories of interdependence and conflict. Doing so is likely to facilitate a better understanding of why and how economic exchange influences the outbreak of armed conflict. Three areas deserve particular attention in further theoretical development: the identification of relevant actors (subnational, national, and supranational), strategic interactions between these actors as they pursue their goals, and the role played by international economic ties in the processes of conflict generation and conflict escalation. We consider these areas in turn. Taking this literature as a whole, the list of relevant actors would seem quite large. Armed conflict between states has been linked in various theories to the interests of consumers, firms, industries, interest groups, national leaders, supranational institutions, and markets. Some arguments are cast at the subnational level of analysis, for example, the widespread claim that firms and consumers have vested interests in commerce that lead them to restrain the state when conflict is on the horizon lest hostilities rupture important economic ties. Other arguments are cast at the state level, as when it is argued that economic dependence motivates leaders to satisfy material needs via conquest rather than trade. Still another set of claims focuses on the dyadic level of analysis, for example, positing that the extent and asymmetry of interdependence between states influence the likelihood that they will resort to force. Finally, a number of different causal explanations are pitched at the Page 6 → supranational or systemic level of analysis. Among these explanations are that trade organizations reduce the likelihood of armed conflict among members, that heightened global trade reduces the prospects of war throughout the system, and that the anticipated negative response of capital markets to war gives national leaders pause before they resort to the sword. As purportedly relevant actors can be found at all levels of analysis, we also see that the choices made by different theories correlate with different schools of thought. Realist explanations tend to view the state as the only relevant actor while liberal explanations tend to focus on subnational and supranational actors. What we lack are more fully articulated models that clearly define the relevant actors and describe how their interests and actions link interdependence to conflict.

Especially important in this regard is the need to explain how the sub-national actors—like consumers, firms, or industries—influence security policy. Surely the ability of these actors to do so depends on the domestic political institutions through which their interests are filtered, a factor given short shrift in most liberal explanations (Mansfield and Pollins 2001, 841–43). In this volume, Beth Simmons argues that liberal claims regarding the pacifying effects of trade—pax mercatoria, as she calls it—will remain incomplete pending the development of a theory of the state that links private commercial interests to public choices about the resort to arms.4 In conjunction with Simmons’s analysis, the subsequent chapters by Christopher Gelpi and Joseph M. Grieco and by Etel Solingen make considerable headway in this regard. Gelpi and Grieco argue that democratic institutions and their associated constraints on national leaders may be a necessary condition for economic interdependence to inhibit conflict. Solingen offers a more complex picture in which the integration of a state into the global economy shapes the interests of varying domestic coalitions. These groups then pursue foreign policies, either pacific or aggressive, to attain their goals. Besides the need to specify the interaction between state actors and societal interests more fully, there is also a need to specify how interdependence influences the process through which wars break out (Barbieri and Schneider 1999, 394). It is widely recognized that wars occur in at least two stages: the outbreak of a dispute between states and the escalation of this dispute to the point where force is applied (Snyder and Diesing 1977). Existing studies have provided considerable insight into how economic exchange influences the outbreak of conflict. However, they have furnished little insight into how economic interdependence influences the escalation of disputes (Mansfield, Page 7 → Pevehouse, and Bearce 1999–2000; Morrow 1999). One possibility is to build on the argument that armed conflict is a consequence of failed interstate bargaining and to link economic ties between nations to the bargaining process. In this vein, interdependence could be viewed as a signal sent in the course of bargaining, the costliness of which is related to the extent of economic exchange between states or the difficulty either state would face in foregoing their economic connections. It might then be feasible to specify the point where a state’s noncommercial interests (for military advantage, say, or defense of the home-land) overcome any effect of trade on hostilities. James D. Morrow’s contribution to this volume revisits these points and offers further directions for theoretical development. Erik Gartzke’s chapter presents an interesting variation on this theme of interdependence, conflict, and “signaling” in strategic interactions. According to Gartzke, highly interdependent states rarely engage in full-blown war because the costs of doing so are prohibitive. If, however, these states realize that war is unlikely, each one may be tempted to engage in acts of brinkmanship against the other(s) to meet its foreign policy goals, since it can rest assured that these acts will not provoke a military reprisal. This scenario implies that interdependence might foster a great deal of low-intensity conflict but that such conflict is unlikely to escalate, thereby helping to reconcile the claims of realists and neomercantilists, on the one hand, and liberals, on the other. Of course, these points are only suggestive. But addressing such issues—all of which involve improving the specification of causal processes—would enrich and deepen our understanding of the interdependence-conflict relationship. Boundedness and Contingency To date, neither liberals nor realists have paid much attention to identifying the conditions under which their claims hold. Instead, arguments about the relationship between interdependence and conflict typically have an air of universality, applying to all actors in all times and places. Yet the contradictory empirical results generated by recent studies suggest that this relationship is more complex than such arguments imply. Moreover, a growing body of empirical literature indicates that the effects of economic exchange on the outbreak of hostilities depend on various domestic and international factors. Devoting more attention to these interaction effects is another important step in promoting a fuller understanding of the interdependence-conflict connection. Page 8 → In their contributions to this volume, Arthur A. Stein, Jack S. Levy, Gregory D. Hess, and Mansfield analyze some of the contingencies in the relationship between economic exchange and hostilities. Stein draws attention to the

contradictions embedded within any trading relationship (i.e., that it contains both cooperative and coercive dimensions) and uses this as a starting point to argue that trade may lead to either conflict or peace under conditions that still await specification. Levy argues, inter alia, that the “trade breeds peace” proposition must be contingent on conditions not identified by the liberals since we live in a world where security interests may trump economic gains. Hess, using a standard, liberal economic framework, shows that interdependence may result in varying equilibrium levels of conflict from a “Kantian peace” to “good wars,” depending on conditions that are once again absent from standard treatments in the literature. Finally, Mansfield argues that the effects of trade flows on conflict are conditioned by the international institutional setting in which foreign commerce is conducted. As noted earlier, theoretical arguments about the interdependence-conflict connection usually are silent on whether their claims are bounded with respect to space or time. Yet the influence of economic exchange on the use of force seems to have changed over time. Many studies have found that heightened trade has inhibited conflict during the period since World War II, and some observers have arrived at similar conclusions based on analyses of the nineteenth and twentieth centuries (Gasiorowski and Polachek 1982; Mansfield 1994; Oneal et al. 1996; Oneal and Russett 1997, 1999a, 1999b; Polacheck 1980; Russett and Oneal 2001; Russett, Oneal, and Davis 1998). John R. Oneal’s chapter provides some additional evidence of this sort. However, other studies focusing on the seventeenth and eighteenth centuries point out how the expansion of major power trade networks within a discriminatory, mercantilist framework aggravated commercial rivalries and sometimes stimulated armed conflict (Holsti 1991; Levy 1999; Levy and Ali 1998; Milton 1999). Commerce therefore has expanded during the past four centuries within two different policy contexts: initially embedded in a more state-directed and imperialist environment during the mercantilist era and later within a more liberal economic regime. Few studies have addressed whether this shift generated a change in either the nature or the strength of the relationship between interdependence and conflict. Instead, large -N studies have focused almost exclusively on the past half-century and have largely ignored whether and how the effects of interdependence have changed over time.5 Taken as a whole, case studies of this relationship have addressed a much longer period. But even these analyses tend to Page 9 → center on the twentieth century (e.g., Copeland 1996, 1999–2000; Papayoanou 1996; Ripsman and Blanchard 1996–97; Solingen 1998), primarily because much of the historical work on the links between interdependence and hostilities addresses World Wars I and II. Important as those wars were, however, it is not clear that they are the best testing ground for theories of these links. One reason is that the primary participants in the world wars were major powers, states that generally had large and relatively well-diversified economies and, as such, were less dependent on economic exchange than their smaller counterparts. Interdependence therefore may have a less pronounced influence on conflict between major powers than on disputes between weaker states. Recent research indicates that the effects of interdependence are conditional on more than just the political power of economic partners. As Mansfield’s chapter discusses, one set of studies found strong evidence that heightened trade flows inhibit the outbreak of military disputes between members of preferential trade arrangements (PTAs)—institutions designed to liberalize commerce among participants (Mansfield and Pevehouse 2000, 2003). In contrast, these same studies found no evidence that trade influences the resort to force among countries that do not belong to such arrangements. Further, Levy (1999; Levy and Ali 1998) concludes that the interaction among commercial rivalry, power relations, domestic politics, and other factors contributed to the friction between England and the Netherlands that bubbled over into war in 1652. In the same vein, Peter Liberman (1999–2000) reports that the effects of interdependence on belligerence during the first half of the twentieth century hinged on the offense-defense balance. And Paul A. Papayoanou (1996, 45) emphasizes that the influence of interdependence on foreign policy is contingent upon the nature of economic ties between status quo and revisionist powers and whether political institutions within the status quo power allow median economic interests a prominent voice. Finally, Håvard Hegre (2000) has shown that domestic economic conditions mediate the impact of interdependence on the likelihood of bilateral hostilities. Specifically, he finds that the liberal claim holds for nation-pairs comprised of advanced industrial societies, but not for developing countries. In short, these studies indicate that whether interdependence promotes or inhibits conflict may depend on the

interactions among various domestic and international factors. To date, however, the ways these factors mediate the relationship have not been addressed in much depth. Additional research on this issue is sorely needed and should help identify the boundaries and limits of liberal and realist claims.6 More generally, too little attention has been devoted to specifying and justifying Page 10 → the appropriate temporal domain for studies of interdependence and hostilities, as well as the set of countries that should be included in empirical analyses. Case-study analyses, for example, have focused primarily on the major powers, although existing theories address a much broader range of countries. Meanwhile, large-N researchers have generated samples made up of a far broader range of countries, but important differences exist among many such samples that deserve closer scrutiny.7 For instance, there is some quantitative evidence that the effects of trade flows on conflict depend on whether all country-pairs or only “politically relevant” dyads (i.e., those that include either geographically contiguous states or a major power) compose the sample being analyzed (e.g., Barbieri 1996a, 1996b; but see Mansfield and Pevehouse 2003; Oneal and Russett 1999a). Taken together, all these issues point to a number of key questions. Should the liberal claim be restricted to market economies because only they develop the private commercial interests with a vested interest in peace? Should it apply primarily to politically relevant dyads? Should the basic argument apply only since the beginning of the nineteenth century—when the virtues of exploiting comparative advantage in trade relations started gaining increased attention— or should it apply to the earlier mercantile era as well?8 Research addressing such questions will help to establish the boundaries of claims about interdependence and conflict. These questions take on still greater importance when we consider that assertions about the relationship between interdependence and conflict are regularly applied in the policy arena. The policy implications of this relationship are examined thoughtfully in two contributions to this volume. Bruce Russett analyzes Sino-American relations and argues that fostering trade ties is an important way to suppress the conflict-generating dimensions of that bilateral relationship. Michael Mastanduno more broadly considers the use of economic exchange as part of engagement strategies whose outcome is uncertain and outlines directions for research that would help us identify the conditions under which such strategies will succeed.

The Conceptualization and Measurement of Interdependence and Conflict Closely intertwined with the theoretical issues raised in the preceding section is a set of important questions concerning how to define and measure both interdependence and conflict. Various conceptualizations have been used, but the differences among them and the empirical implications of these differences Page 11 → have generated relatively little discussion. In this section, we therefore address the operationalization of interdependence and conflict. Conceptualizing Interdependence In the field of international relations, “economic interdependence” has two meanings. First, a group of countries is considered interdependent if economic conditions in one are contingent on those found in the others, for example, if inflation in France quickly places upward pressure on German prices. Second, countries are considered interdependent if it would be costly for them to rupture or forego their relationship, as would be the case if relations between the Organization of Petroleum Exporting Countries and the advanced industrial countries (which rely heavily on petroleum imports) were severed. The first of these is generally referred to as sensitivity interdependence; the second is typically referred to as vulnerability interdependence (Baldwin 1980). The key difference between sensitivity and vulnerability interdependence hinges on the costs that countries would bear should relations between them be disrupted. Although these forms of interdependence—and the differences between them—are fairly straightforward, developing adequate indicators of them is not. First, distinct measures are needed for each of them because they often do not move in lockstep. While there may be extensive and complex economic connections between states (yielding a high level of sensitivity interdependence), they might not find it especially costly to replace these

connections, either by expanding economic interactions with third parties or by making domestic economic adjustments (yielding a low level of vulnerability interdependence). Second, the best measures of sensitivity and vulnerability interdependence involve information about a counterfactual situation, namely, what the costs would be to country A if economic conditions changed in or relations were interrupted with country B. Difficulties obtaining reliable estimates of that situation complicate efforts to measure interdependence, but it is nonetheless important for studies of its effect on conflict to demonstrate an awareness of these costs. Observing Interdependence Economic interdependence has been measured in various ways, with most indicators being closely linked to the flow of international trade. In part, this reflects the paucity of data available on other forms of economic exchange. Though varied, measures of interdependence usually emphasize one of three Page 12 → themes: openness, vulnerability, or gain. Openness indicators are based in one way or another on the ratio of trade to total economic output. They rely on the idea that the higher the fraction of total output crossing state boundaries, the more costly would be the interruption of such flows. Researchers who emphasize the vulnerability theme do not share the same level of consensus regarding measurement. However, they frequently rely on indicators of trade asymmetry. Typically, such indicators are constructed using the portion of trade (imports and/or exports) between a given pair of states, A and B, represented in the total trade of A and the total trade of B. The more these two figures differ, the greater the asymmetry between A and B. The gain theme is somewhat different. As Solomon W. Polachek (1980) points out, the microtheory underpinning the central liberal claim hinges not on trade flows, per se, but on the gains from trade. At best, these gains can only be measured indirectly since, strictly speaking, they presume the observation of a counterfactual condition (viz., what total product would be if there was no cross-border trade). Economists argue that the gains from trade are correlated with import (or export) price elasticities, and Polachek has used this indicator in various studies (Polachek 1992; Polachek and McDonald 1992; Polachek, Robst, and Chang 1999). Unfortunately, the limited availability of price data has severely restricted the range of countries and years over which such elasticities can be used. Polachek (forthcoming), however, recently reported that considerable progress is being made in the collection of such data, a promising development for researchers interested in the gain-from-trade dimension of interdependence. Of these three main conceptualizations, openness has been the most widely employed by far in the literature on interdependence and conflict. Mansfield (1994) uses such an indicator and finds that heightened global trade (as a percentage of global output) was inversely related to the frequency of war throughout the international system during the nineteenth and twentieth centuries. Oneal and Russett (1997, 1999a, 1999b; Oneal et al. 1996; Russett and Oneal 2001; Russett, Oneal, and Davis 1998) employ a related measure in a series of studies cast at the dyadic level of analysis and also report results consistent with liberal claims. Similarly, at the unit level, William Domke (1988, 131) concludes that nations more closely connected to the global economy are less likely to go to war. Thus, studies based on the openness dimension of interdependence offer considerable support for the liberal view.9 Need we look any further? It is frequently argued that the ratio of trade to output—the leading indicator of commercial openness—is a valid measure of both sensitivity and vulnerability interdependence (e.g., Oneal and Russett Page 13 → 1997). This ratio does provide a useful measure of sensitivity interdependence, since it captures the extent to which trade partners’ economies are intertwined. Its validity as an indicator of vulnerability interdependence, however, rests on the claim that as commerce between countries makes up a larger portion of each country’s total economic output, it is increasingly costly for either partner to replace the trade conducted with the other. The basis of this claim can be questioned on three grounds. First, the size of the flow of trade between states (taken either by itself or as a percentage of national income) may not accurately reflect the costs to them if their economic relations were disrupted. Yet, as noted earlier, the magnitude of these costs is central to assessing the extent of vulnerability interdependence (Baldwin 1980; Gasiorowski 1986; Hirschman [1945] 1980; Keohane and Nye 1977). States trading heavily that can easily locate

close substitutes for the goods being exchanged clearly are not very dependent on each other. At the same time, states conducting little trade that would have great difficulty locating substitutes for the goods being exchanged may be highly vulnerable. In this light, it is interesting to note that Norrin M. Ripsman and Jean-Marc F. Blanchard (1996–97), who focus on vulnerability by tracking trade in “strategic” goods, present results at odds with the liberal position. Similarly, the indicator of interdependence offered by Katherine Barbieri (1998) combines measures of the level of trade and trade “salience” (similar to trade concentration, which could be expected to correlate with a difficulty to find substitutes). She reports a positive association between interdependence and conflict.10 A second problem with measuring interdependence based on the ratio of trade flows between states to the national income of each trade partner is that this value tends to be highly correlated with each partner’s economic size (Hegre 2000; Mansfield and Pevehouse 2000). Moreover, it is well known that economically large states tend to be politically powerful and that powerful states are disproportionately likely to become involved in military conflicts. As such, it is important to control for the independent effects of national income in studies of conflict that include the ratio of bilateral trade to national income; otherwise an inverse relationship between this ratio and hostilities might simply reflect the influence of national income alone. Finally, the “cost” conception of vulnerability may be too restrictive since some claims regarding interdependence and conflict center not on the economic consequences of disrupting commerce but rather on the security implications of dependency or highly asymmetric trade relations. Some realists, for example, argue that highly asymmetric interdependence may restrain the Page 14 → weaker partner in a dyad but is unlikely to deter the stronger partner from resorting to force should their strategic interests collide (Hirschman [1945] 1980). Thus, economic ties between states may restrain only one party from resorting to armed force should a dispute arise, while having no effect on (or possibly even inflaming the aggressiveness of) the stronger party. Meanwhile, some Marxist and world-systems scholars view asymmetric trade relations as innately exploitive and argue that this situation may heighten the prospect of conflict (Chase-Dunn 1989; Wallerstein 1984). Empirical exploration of these possibilities within the interdependence-conflict research community is rare. Equally rare are empirical studies employing the gain conceptualization of interdependence, despite the centrality of the efficiency gains from trade to most liberal arguments, as well as to some criticisms of these arguments (e.g., Gowa 1994). While the aforementioned “counterfactual measurement” problem is one reason for the infrequent use of this theme, economists argue that the gains from trade correlate with import (or export) price elasticities and often use this measure as a surrogate. Still, few scholars have incorporated such a measure in trade-conflict studies (Gasiorowski 1986; Polachek 1992; Polachek and McDonald 1992; Polachek, Robst, and Chang 1999). The nature of the key questions on which such studies focus rightly impels most researchers to include a large number of countries in their analysis or to examine cases reaching back a century or more, but as noted earlier, the limited availability of price data makes this sort of wide-ranging inquiry virtually impossible. Nonetheless, future research needs to confront the implications of the gain dimension of interdependence, given its theoretical importance to debates in this field. Regardless of whether empirical studies stress openness, vulnerability, or gain, they almost always rely on trade data to measure interdependence. In many cases, this strategy seems to stem from an implicit assumption that other forms of economic exchange are highly correlated with trade flows. The appropriateness of this assumption, however, is open to question, especially in an era when merchandise trade composes a dwindling fraction of all economic exchange. Further, it is not clear that all aspects of economic interdependence (commercial, capital, monetary, etc.) should have the same effect on hostilities (Barbieri and Schneider 1999; Russett and Oneal 2001, 141). A recent study by Gartzke, Quan Li, and Charles Boehmer (2001) breaks new ground on this question by comparing the effects of international trade flows, monetary relations, and the cross-border movement of capital. More studies of this sort should follow. Similarly, recent work suggests that considering both the international institutions that guide commerce as well as the flow of trade may enrich conventional Page 15 → measures of economic interdependence and add explanatory power to models of conflict (Mansfield and Pevehouse 2000, 2003; Mansfield, Peve-house, and Bearce

1999–2000). Particularly important among such institutions are PTAs, arrangements that can foster interdependence through a variety of channels. The chapter by Mansfield in this volume argues that PTAs inhibit conflict among members and that these arrangements influence the relationship between trade flows and military disputes. A related set of problems bearing on the use of trade flows to measure interdependence is data driven. In this volume, both Barbieri and Russett discuss data issues that have affected the operational indicators of interdependence chosen by different researchers; the cases that are (or are not) included in their studies; and, by extension, the conclusions they come to. These problems do not have simple solutions. As one example, the overwhelming number of studies analyzing the period since World War II rely on the International Monetary Fund’s (IMF’s) Direction of Trade (DoT) statistics on commercial flows. The long period of time during which various Communist and less developed countries did not belong to the IMF generates a considerable amount of missing data for Eastern Europe, the Soviet Union, China, and many other states that need to be included in any comprehensive analysis of interstate conflict. Furthermore, trade data are quite spotty prior to the IMF’s formation in the late 1940s, which has limited the temporal scope of much of the literature on the relationship between trade and conflict. There are also concerns about the IMF’s treatment of missing data. Specifically, the DoT statistics do not distinguish between situations in which two states conduct no trade (or a trivially small amount of trade) in a given year and situations in which trade data is missing (either because the two states do not engage in trade or because no reportable information on the trade they do conduct is available). Unfortunately, a number of key cases for conflict research involve dyads made up of states that engaged in trade, but whose commercial activities are not recorded by the IMF. East Germany and West Germany, Israel and its Arab neighbors, and apartheid South Africa and the “front-line states” are just some examples. This data limitation presents scholars addressing the relationship between trade and conflict with the dilemma of whether to treat all such observations as ones in which no trade was conducted or as missing.11 Either choice ensures that some observations will be coded incorrectly. Calls for better measures of interdependence are hardly new. Indeed, a considerable amount of ink was spilled over this issue a few decades ago (Baldwin 1980; Gasiorowski 1986; Rosecrance and Stein 1973; Rosecrance et al. 1977; Tetreault 1980). But relatively little has been done to heed such calls, and the Page 16 → need for better measures of interdependence is pressing if we are to resolve debates over the relationship between interdependence and conflict. It is also noteworthy that scholars’ choice of theme—openness, vulnerability, or gain— seems to be strongly correlated with their position on the strength and nature of this relationship. Supporters of liberal claims tend to employ indicators emphasizing the themes of openness and absolute gain, whereas supporters of realist and neomercantilist arguments tend to highlight the theme of vulnerability and relative gain. We are not implying that the larger debate reduces to this single dichotomy, but more attention needs to be devoted to assessing why certain indicators of interdependence seem to provide greater support for one set of theories than another set. Conceptualizing Conflict The influence of interdependence hinges not only on the form and facet of economic intercourse being analyzed but also on the type of international conflict being explained. Taken as a whole, research in this area has addressed an extremely broad spectrum of interstate conflict behavior, from hostile statements to full-scale war, while leaving unclear—both theoretically and empirically—whether economic interdependence should affect lowintensity conflict, high-intensity conflict, or both. Similarly, the question of how economic relations influence the escalation of political conflict remains open, as we discussed earlier. The problem is not that studies fail to define the type of conflict being analyzed: indeed, most empirical research is quite clear on that score. Rather, the problem is that the theoretical literature tends to be quite murky about what type of conflict should be analyzed, and, partly as a result, empirical studies tend to define conflict based on the data at hand. The upshot is that existing research focuses on a wide range of different types of conflict, and at least some of the disagreement in the empirical literature can be traced to these differences.

What forms of interstate conflict should this research community address? The most ardent advocates of the liberal position would expect interdependence to inhibit political conflict at all levels of intensity, though much of the oft-cited work by Immanuel Kant, the Manchester liberals, and others centers on war (Doyle 1997, chaps. 7–8). Realists, meanwhile, might readily concede that trade could suppress less salient interstate conflicts, while denying any systematic effect as conflicts become more serious, placing core national interests at stake (e.g., Viner 1951; Waltz 1970). Still other advocates of the liberal proposition might argue just the opposite. States may continue to voice their differences—and perhaps even threaten sanctions or the use of armed force—but domestic trade interests will restrain them from acting on such threats (thus Page 17 → preventing escalation to the highest levels of conflict) lest commerce be disrupted. Once again, the theoretical literature is composed of so many different claims that almost any empirical result can be fit to some extant theory. Improving the microfoundations of theories linking interdependence to conflict and further specifying the contingencies and boundary conditions of these theories will certainly help matters. But more attention to the different types of interstate conflict is also required to resolve existing debates in this research community. Future work should explicitly consider the likely effects of interdependence on lower-intensity conflict (trade disputes, sanctions, and threats of force), higherintensity conflict (mobilization, the use of armed force, and full-blown wars), and the escalatory and de-escalatory processes that move conflicts from one level to another. Observing Conflict Among empirical studies, methodological orientation seems to play a large role in determining the facet of conflict that is addressed. The vast bulk of the historically oriented case studies focus on international war—especially major power war. In contrast, most statistical analyses center on a much broader range of interstate disputes, although this body of research is marked by considerable disagreement about which type of conflict should be addressed. Some of the earliest statistical research on interdependence and hostilities (Polacheck 1980; Pollins 1989a, 1989b) relied exclusively on event data sets, such as the Conflict and Peace Data Bank (COPDAB) and the World Event Interaction Survey (WEIS) (Azar 1980; McClelland and Haggard 1969). The overwhelming number of such studies conducted during the past ten years has focused on militarized interstate disputes (MIDs), which are episodes in which one state threatens, displays, or uses force against another state (Gochman and Maoz 1984). But there has been remarkably little discussion of either why this focus is theoretically appropriate or the implications of shifting among these different data sets.12 The COPDAB, WEIS, and MID data sets capture markedly different types of foreign policy behavior. COPDAB and WEIS record events over the broadest spectrum of international interactions—cooperative as well as conflictual— from low-intensity hostility (such as a verbal protest) to wars. In contrast, the MID data set only records instances involving the threat, display, or use of armed force. Subsequent interactions surrounding this triggering incident are aggregated into a single data point or observation, an “event” that might persist for years. The differences between these two types of data may be substantial in terms of both the conceptualization and the observation of conflict. In Page 18 → the COPDAB tradition, conflict is conceived as a continuous flow marked at regular time intervals. The MID tradition, meanwhile, views conflict as a discrete episode whose time span may be very brief or sometimes quite prolonged. Consider a well-known result from earlier research on interstate conflict: the correlation between “flows” of conflict and cooperation for a given pair of states tends to be high (Dixon 1983). In other words, states that interact often will engage in both cooperative and conflictual ways—finding disagreement though they may be essentially friendly or working to resolve emerging conflicts through more cooperative behavior. This widespread pattern led many researchers using COPDAB or WEIS to construct a “net conflict” or “net cooperation” indicator that captured the overall diplomatic climate between pairs of states (Polachek 1980; Pollins 1989a). But such measures of diplomatic relations are based on a very different way of conceptualizing friendliness and hostility than MIDs, which are by definition episodic, sometimes brief, sometimes prolonged, and at least fairly conflictual.13 How would COPDAB-based and MID-based pictures of interstate conflict patterns compare? Do MIDs simply map onto the highest end of the COPDAB conflict scale? Do countries engaging in MIDs (especially at the lower levels, which include episodes like fishing disputes) also exhibit high flows of cooperative behavior to settle such controversies? Jon C. Pevehouse’s contribution to this volume provides an important initial

look at these questions, while Rafael Reuveny’s chapter calls for an end to exclusive reliance on MID data in this research community and the reintroduction of event-based measures of conflict and cooperation. In sum, more attention needs to be paid to the aspect and type of political conflict that should be the focus of work on interdependence and hostilities. It is clear that the prevailing diplomatic climate, the occurrence or absence of a militarized dispute, and war are only weakly linked. Consequently, they should not be used interchangeably in empirical studies as all-encompassing indicators of “conflict.” The tendency to do just that (albeit implicitly) is one reason why an understanding of the relationship between interdependence and conflict remains elusive. Indeed, research probing more deeply the meaning and measurement of both interdependence and conflict would itself make an important contribution to our understanding of the links between the two.

Methodological Issues and Advances Relevant to the Study of Interdependence and Conflict Despite the lack of attention devoted to the implications of analyzing different types of interstate conflict, quantitative research on the relationship between Page 19 → interdependence and various aspects of conflict has led to the development of important methodological advances that, in turn, seem likely to improve our understanding of these links. Several recent advances are discussed and used in the following chapters. For example, liberal theories typically posit a simultaneous relationship between interdependence and conflict. As Pollins (1989b) points out, underlying many such theories is the presumption that trade-reliant groups in society recognize the welfare-damaging effects of conflict and restrain their governments from resorting to force because conflict reduces trade. Limitations of data and estimation techniques, however, have greatly inhibited attempts to estimate models of interdependence and hostilities in simultaneous form. These limitations grew more severe as the field moved toward use of MID data to measure conflict, since techniques for simultaneous estimation of models including a dichotomous endogenous variable have only begun to emerge. Nonetheless, simultaneousequations models of trade and certain aspects of conflict have been estimated by Polachek (1980) and Mansfield (1994, 186–90), both of whom found that heightened commerce dampens hostilities. In addition, efforts to test for simultaneity using Granger methods were made by Mark Gasiorowski and Polachek (1982) and by Reuveny and Heejoon Kang (1998), although these efforts yielded conflicting results. Finally, Soo Yeon Kim (1998) and Pollins and Reuveny (2000) have applied some newer maximum likelihood techniques to analyze the relationship between trade and military disputes in simultaneous form. Future efforts to specify and estimate the reciprocal nature of the interdependence-conflict relationship will likely be aided by Richard J. Timpone’s contribution to this volume, which provides a thorough exploration of the issue of endogeneity and the current state of simultaneous-equation estimation techniques. In addition, various methodological advances have improved single-equation models of interdependence and conflict, which remain the workhorse of statistical research on this topic. We noted earlier that most statistical analyses of the liberal claim during the past decade have focused on explaining MIDs, events that are almost always coded as a dichotomous variable. Also, such analyses almost always include observations for many dyads in the same year, as well as observations for each dyad over time. This raises two potentially thorny issues for model estimation: temporal dependence across observations and the possibility of heterogeneity (i.e., the possibility that qualitative differences may exist between dyads or even within a given dyad over time) (Beck, Katz, and Tucker 1998; Beck and Katz 2001; Green, Kim, and Yoon 2001; King 2001; Oneal and Russett 2001). Equally, Gary King and Langche Zeng (2001) Page 20 → have argued that the rarity with which MIDs occur can lead to biased coefficient estimates when standard statistical techniques (i.e., logit and probit models) are used to analyze militarized interstate disputes. They have developed a statistical technique for analyzing models where the dependent variable is a rare event that may prove useful in estimating the effects of commerce on conflict. The most recent methodological advances pertaining to this set of topics are discussed by Janet M. Box-Steffensmeier, Dan Reiter, and Christopher J. Zorn. Their chapter shows how the family of duration models (and, more specifically, proportional hazard, “cure,” and “frailty” models) can be used to address these estimation issues. While recent advances in statistical methods hold great promise for this research community, it is important to

underscore our belief that the study of interdependence and conflict will be impoverished if we limit our attention to large-N, data-analytic studies. Resolving issues of historical boundedness, causal mechanisms, and contingency that are central to the relationship between interdependence and conflict—as well as issues surrounding the definition and measurement of both factors—will surely be aided by carefully constructed case studies. Readers will find thoughtful guidance on this approach in the chapter by Norrin M. Ripsman and Jean-Marc F. Blanchard. Finally, the complexity, endogeneity, and possible nonlinear nature of the interdependence-conflict relationship suggests that simulation, or “computational modeling,” could be a powerful tool for this research community. Early efforts to employ this method were made by Pollins (1985) and Pollins and Peter K. Brecke (1987), but relatively few subsequent studies exploring the influence of interdependence on hostilities have used it. In this volume, David H. Bearce and Eric O’N. Fisher demonstrate how computer simulation can provide unique insights into the connection between international economic relations and war. Conclusion Just over a decade ago, a well-known review of the causes of war lamented the dearth of research on the relationship between economic interdependence and hostilities (Levy 1989, 261). Since then, scholars of international relations have addressed this issue with considerable enthusiasm, stimulating a still modest but rapidly growing literature. These recent studies have made considerable headway in assessing some key aspects of the influence of interdependence on political tensions. Nonetheless, this body of literature has yet to resolve many core issues. First, Page 21 → a stronger theoretical foundation is needed for many of the competing claims on the relationship between interdependence and conflict. Second, too little stress has been placed on whether this relationship is stable over time—especially over periods before World War II—and across countries. More generally, there is a growing indication that the strength and nature of the effects of interdependence depend on various domestic and international factors. A better understanding of these factors and how they affect the links between economic exchange and political antagonism is badly needed. Third, existing studies often rely on different definitions and measures of both interdependence and conflict. While that poses no inherent problem, these differences seem to contribute to variations in the results of empirical studies, and existing theories offer no clear guidance as to which definitions and measures are most appropriate. Moreover, the most widely used measures of interdependence are excessively narrow, focusing on trade flows. There is a glaring need to resolve questions about the merits of relying on particular indices of interdependence and conflict, as well as to assess the sensitivity of empirical results to the use of different measures. The following chapters provide fresh insights into these crucial issues. The diversity of theoretical approaches and methodological perspectives represented among the contributors prevents premature closure on these central questions while offering a broad range of avenues for future research. Taken as a whole, this volume makes considerable headway in addressing how and to what extent interdependence influences hostilities, the causal mechanisms driving this relationship, and the most appropriate ways to model and test it. These advances are sure to improve our understanding of the political economy of national security, the causes of war, and the politics of global economic relations. The following chapters also bear on key foreign policy issues. Various Western governments—most recently the Clinton and Bush administrations—have argued that fostering international economic openness will promote both peace and prosperity. Existing studies offer general support for this position, though central questions remain about how to move from declarative statements and claims of empirical regularity about the relationship between interdependence and conflict to the normative realm of policy prescriptions. Equally, using findings grounded at the dyadic or systemic level of analysis as a basis for policy recommendations for individual nations requires that we gain a clearer understanding of exactly how economic statecraft bears on states’ security policy. More generally, additional research is sorely needed to determine more precisely how, when, and to what extent economic interdependence Page 22 → affects the tenor of international politics. The contributions to this volume chart a course for that work.

NOTES Parts of this essay appeared in an earlier form in “The Study of Interdependence and Conflict: Recent Advances, Open Questions, and Directions for Future Research,” Journal of Conflict Resolution 45 (2001): 834–59. 1. For an overview of the various strands of this argument, see Doyle 1997; Keohane 1990; and Stein 1993. 2. On this point, see also Baldwin 1980; Grieco 1990; and Keohane and Nye 1977. 3. The burgeoning empirical literature that has emerged on this issue over the past decade or so includes Barbieri 1996a, 1996b, 1998; Beck, Katz, and Tucker 1998; Beck and Tucker 1996; Blanchard, Mansfield, and Ripsman 2000; Copeland 1996, 1999–2000; Domke 1988; Gartzke 1998; Gartzke, Li, and Boehmer 2001; Gasiorowski 1986; Gelpi and Grieco 2000; Kim 1998; Levy 1999; Levy and Ali 1998; Liberman 1999–2000; Mansfield 1994; Mansfield and Pevehouse 2000, 2003; Mansfield, Pevehouse, and Bearce 1999–2000; Oneal et al. 1996; Oneal and Russett 1997, 1999a, 1999b; Papayoanou 1996; Polachek 1992; Polachek and McDonald 1992; Polachek, Robst, and Chang 1999; Pollins 1989a, 1989b; Pollins and Reuveny 2000; Reuveny and Kang 1998; Ripsman and Blanchard 1996–97; Russett and Oneal 2001; Russett, Oneal, and Davis 1998; and Solingen 1998. For overviews of this literature, see Barbieri and Schneider 1999; Mansfield and Pollins 2001; and McMillan 1997. 4. Polachek (1980) offers the most explicit microtheory for the connection between economic interdependence and interstate conflict. Pollins (1989b) presents a parallel specification to connect international conflict and cooperation on changing levels of trade. But in both works, the linkages between specific actors and state policy are only implicit within the larger argument. Their estimation models, we might say, are simply reduced forms of the theoretical stories they tell. We argue here that this research community should now move beyond such reduced-form specifications. 5. Some studies of the impact of interdependence on conflict have analyzed the period prior to World War II (Barbieri 1996a, 1996b; Oneal and Russett 1999b; Russett and Oneal 2001), as have some analyses of the influence of conflict on interdependence (Gowa 1994; Gowa and Mansfield 1993; Morrow, Siverson, and Tabares 1998). However, a number of these studies focus only on the great powers, and none of them reach back further than the last quarter of the nineteenth century—still long after the mercantilist era—due largely to the lack of reliable economic data for many states. 6. On this issue, see also Keohane 1990; Mastanduno 1999–2000; and Stein 1993. 7. Another complicating aspect of any commercial network is that all bilateral relationships exist within an interconnected web. A change in any bilateral relationship will ripple through many other dyads. Pollins and Kirkpatrick (1987) tried to estimate parameters Page 23 → for the trade-conflict relationship in a system of equations, realizing only limited success. Recently, Penubarti and Ward (2000) have employed modern methods in spatial autocorrelation to address the same question and find that our understanding of the main relationship may be sensitive to such network effects. 8. On the development of both mercantilism and liberal economic thought, see Irwin 1996. 9. However, such support is by no means universal. See, for example, Beck and Tucker 1996; and Beck, Katz, and Tucker 1998. 10. Oneal and Russett (1999a) and Oneal (this volume) adjust Barbieri’s measure and find that an inverse relationship exists between interdependence and conflict. Our point is not to settle this controversy but rather to stress that studies employing explicit measures of interdependence that go beyond the popular “openness” theme are warranted. 11. On this issue, see Barbieri 1996a, 1996b; and Oneal and Russett 1999a. 12. To be clear, our point is not that scholars should necessarily focus on a single type of conflict but rather that it is important to link the hypothesized effects of interdependence more explicitly to different conflict stages and processes. 13. Penubarti and Ward (2000, 10) argue that these very characteristics could make MIDs particularly unsuitable to testing the relationship between trade and conflict. Whether one accepts this point or not, we submit that researchers would be better served by considering various meanings of “interstate conflict” rather than relying exclusively on the given concept and measurement of MIDs.

REFERENCES Azar, Edward E. 1980. The Conflict and Peace Data Bank (COPDAB) Project. Journal of Conflict Resolution 24:143–52. Baldwin, David A. 1980. Interdependence and Power: A Conceptual Analysis. International Organization 34:471–506. Barbieri, Katherine. 1996a. Economic Interdependence: A Path to Peace or a Source of Conflict? Journal of Peace Research 33:29–49. _____. 1996b. Explaining Discrepant Findings in the Trade-Conflict Literature. Paper presented at the 37th annual convention of the International Studies Association, April 16–20, San Diego. _____. 1998. International Trade and Conflict: The Debatable Relationship. Paper presented at the 39th annual convention of the International Studies Association, March 18–21, Minneapolis. Barbieri, Katherine, and Gerald Schneider. 1999. Globalization and Peace: Assessing New Directions in the Study of Trade and Conflict. Journal of Peace Research 36:387–404. Beck, Nathaniel, and Jonathan N. Katz. 2001. Throwing out the Baby with the Bath Water: A Comment on Green, Yoon, and Kim. International Organization 55:487–95. Page 24 → Beck, Nathaniel, Jonathan N. Katz, and Richard Tucker. 1998. Beyond Ordinary Logit: Taking Time Seriously in Binary Time-Series Cross-Section Models. American Journal of Political Science 42:1260–88. Beck, Nathaniel, and Richard Tucker. 1996. Conflict in Space and Time: Time-Series-Cross-Section Analysis with a Binary Dependent Variable. Paper presented at the 92nd annual meeting of the American Political Science Association, August 29–September 1, San Francisco. Blanchard, Jean-Marc F., Edward D. Mansfield, and Norrin M. Ripsman, eds. 2000. Power and the Purse: Economic Statecraft, Interdependence, and National Security. London: Frank Cass. Brandt, Patrick T., John T. Williams, Benjamin O. Fordham, and Brian M. Pollins. 2000. Dynamic Modeling for Persistent Event-Count Time Series. American Journal of Political Science 44:823–43. Buzan, Barry. 1984. Economic Structure and International Security: The Limits of the Liberal Case. International Organization 38:597–624. Chase-Dunn, Christopher. 1989. Global Formation. London: Basil Blackwell. Copeland, Dale C. 1996. Economic Interdependence and War: A Theory of Trade Expectations. International Security 20:5–41. _____. 1999–2000. Trade Expectations and the Outbreak of Peace: Détente, 1970–74, and the End of the Cold War, 1985–91. Security Studies 9:15–58. Dixon, William J. 1983. Measuring Interstate Affect. American Journal of Political Science 27:828–51. Domke, William. 1988. War and the Changing Global System. New Haven: Yale University Press. Doyle, Michael W. 1997. Ways of War and Peace: Realism, Liberalism, and Socialism. New York: W. W. Norton.

Earle, Edward Mead. 1986. Adam Smith, Alexander Hamilton, Frederich List: The Economic Foundations of Military Power. In Makers of Modern Strategy from Machiavelli to the Nuclear Age, ed. Peter Paret, 217–61. Princeton: Princeton University Press. Gartzke, Erik. 1998. Kant We All Just Get Along? Opportunity, Willingness, and the Origins of the Democratic Peace. American Journal of Political Science 42:1–27. Gartzke, Erik, Quan Li, and Charles Bohmer. 2001. Investing in the Peace: Economic Interdependence and International Conflict. International Organization 55:391– 438. Gasiorowski, Mark. 1986. Economic Interdependence and International Conflict: Some Cross-National Evidence. International Studies Quarterly 30:23–38. Gasiorowski, Mark, and Solomon W. Polachek. 1982. Conflict and Interdependence: East-West Trade Linkages in the Era of Detente. Journal of Conflict Resolution 26:709–29. Gelpi, Christopher, and Joseph M. Grieco. 2000. Democracy, Interdependence, and the Liberal Peace. Typescript, Department of Political Science, Duke University. Page 25 → Gilpin, Robert. 1981. War and Change in World Politics. New York: Cambridge University Press. _____. 1987. The Political Economy of International Relations. Princeton: Princeton University Press. Gochman, Charles S., and Zeev Maoz. 1984. Militarized Interstate Disputes, 1816–1976: Procedures, Patterns, and Insights. Journal of Conflict Resolution 28:585–615. Gowa, Joanne. 1994. Allies, Adversaries, and International Trade. Princeton: Princeton University Press. Gowa, Joanne, and Edward D. Mansfield. 1993. Power Politics and International Trade. American Political Science Review 87:408–20. Green, Donald P., Soo Yeon Kim, and David Yoon. 2001. Dirty Pool. International Organization 55:441–68. Grieco, Joseph M. 1990. Cooperation among Nations: Europe, America and Non-Tariff Barriers to Trade. Ithaca: Cornell University Press. Hegre, Håvard. 2000. Development and the Liberal Peace: What Does It Take to Be a Trading State? Journal of Peace Research 37:5–30. Hirschman, Albert O. [1945] 1980. National Power and the Structure of Foreign Trade. Reprint, Berkeley: University of California Press. _____. 1977. The Passions and the Interests: Political Arguments for Capitalism before Its Triumph. Princeton: Princeton University Press. Holsti, Kalevi J. 1991. Peace and War: Armed Conflicts and International Order, 1648–1989. Cambridge: Cambridge University Press. Irwin, Douglas A. 1996. Against the Tide: An Intellectual History of Free Trade. Princeton: Princeton University Press. Keohane, Robert O. 1990. Economic Liberalism Reconsidered. In The Economic Limits to Politics, ed. John Dunn, 165–94. Cambridge: Cambridge University Press. Keohane, Robert O., and Joseph S. Nye. 1977. Power and Interdependence: World Politics in Transition. Boston:

Little, Brown. Kim, Soo Yeon. 1998. Ties That Bind: The Role of Trade in International Conflict Processes, 1950–1992. Ph.D. diss., Yale University. King, Gary. 2001. Proper Nouns and Methodological Propriety: Pooling Dyads in International Relations Data. International Organization 55:497–507. King, Gary, and Langche Zeng. 2001. Explaining Rare Events in International Relations. International Organization 55:693–715. Levy, Jack S. 1989. The Causes of War: A Review of Theories and Evidence. In Behavior, Society, and Nuclear War, ed. Philip E. Tetlock, Jo L. Husbands, Robert Jervis, Paul C. Stern, and Charles Tilly, 1:209–313. New York: Oxford University Press. _____. 1999. The Rise and Decline of the Anglo-Dutch Rivalry, 1609–1689. In Great Power Rivalries, ed. William R. Thompson, 172–200. Columbia: University of South Carolina Press. Levy, Jack S., and Salvatore Ali. 1998. From Commercial Competition to Strategic Rivalry to War: The Evolution of the Anglo-Dutch Rivalry, 1609–52. In The Page 26 → Dynamics of Enduring Rivalries, ed. Paul F. Diehl, 29–63. Urbana: University of Illinois Press. Liberman, Peter. 1996. Does Conquest Pay? The Exploitation of Occupied Industrial Societies. Princeton: Princeton University Press. _____. 1999–2000. The Offense-Defense Balance, Interdependence, and War. Security Studies 9:59–91. Mansfield, Edward D. 1994. Power, Trade, and War. Princeton: Princeton University Press. Mansfield, Edward D., and Jon C. Pevehouse. 2000. Trade Blocs, Trade Flows, and International Conflict. International Organization 54:775–808. _____. 2003. Institutions, Interdependence, and International Conflict. In Globalization and Armed Conflict, ed. Gerald Schneider, Katherine Barbieri, and Nils Petter Gleditsch, 233–50. Lanham, MD: Rowman and Littlefield. Mansfield, Edward D., Jon C. Pevehouse, and David Bearce. 1999–2000. Preferential Trading Arrangements and Military Disputes. Security Studies 9:92–118. Mansfield, Edward D., and Brian M. Pollins. 2001. The Study of Interdependence and Conflict: Recent Advances, Open Questions, and Directions for Future Research. Journal of Conflict Resolution 45:834–59. Mastanduno, Michael. 1999–2000. Economic Statecraft, Interdependence, and National Security: Agendas for Research. Security Studies 9:288–316. McClelland, Charles A., and Gary D. Hoggard. 1969. Conflict Patterns in the Interactions among Nations. In International Politics and Foreign Policy: A Reader in Research and Theory, ed. James N. Rosenau, 711–24. New York: Free Press. McMillan, Susan M. 1997. Interdependence and Conflict. Mershon International Studies Review 41:33–58. Mearsheimer, John J. 1990. Back to the Future: Instability in Europe after the Cold War. International Security 15:5–56.

Milton, Giles. 1999. Nathaniel’s Nutmeg. New York: Farrar, Straus and Giroux. Morrow, James D. 1999. How Could Trade Affect Conflict? Journal of Peace Research 36:481–89. Morrow, James D., Randolph M. Siverson, and Tressa E. Tabares. 1998. The Political Determinants of International Trade: The Great Powers, 1907–1990. American Political Science Review 97:649–61. Oneal, John R., Frances H. Oneal, Zeev Maoz, and Bruce M. Russett. 1996. The Liberal Peace: Interdependence, Democracy, and International Conflict, 1950–1985. Journal of Peace Research 33:11–28. Oneal, John R., and Bruce M. Russett. 1997. The Classical Liberals Were Right: Democracy, Interdependence, and Conflict, 1950–1985. International Studies Quarterly 41:267–94. _____. 1999a. Assessing the Liberal Peace with Alternative Specifications: Trade Still Reduces Conflict. Journal of Peace Research 36:423–42. _____. 1999b. The Kantian Peace: The Pacific Benefits of Democracy, Interdependence, and International Organizations. World Politics 52:1–37. Page 27 → _____. 2001. Clear and Clean: The Fixed Effects of the Liberal Peace. International Organization 55:469–86. Papayoanou, Paul A. 1996. Interdependence, Institutions, and the Balance of Power: Britain, Germany, and World War I. International Security 20:42–76. Penubarti, Mohan, and Michael D. Ward. 2000. Commerce and Democracy. Working paper 6, Center for Statistics and the Social Sciences, University of Washington. Polachek, Solomon W. 1980. Conflict and Trade. Journal of Conflict Resolution 24:55–78. _____. 1992. Conflict and Trade: An Economics Approach to Political International Interactions. In Economics of Arms Reduction and the Peace Process, ed. Walter Isard and Charles Anderton, 89–120. Amsterdam: North-Holland. _____. Forthcoming. Trade-Based Interactions: An Interdisciplinary Perspective. Conflict Management and Peace Science. Polachek, Solomon W., and Judith McDonald. 1992. Strategic Trade and the Incentive for Cooperation. In Disarmament, Economic Conversions and the Management of Peace, ed. Manas Chatterji and Linda Rennie Forcey, 273–84. New York: Praeger. Polachek, Solomon W., John Robst, and Yuan-Ching Chang. 1999. Liberalism and Interdependence: Extending the Trade-Conflict Model. Journal of Peace Research 36:405–22. Pollins, Brian M. 1985. Breaking Trade Dependency: Third World Proposals for Alternative Trade Regimes in a Global Simulation. International Studies Quarterly 29:287–312. _____. 1989a. Conflict, Cooperation, and Commerce. American Journal of Political Science 33:737–61. _____. 1989b. Does Trade Still Follow the Flag? American Political Science Review 83:465–80.

Pollins, Brian M., and Peter K. Brecke. 1987. International Economic Processes. In The GLOBUS Model, ed. Stuart A. Bremer, 459–567. Boulder: Westview. Pollins, Brian M., and Grant Kirkpatrick. 1987. Modeling an International Trade System. In Communication and Interaction in Global Politics, ed. Claudio Cioffi-Revilla, Richard L. Merritt, and Dina Zinnes, 65–88. Beverly Hills: Sage. Pollins, Brian M., and Rafael Reuveny. 2000. The Liberal Peace: Testing Propositions in a Simultaneous Framework. Paper presented at the 41st annual convention of the International Studies Association, March 15–18, Los Angeles. Reuveny, Rafael, and Heejoon Kang. 1998. Bilateral Trade and Political Conflict/Cooperation: Do Goods Matter? Journal of Peace Research 35:581–602. Ripsman, Norrin M., and Jean-Marc F. Blanchard. 1996–97. Commercial Liberalism under Fire: Evidence from 1914 and 1936. Security Studies 6:4–50. Rosecrance, Richard. 1986. The Rise of the Trading State: Commerce and Conquest in the Modern World. New York: Free Press. Rosecrance, Richard, Alan Alexandroff, Wallace Koehler, John Kroll, Shlomit Laqueur, and John Stocker. 1977. Whither Interdependence? International Organization 31:425–71. Page 28 → Rosecrance, Richard, and Arthur A. Stein. 1973. Interdependence: Myth or Reality? World Politics 16:1–27. Russett, Bruce M., and John R. Oneal. 2001. Triangulating Peace: Democracy, Interdependence, and International Organizations. New York: W. W. Norton. Russett, Bruce M., John R. Oneal, and David R. Davis. 1998. The Third Leg of the Kantian Tripod for Peace: International Organizations and Militarized Disputes, 1950–1985. International Organization 52:441–67. Snyder, Glenn H., and Paul Diesing. 1977. Conflict among Nations: Bargaining, Decision Making, and System Structure in International Crises. Princeton: Princeton University Press. Solingen, Etel. 1998. Regional Orders at Century’s End: Global and Domestic Influences on Grand Strategy. Princeton: Princeton University Press. Staley, Eugene. 1939. The World Economy in Transition. New York: Council on Foreign Relations. Stein, Arthur A. 1993. Governments, Economic Interdependence, and International Cooperation. In Behavior, Society, and Nuclear War, ed. Philip E. Tetlock, Jo L. Husbands, Robert Jervis, Paul C. Stern, and Charles Tilly, 3:241–324. New York: Oxford University Press. Tetreault, Mary Ann. 1980. Measuring Interdependence. International Organization 34:429–43. Viner, Jacob. 1951. International Economics. Glencoe, IL: Free Press. Wallerstein, Immanuel. 1984. Long Waves as a Capitalist Process. Review 7:559–75. Waltz, Kenneth. 1970. The Myth of National Interdependence. In The International Corporation, ed. Charles P. Kindleberger, 205–23. Cambridge, MA: MIT Press.

Page 29 →

Part 1. Theoretical Foundations Domestic Politics Page 30 →

Page 31 →

Pax Mercatoria and the Theory of the State Beth Simmons No idea is more enticing to policymakers and academics alike than the proposition that economic interdependence encourages peaceful international relations. Policymakers are gratified that trade is a policy lever that governments can influence. Academics are encouraged by the (relative) ease of constructing long time series of bilateral cross-border transactions in goods for most countries. On top of this, economists tell us trade is economically efficient. Policy-makers, scholars, and consumers should all be thrilled that trade and peace are robustly correlated. This is why it is essential to submit the pax mercatoria hypothesis to severe scrutiny, both methodologically and theoretically. This chapter does the latter and focuses on one particular theoretical issue to which few scholars have given serious attention: what is the theory of the state that provides a plausible mechanism linking private trade to public conflict behavior? The first section argues that this question deserves attention. The second section outlines three general approaches to state-society relations and discusses the implications of these for empirical research. The third section concludes and calls for research that includes more meaningful tests, informed by more explicit theories of state-society relations.

The Need for a Theory of the State Why are governments at all influenced by interdependence in their crisis or conflict behavior? Why should a government that has a political dispute with another government be restrained in the way it pursues that dispute if its citizens trade a lot with the citizens of the adversary state? Or, alternatively, why in Page 32 → some cases should trade impel a government to raise the level of conflict behavior in a dispute? What do societal preferences and activities in the commercial area ultimately have to do with the way the state formulates and pursues its foreign policy objectives? These questions have been neglected in the research on interdependence and conflict, largely because scholars doing the empirical work have not devoted sufficient attention to elucidating the theory of the state that guides their thinking. Almost no attention has been given to theorizing the relationship of the state to society that might give rise to particular forms and intensities of interstate conflict behavior. This is supremely ironic, since trade (an activity carried out primarily by private actors) is said to influence the tendency to resort to force (a decision implemented by agents of the state). How do trading interests come to prevail on the state machinery to influence the basic course of foreign policy? To complicate theoretical matters, there are two apparently opposing propositions that have received sustained testing in the empirical literature. One is that interdependence causes or, more likely, inflames violent interstate disputes (Waltz 1970). The other claims that interdependence dampens disputes (Gartzke 1998; Oneal et al. 1996; Polachek 1980). Now, these seem to be mutually exclusive propositions, at least for a particular temporal and spatial domain, unless we mean something different by interdependence in each case.1 Yet few studies have spent much time theorizing the state-society relationship that could potentially provide a causal mechanism. Empirical spadework has far outstripped theoretical specification. This is unfortunate. It not only leads to an uninformative tallying of papers that support one proposition over another but also misses an opportunity to explore the varieties of state-society relationships and the kinds of interdependence that can contribute to serious interstate disputes. The empirical project has largely been stylized as a “realist” versus “liberal” horse race. When it is formulated in this way, scholars are missing rich opportunities to specify the nature of state-society relations that link private profiteering with national foreign (even military) policy. After all, the conception suggested by most neorealist analysis—that conflict behavior is largely the outcome of national governments formulating and pursuing national objectives independently of the interests of civil society, in a strategic confrontation with an adversary—is but one possibility in which the state is conceived as largely autonomous from societal demands. But studies of

interdependence cry out for an explicit consideration of the links that bind a state to its society. How we conceive of these links has profound implications for research. Page 33 →

Theories of the State The Autonomous State One theoretical possibility is the autonomous state assumed in neorealist thought (see, e.g., Skidmore 1993).2 Statist theories posit the state itself as possessing goals distinct from the immediate demands of the society over which it rules. The inspiration for much statist thought is plausibly Weberian. In sharp contrast to Marx (discussed subsequently), Weber asserted that the state cannot somehow be reduced to an extrapolitical phenomenon. On the contrary, the state is useful for those operating in markets precisely because its incumbents obey a logic quite different from that of utilitarian exchange. The ideal-typical bureaucracy is a meritocratic, corporately coherent entity in which individual and corporate goals are incentive compatible. Corporate coherence and autonomy from private entrepreneurial groups require that individual incumbents be to some degree insulated from the surrounding society: in the Weberian tradition, statists are willing to assume that governments (even democratic ones [Nordlinger 1981]) have a significant degree of autonomy from civil society—sufficient autonomy, at any rate, that we can understand major foreign policy decisions as relatively unconstrained by civil society's current attitudes or preferences. In this view policymakers are active, determined individuals who shape the patterns of societal demands, manipulate political resources and alignments, and may even in some circumstances act contrary to the demands of well-endowed private actors.3 It remains to specify just what the autonomous state is trying to achieve. Neorealists stress security above all other objectives, but more eclectic realists suggest that “ideological goals related to beliefs about how societies should be ordered” could drive the United States to resort to the use of force (Krasner 1978, 14). With Vietnam in mind, Krasner argues, “It was the ability of American leaders to free themselves from societal constraints… that allowed them to define and pursue ideological goals in a nonlogical manner, even though this activity weakened the fabric of American society” (16). Mainstream realists, however, view national security as the prime goal of an autonomous state and argue that trade creates externalities that can ultimately threaten security (Gowa 1989; Grieco 1990). In this view, the state, or in some models the nation as a whole, bears the “costs” of interdependence in terms of national security. Foreign policy-making is guided by the “national interest” and either insulated from societal demands or fully consonant with society's long-term interests. There is no need here to examine the role of domestic societal Page 34 → groups in foreign policy formulation. The assumption of state autonomy overrides domestic political and societal constraints.4 Theories of the state that rest on assumptions of autonomy from societal forces and the dominance of security in the autonomous state's utility function plausibly link trade and conflict behavior by examining a specific kind of interdependence: vulnerability and extremely asymmetrical relative gains. Vulnerability interdependence involves serious costs to one party of an interruption to the economic relationship with a potential adversary state (Keohane and Nye 1977). The essence of vulnerability interdependence is the high cost for the state or the nation as a whole of substituting an alternative trade partner in the event of a dispute. A good indicator of this kind of vulnerability interdependence would be price elasticities for essential or strategic goods traded between the two economies. If a country's purchases from a particular trade partner are fairly inelastic with respect to price (typically the case with, e.g., petroleum), we can infer it is quite costly, or perhaps not even possible in the short run, to switch to an alternative supplier. Hirschman's writings supply good reasons to expect the politicization of such relations (Hirschman [1945] 1980). What this mechanism does not imply, however, is that trade per se (imports plus exports as a proportion of gross domestic product) engages a national interest. The state, in this conception, is hardly concerned if The Gap is vulnerable to an interruption in jeans deliveries from Guatemala.

If extreme relative gains can create externalities that threaten security one might expect an autonomous state to take preemptive actions that could escalate into violent interstate conflict to deal with these externalities. This is, however, an inherently weak argument, since security externalities are said by some realists to deter trade in the first place (Gowa 1994). Moreover, trade probably accounts for a small proportion of the differential productivity growth that some scholars suggest leads to major power conflicts (Gilpin 1981). Studies that rely on such mechanisms should at a minimum control for the obvious non-trade sources of conflict: differential domestically generated productivity gains and growth rates. To suggest that these are the result of trade (and not domestic innovations that lead to economic efficiencies, as emphasized by North and other institutional economists [North 1981]) is probably unwarranted. Other kinds of models are also consistent with (indeed, depend on) the assumption of an autonomous state free from societal constraints to pursue the “national interest.” James D. Morrow's suggestion (this volume) that states that govern highly interdependent economies have a fuller range of “signals” they can use to inform their friends and adversaries of their intentions is one such model. The clarity and hence usefulness of such a signal decline if civil Page 35 → society interferes in the state's attempted transmission. On the other hand, some models emphasize that various forms of societal constraints that render a signal costly (e.g., sending the signal entails electoral risks) are precisely the reason that certain governments can signal more credibly than can others. Perhaps there is some optimal degree of autonomy that allows certain states to capitalize on the signaling value of trade interdependence, but it is useful to be explicit about the state/society relationship that underlies a model of this kind. It seems plausible that intensive trading relationships have a positive impact on the information environment in which autonomous states design and execute security policy. After all, economic interdependence necessitates contact on several levels: across members of civil society (private contracting), between civil society and official circles (passing inspections, paying duties, applying for import licenses), and at the government-to-government level (trade negotiations, trade-dispute resolution). It is completely consistent with theories that assume a high degree of state autonomy to attribute the pacific influence of trade to better information, which reduces the possibility of miscalculation and unnecessary conflict escalation. All of these arguments assume and ought explicitly to theorize a state relatively free from societal constraints to pursue the national interest even if this means that civil society—or some segment of it—bears the costs. It is crucial in empirical work to be sure data on interdependence is relevant to security arrangements, if that is what theory suggests autonomous states are responding to. This approach contrasts strongly with society-centered theories of the state, discussed in the following section. Society-Centered Theories of the State The major theoretical alternative to an autonomous state able to fashion foreign policy autonomously is a governing organ that is highly constrained by societal forces. Scholars have long recognized that state-society relations, particularly those that link states and entrepreneurial elites, are crucial (Gerschenkron 1962). Three plausible kinds of societal constraints seem most pertinent to the interdependence-conflict connection. Each of these approaches involves different assumptions about just how the state is constrained and therefore has different implications for empirical testing. Median-Voter Theories

One way to think about how governments go about foreign policy-making is that it is an extension of domestic politics, which we can think of as influenced Page 36 → by the range of “voter preferences.” Electoral logic in these models suggests that governments will be influenced by the preferences of the median voter. In a polity stripped for simplicity's sake of social or economic institutions, democratic governments should be expected to adopt policies that reflect the preferences of the bulk of the electorate. Governments in this model attempt first and foremost to retain office and thus choose policies that maximize political support. Economic interdependence is first and foremost an electoral issue, rather than a security dilemma. Facing an electorate with free-trade proclivities, governments that need to worry about reelection will foster trade and avoid conflict escalation.

The mechanism offered by Christopher Gelpi and Joseph M. Grieco (this volume) is an example of a median-voter approach at work. They hypothesize that the interaction between democracy and trade should reduce conflict escalation. The logic depends on the assumption that, given an electoral constraint, democracies are relatively more likely than autocracies to foster growth (highly appealing to the median voter) and trade (likely to produce growth). Trade is therefore conceived of as a public good. Democratic leaders, ever cognizant of the preferences of most voters for growth and employment, are worried about the macroeconomic effects of trade disruption. It should be obvious that the state in this model is radically constrained in a very positive way: democratic governments find it in their best interest to avoid disruptive international conflict. They do well by doing good, that is, by catering to the demands of electorally empowered (but unorganized) voters who want growth. In contrast to the models discussed subsequently, trade produces a public good, not simply a private benefit (and certainly not a negative security externality). This approach has very little to say about vulnerability interdependence. It is most plausibly tested with broad interdependence measures such as imports plus exports as a proportion of gross domestic product (GDP). It almost certainly is correctly specified as an interaction model, in which democracy enhances the pacific qualities of trade. But it is based on a naive notion of societal constraints on foreign and foreign economic policy; after all, models of trade policy-making in democracies rarely depend on median-voter assumptions. Indeed, if the argument is correct, why don't trade barriers among democracies immediately fall to zero? The answer usually lies in group politics of some kind, explored in the models that follow. Finally, because it views trade as a public good, this approach cannot consider more sophisticated models that speak to the distributive consequences of liberalization and closure (e.g., Rogowski 1989) and the coalitional bases for policy choice. Page 37 → Pluralist Theories

Various forms of pluralism overcome some of the weaknesses of the median-voter approach. Pluralism recognizes groups, rather than individual voters, as the most important unit to influence policy. Groups are a means of articulating interests of various sectors of society and representing those interests to government. Power is dispersed throughout society rather than concentrated within the state. Individuals have various identities, but, crucially, they join together to express their interests to government, which reflects on these and takes the various pressures into account in the public interest. In optimistic versions of pluralist theory, no single interest can come to dominate the policy-making process.5 Some form of pluralist theory lurks behind liberal explanations of the pax mercatoria. Rather than positing publicgoods effects of trade, the assumption more often is that traders gain from undisturbed trade. Traders —not “the state,” not “the public”—face opportunity costs when lucrative trade relationships are disrupted. Hence they organize to defend their collective interests in maintaining peaceful relations. In formulating policy, the state is most receptive to those groups that are best organized and endowed politically.6 Assuming industries comprising the traded-goods sector are well organized, this approach predicts the “pacific effect” generally attributed to liberal theory. Empirical work based on some form of pluralist theory should do more than repeatedly test the statistical relationship between trade and international conflict escalation. If the logic of the argument is correct, then we might also expect countries with large traded-goods sectors to be connected with appeasement policies. Countries that are highly interdependent should be especially bad at deterring, because the threat of force is not credible, and it is not credible because of the domestic interests that are pressing for peace. (Contrast this with the ability of the unconstrained state to send sensitive signals and thus to deter in James D. Morrow's formulation.) The nature of the assumptions linking state and society will have a significant impact on outcomes. For pluralist models, researchers should be looking for evidence that anomalous cases—in which conflict escalates contrary to expectation—are plausibly cases of deterrence failure. The research agenda might then turn to the question of the conditions under which a societally constrained state can credibly deter undesired policies of a potential adversary with whom trading relationships are significant.

As is the case with median-voter approaches, pluralist theory can probably be roughly captured by using traditional measures of interdependence (imports plus exports as a proportion of GDP). The idea of using this empirical Page 38 → measure is that it very roughly picks up the size (though not the organizational capacity) of trading interests within a polity. A more sophisticated empirical treatment would include some measure of political organization of the traded-goods sector. One way to do this is to look at industrial concentration. Where few firms dominate the major industries within the traded-goods sector in a bilateral trading relationship, a pluralist story can be told about their ability to overcome problems of collective action and lobby the government to moderate its conflict behavior in order to protect the sector's opportunity to profit. It is also important to note that liberal pluralism usually assumes that traders prefer to be left alone to trade peacefully without government interference. Critics have for decades questioned this halcyon assumption. Polanyi ([1944] 1957, 140) is one of many to note that markets depend on state action: “The road to the free market was opened and kept open by an enormous increase in continuous, centrally organized and controlled interventionism.” In whose interest laissez-faire shades into puissez-faire has long been the domain of Marxist /Leninist theory, discussed following. Marxist Theories

It is truly ironic that empirical studies of interdependence and conflict have barely acknowledged the possibility that Marxist/Leninist theories might have something interesting to say about the mechanisms that link economic policy and foreign policy. Lenin was clear about the linkages; Marx himself was less so, although theorists drawing from his insights have since fleshed out a theory of the state that speaks directly to the empirical problem. Societal constraints of a particular type are central to Marxist thinking (Miliband 1969). Two broad categories of Marx-inspired theories of the state have developed. Both point to tight linkages between the state and the capitalist class. “Instrumental” approaches generally derive from Marx and Engel's position in the Communist Manifesto that the state is nothing but an Executive Committee for the bourgeoisie. “Structural” Marxists, on the other hand, see the state as somewhat more autonomous and able to foster the accumulation of capital with greater effectiveness than private capitalists themselves (Kolko 1969; Magdoff 1969; Poulantzas 1973). Structural Marxists posit the relative autonomy of the state as necessary to ensure the long-term interest of the capitalist class. In both instrumental and structural versions, the state is seen as distinct from society, but nonetheless as epiphenomenal; that is, the state is essentially a superstructure that serves to rationalize more fundamental forces stemming from class relations based on the means of production.7 Page 39 → Marxist thought is most explicit about the likelihood that “the state” is subject to capture by capitalist interests. Though this is clearly a society-centered approach, gone is the assumption of the liberal median voter for whom the government attempts to generate “public goods.” Gone, too, is the liberal assumption that trading interests are essentially pacific; Lenin's writing exposed the simplicity of such an unqualified assumption. Indeed a possible hypothesis inspired by Marxist/Leninist thought is that, under certain conditions, governments are constrained to intervene, sometimes with a show of force, because of certain firms' preferences or, more structurally, to further the capitalist project generally. Such a formulation might expect conflict to break out on the periphery of the capitalist world, among capitalist competitors (as predicted in Lenin's Imperialism) or between a capitalist power and a highly resistant developing region (e.g., the Opium Wars). Here, we would expect possibly forcible intervention on behalf of threatened trading or investing interests or in areas where economic interdependence (or, in Marxist terms, exploitation) lags far behind its potential. Conventional wisdom views the nineteenth century as the paragon of these dynamics: European governments were called on to intervene to enforce property rights, protect nationals, keep ports and shipping lanes open, and enforce sovereign debt repayment to private creditors on more than one occasion.

Implications for Research: Where Do We Go from Here?

Society-centered approaches assume that state policies “reflect vested interests in society” (Collander 1984, 2), while theories based on state autonomy posit decision making and policy formulation based on independent preferences of those individuals collectively authorized to govern. Research that links high foreign policy with patterns of civil-society interactions cannot avoid being explicit about whether, to what extent, and how states are bound by societal preferences. I have argued that a theory of the state should inform the empirical expectations we choose to test and hence the evidentiary commitments we make. The relatively autonomous state, for example, is far more likely to be influenced by the possibility of vulnerability interdependence than broad-based but highly diffuse mutual trade. Since the autonomous state is so crucial to realist theory, it is essential that better measures of vulnerability be tested. One possibility is to look at trade elasticities, which give a general sense of how costly it would be to substitute one trade partner for another. Another is to look more carefully at the products being traded and attempt to gauge just how Page 40 → strategically important they are to “national security.” Weapons, strategic materials, and fuels are a good starting point, but such a list would have to be sensitive to technological change over time. It may be possible to create an index of interindustry trade that would be related inversely to vulnerability (since similar goods could in a pinch be supplied from one's domestic industrial capacity). Even if data availability makes it difficult to retain a global data set, it is worth eliminating observations from our analyses in order to test more precise specifications of the autonomy-vulnerability argument. One might want to explicitly test a conditional argument that hypothesizes that autonomous states react to greater vulnerability with more highly conflictual behavior than do less autonomous, more societally constrained states. Here, the interaction of autonomy and vulnerability would predict a higher incidence or intensity of military conflict. If we want to capture the strength and activism of the state, it may be possible and desirable to include measures of the size of the public sector as a proxy (Carnoy 1984). It is important to keep in mind that not all theories that assume state autonomy involve expectations about vulnerability interdependence. Broad-based economic relations could change the information environment in which autonomous states operate, in ways that might resemble more pluralist arguments but are logically distinct. Autonomous and constrained states might still differ in their ability to deter given high levels of interdependence. If Morrow's signaling argument is correct, high degrees of autonomy interacted with broad-based trade dependence should lead to deterrence success (such governments have a nuanced set of clear policy signals at their disposal), while highly constrained states should fail to deter when broadly entangled in trade interdependence (the cost to civil society is so prohibitive as to make any threat incredible). The alternative to a statist interpretation is not simply “liberalism.” There are many ways in which governments are constrained in foreign policy-making by civil society. Whether we choose to see the state as responding to a public interest or a particularist interest is a central analytical choice with empirical implications. Median-voter approaches suppose governments moderate their foreign policy behavior to provide macroeconomic benefits; pluralist models modify this proposition by looking at the political power and access of specific groups. Marxists link government behavior directly or indirectly to the interests of the capitalist class. Each of these approaches requires more nuanced evidence than to date has been brought to bear. Pluralist models typically specify only one relevant Page 41 → group—”traders”—when they regress total imports plus exports as a proportion of GDP on conflict behavior. Of course, such a measure does not get at the organizational capacity of traders to influence policy. Data on industrial concentration within a bilateral trading relationship would help. An even more complete specification might include the weight of interests specifically interested in or prepared for escalation—military spending as a proportion of GDP, for example. Once again, I am aware that data are limited, but I would advocate developing these more nuanced models that reflect the relationships actually theorized even if it means observations have to be dropped from the analysis. Finally, we need to consider models that respect the obvious fact that normative and institutional differences affect state-society relations over space and time (Spaulding 1991). If we are prepared to think through how evolving institutional arrangements and ideas have altered how governments relate to the governed, we will be in a position to contextualize any observed regularities by period and region. Global generalization may not be nearly as

interesting as variations based on distinctive state-society relationships.

NOTES 1. But see studies that argue that the relationship is not consistent over time (Barbieri 1996; Gasiorowski 1986).2. The “state” here refers to individuals with the authority to make national public policy decisions, that is, public officials and the government (Ikenberry 1986; Mastanduno, Lake, and Ikenberry 1989). This meaning is narrower than in some accounts (e.g., Katzenstein 1978; Krasner 1978) in which the state is practically synonymous with the national political arena (Jackman 1993, 62–72). 3. The state has the ability to neutralize active opposition to its policies. Special privileges, office, and public recognition provide one way (Staley 1935, 289). The state can also change the ways actors understand particular issues, ultimately convincing them of the best course to follow (Krasner 1978, 81). 4. This is intentionally a stark presentation of “state autonomy” so as to facilitate contrast with societycentered approaches discussed later. For a more nuanced theory of the semiautonomous state, see Peter Evan's (1995) account of “embedded autonomy.” 5. Neoutilitarians of various stripes have of course extended this happy model of balancing interests to examine models of various forms of capture, which typically injects various degrees of inefficiency into economic policy-making. See, for example, Bates 1981. 6. In some versions of pluralist theory, “policy networks” are the central unit of analysis that explain policy outcomes. See Smith 1993. 7. Page 42 →Other Marxist theorists emphasize the suffusion of the state throughout civil society (Gramsci 1971). Althusser (1971) sees organizations such as the church, schools, and even trade unions as part of the Ideological State Apparatus.

REFERENCES Althusser, L. 1971. Lenin and Philosophy. London: New Left Books. Barbieri, Katherine. 1996. Economic Interdependence: A Path to Peace or a Source of Interstate Conflict? Journal of Peace Research 33:29–49. Bates, Robert H. 1981. Markets and States in Tropical Africa: The Political Basis of Agricultural Policies. Berkeley and Los Angeles: University of California Press. Carnoy, Martin. 1984. The State and Political Theory. Princeton: Princeton University Press. Collander, David, ed. 1984. Neoclassical Political Economy: The Analysis of Rent Seeking and DUP Activities. Cambridge, MA: Ballinger. Evans, Peter. 1995. Embedded Autonomy: States and Industrial Transformation. Princeton: Princeton University Press. Gartzke, Erik. 1998. Kant We All Just Get Along? Opportunity, Willingness, and the Origins of the Democratic Peace. American Journal of Political Science 42:1–27. Gasiorowski, Mark. 1986. Economic Interdependence and International Conflict: Some Cross-National Evidence. International Studies Quarterly 30:23–38. Gerschenkron, Alexander. 1962. Economic Backwardness in Historical Perspective. Cambridge, MA: Belknap. Gilpin, Robert. 1981. War and Change in World Politics. Cambridge: Cambridge University Press. Gowa, Joanne. 1989. Bipolarity, Multipolarity, and Free Trade. American Political Science Review 83:1245–56.

______. 1994. Allies, Adversaries, and International Trade. Princeton: Princeton University Press. Gramsci, A. 1971. Selections from Prison Notebooks. London: Lawrence and Wishart. Grieco, Joseph M. 1990. Cooperation among Nations: Europe, America, and Non-tariff Barriers to Trade. Ithaca: Cornell University Press. Hirschman, Albert O. [1945] 1980. National Power and the Structure of Foreign Trade. Reprint, Berkeley: University of California Press. Ikenberry, G. John. 1986. The State and Strategies of International Adjustment. World Politics 39:53–77. Jackman, Robert W. 1993. Power without Force: The Political Capacity of Nation-States. Ann Arbor: University of Michigan Press. Katzenstein, Peter J., ed. 1978. Between Power and Plenty: Foreign Economic Policies of Advanced Industrial States. Madison: University of Wisconsin Press. Page 43 → Keohane, Robert O., and Joseph S. Nye. 1977. Power and Interdependence: World Politics in Transition. Boston: Little, Brown. Kolko, Gabriel. 1969. The Roots of American Foreign Policy. Boston: Beacon Press. Krasner, Stephen D. 1978. Defending the National Interest: Raw Materials Investments and U.S. Foreign Policy. Princeton: Princeton University Press. Magdoff, Harry. 1969. The Age of Imperialism. New York: Monthly Review Press. Mastanduno, Michael, David Lake, and G. John Ikenberry. 1989. Toward a Realist Theory of State Action. International Studies Quarterly 33:457–74. Miliband, Ralph. 1969. The State in Capitalist Society. New York: Basic Books. Nordlinger, Eric A. 1981. On the Autonomy of the Democratic State. Cambridge, MA: Harvard University Press. North, Douglass C. 1981. Structure and Change in Economic History. New York: W. W. Norton. Oneal, John R., Frances H. Oneal, Zeev Maoz, and Bruce Russett. 1996. The Liberal Peace: Interdependence, Democracy, and International Conflict, 1950–85. Journal of Peace Research 33:11–28. Polachek, Solomon W. 1980. Conflict and Trade. Journal of Conflict Resolution 24:57–78. Polanyi, Karl. [1944] 1957. The Great Transformation. Boston: Beacon Press. Poulantzas, Nicos. 1973. Political Power and Social Classes. London: NLB and Sheed and Ward. Rogowski, Ronald. 1989. Commerce and Coalitions: How Trade Affects Domestic Political Alignments. Princeton: Princeton University Press. Rosecrance, Richard. 1986. The Rise of the Trading State: Commerce and Conquest in the Modern World. New York: Basic Books. Skidmore, David, and Valerie M. Hudson, eds. 1993. The Limits of State Autonomy: Societal Groups and Foreign Policy Formulation. Boulder, CO: Westview Press.

Smith, Martin J. 1993. Pressure, Power, and Policy: State Autonomy and Policy Networks in Britain and the United States. Pittsburgh: University of Pittsburgh Press. Spaulding, Robert Mark, Jr. 1991. German Trade Policy in Eastern Europe, 1890–1990: Preconditions for Applying International Trade Leverage. International Organization 45:343–68. Staley, Eugene. 1935. War and the Private Investor. Garden City, NJ: Doubleday, Doran. Waltz, Kenneth. 1970. The Myth of National Interdependence. In The International Corporation, ed. C. P. Kindleberger, 205–23. Cambridge, MA: MIT Press.

Page 44 →

Economic Interdependence, the Democratic State, and the Liberal Peace Christopher Gelpi and Joseph M. Grieco The debate about the relationship between economic interdependence and military conflict is one of the oldest in the field of international relations. Most recently, this debate has been defined by the starkly contrasting conclusions of John R. Oneal and Bruce M. Russett (Oneal and Russett 1997, 1999a; Russett and Oneal 2001; see also Oneal et al. 1996) on the one hand, and Katherine Barbieri (1996, 1998) on the other. While Oneal and Russett contend that economic interdependence promotes peace, Barbieri suggests that it can actually promote conflict. These scholars differ in their research choices with regard to such important matters as the rules for data inclusion and the range of countries that should be analyzed. However, they converge in a decision to conceptualize their research question as one of trying to understand the individual, independent impacts made by economic interdependence and democracy on the occurrence of military conflicts between nations. We seek in this chapter to explore a new path for analyzing democracy, interdependence, and war. Specifically, we explore the possibility that the impact of trade and democracy may be contingent upon one another. We ground this interest in the possible interaction between economic interdependence and democracy in a review of the work of Immanuel Kant and a number of modern writers on interdependence and on the domestic incentives of political leaders. We contend that this literature implies that economic interdependence may reduce the risk of war between democracies but exacerbate the risk of such conflicts between nondemocracies. Thus—along with Oneal and Russett—we suggest that the classic liberals may indeed have been right, but in a Page 45 → manner more complex than anticipated by many modern scholars. Rather than the two acting independently, the combined influence of democracy and interdependence may create a powerful web of constraint that reinforces the zone of peace among increasingly interdependent democracies. However, the absence of mutual democracy may vitiate the pacifying effect of economic interdependence between nations. In terms of the paths of future research discussed by Edward D. Mansfield and Brian M. Pollins in the introduction to this volume, we view this work as an effort to substantiate the microfoundations of the liberal peace. Our contention is that the liberal peace cannot be understood without taking into account the incentives that state leaders have to engage in international conflict and the way that domestic political institutions shape and determine those incentives.

Democracy, Trade, and Peace: The Current Debate As Mansfield and Pollins detail in their introduction, liberal scholars have long contended that economic interdependence can be a powerful source of peace, while realists have remained skeptical of any such pacifying influences. Recently this debate has been centered on the contrasting conclusions of liberal scholars such as Oneal and Russett (Oneal and Russett 1997, 1999a; Russett and Oneal 2001) and realists such as Barbieri (1996, 1998). In particular, the debate has focused on issues of measurement, data selection, and statistical modeling. We contend, however, that this debate may be hampered by conceptual problems as much as by empirical ones. In particular, we believe that scholars have addressed only one of the two traditional liberal arguments relating economic interdependence to the reduction of the probability of war. A second liberal argument linking interdependence to peace has yet to be evaluated systematically. Pursuit of this second argument might help us move forward in understanding the connection between economic integration and world peace. The first traditional liberal argument, put forward by such writers as Adam Smith, Thomas Paine, Norman Angell, and, in recent times, Edward Morse, suggests that economic interdependence by itself increases the value of peace between nations that rationally calculate their interests and thereby reduces the danger of war between them (Doyle 1997; Paine 1956; Angell 1913; Morse 1970). However, there is a second liberal argument about economic interdependence and war. This second argument can be traced to Immanuel Kant and his 1795 work Perpetual

Peace. Kant's basic hypothesis regarding economic interdependence is that it may dampen the risk of war between states if the governments of Page 46 → those states are responsive to and representative of a wider rather than a narrower range of societal interests. In modern terms we would call such states democracies. Kant's argument would lead us to expect that it is the joint appearance of, and the interaction between, growing economic interdependence and mutual democracy that bring about a diminution of the probability of militarized conflicts between states.1 Kant's emphasis on the joint operation of representative government and economic interdependence can be observed in the sharp contrast he draws between the probable impact of economic interactions on his idealized republics and the effects of actual economic interactions between the European nations and foreign groups around the world (Doyle 1997, 252–87). In the latter instances, according to Kant, economic interaction is clearly and closely associated with hostility and war. In contrast, Kant argues that commercial relations that take place between nations that are republics and hence are responsive to the true interests of their populations would be characterized by “hospitality.” For Kant, then, economic exchange between nonrepublics is likely to be associated with inequality, exploitation, and ultimately violence. Between republics, in contrast, the effects of commerce would be different. In this context, because of the opportunity for commerce to be shaped both by mutual respect and by international legal protection, commerce would spur the development of more cosmopolitan populations who would be more amenable to peaceful interchange rather than be prone to war.

Modern Microfoundations for the Kantian Hypothesis Given the longstanding intuition that democracy and trade may somehow act together to prevent military conflict, we seek to embed this expectation in a more systematic theoretical framework that accounts more explicitly for the incentives that individual state leaders have regarding the decision to initiate a militarized conflict. We construct our argument from two basic microlevel assumptions. First, we assume that international military conflict is a result of deliberate cost/benefit decisions made by national leaders. Second, regardless of the political institutions in which they operate, we assume that national leaders have one fundamental objective: to retain office. Leaders may wish to do so out of selfish motives (the desire to use their offices to acquire wealth, prestige, and even more power), altruistic concerns (the desire to produce policy outcomes the leader believes appropriate for his or her community), or some combination of these motives. But, in all circumstances, retaining office Page 47 → is a necessary condition for the achievement of whatever goals are motivating leaders. Given these two assumptions—that national leaders are the key foreign policy actors and that they want to retain office—we ask how the disruption of trade as a result of military conflict will influence the ability of democratic and nondemocratic leaders to retain office. In developing our response, we draw on recent theoretical work on the impact of domestic regime-types on national foreign policy performance put forward by Bueno de Mesquita et al. (1999a, 1999b). These authors share our assumption that leaders in all types of polities have retention of office as their fundamental objective. Furthermore, they contend that two central characteristics define the structure of any polity: (1) the range of citizens who participate in the selection of national leaders (the “selectorate”) and (2) the minimum size of the coalition within the selectorate whose support leaders must enjoy in order to attain and retain national leadership (the “minimum winning coalition”). Bueno de Mesquita et al. observe that leaders believe they are most likely to stay in office by providing benefits to their minimum winning coalition and that such benefits may either be “private goods” (i.e., goods granted directly and enjoyed exclusively by the coalition members) or goods associated with public policy successes that are enjoyed not just by the members of the coalition but by the citizenry as a whole. Democracies tend to be characterized by large selectorates and large minimum winning coalitions that are relatively clearly defined and institutionalized. Most important for our concerns, the winning coalitions within democratic states generally have to be quite broad and tend toward the inclusion of the median member of the selectorate.2 The breadth of these winning coalitions in democracies makes it difficult to satisfy the necessary number of constituents through payments of such private goods as money, desirable work or housing, or other

preferential treatment. Democratic leaders may (and indeed do) engage in these practices, but in order to retain a sufficiently large coalition they will generally also have to provide public policy successes whose benefits are more widely consumable by the public. One key example of such a public policy success is economic growth.3 In authoritarian states the breadth of the selectorate may vary widely. In corporatist or communist authoritarian systems, large sections of the population—such as organized labor or public-sector bureaucrats—may be considered a part of the selectorate. In highly personalized dictatorships, on the other hand, the selectorate may be exceedingly small. In general, however, authoritarian systems are likely to be characterized by relatively small winning coalitions. Because leadership competition is neither routinized nor institutionalized, Page 48 → relatively small groups are capable of seizing and holding power—particularly if they control the use of force. Since authoritarian leaders only need to satisfy a relatively narrow winning coalition, it will often be possible for them to retain office through the provision of private goods to their supporters. These goods may include preferential treatment of various types, but they all include discriminatory practices that advantage the leader's supporters.4 Of course, authoritarian leaders may retain office through the provision of widely consumable public policy benefits as well. For a number of reasons, however, we believe they are likely to prefer the provision of private goods. First, public policy success is difficult to achieve. For example, to the extent that it depends upon international or cyclical influences, the maintenance of economic growth may be beyond the capacity of state leaders. The provision of domestic advantages, however, is generally under the control of the authoritarian leader and can be provided with relatively little difficulty. Second, the provision of widely consumable benefits such as economic growth may actually be dangerous for authoritarian leaders, since resources accrue to the leaders' opponents as well as to his or her supporters.5 These resources may be turned toward efforts to remove the leader so that a different coalition can hold power and extract resources from the society. Consequently, in the absence of clear rules and the abjuration of violence, non-democratic leaders can benefit most by seeking to reward those who support them while simultaneously punishing opponents, often through violence or intimidation, and undermining their future influence in the political process. Bueno de Mesquita et al. provide an institutional model that appears to account for a number of important empirical findings in the international relations literature, including not just the democratic-peace finding but also the finding that democracies on average do better in international conflicts than do nondemocracies (1999a). Most important for our concerns, they suggest further that the need for public policy successes will instill in democratic leaders a higher level of interest in promoting overall national economic growth than is true of nondemocratic states (1999a). Moreover, Bueno de Mesquita et al. (1999a, 157–59) demonstrate empirically that larger winning coalitions are a positive predictor of stronger national economic performance, and on the basis of their analysis they conclude (159) that “big winning coalitions pressure leaders to perform especially well on public policy issues. These leaders have the greatest incentive to provide prosperity for their citizens.”6 We build our argument on these two key insights generated by Bueno de Mesquita et al.: first, democratic leaders need public policy successes to stay in office to a much greater degree than is true for nondemocratic leaders; and second, Page 49 → the need for public policy success gives democratic leaders a greater incentive than autocrats to promote aggregate economic growth. We believe these two facts have major implications for our theoretical understanding of the effects of economic interdependence, because they clarify how democratic and nondemocratic leaders assess the value of economic interdependence and thus how trade affects the prospects for war. The key to this theoretical insight lies in the relationship between trade and economic growth. The standard model of trade has long recognized the capacity for international trade integration to improve the aggregate growth rate of a country during the period in which integration is occurring.7 Moreover, the many different strands of economic analysis that constitute the new “endogenous growth theory” highlight opportunities for trade to boost the national economic growth rates of countries well beyond that anticipated by the standard model of trade. Trade is a powerful mechanism for the expansion and deepening of markets and thereby may by itself encourage new investments in technology; it also creates additional opportunities for local firms to gain access to new technologies. Through these and other mechanisms, trade can directly and indirectly enhance the productivity of the factors of production and thereby raise national growth rates.8

Consequently, we expect democratic leaders to be relatively more concerned than nondemocratic leaders about the prospective effects of the breakdown of foreign trade as the result of a militarized dispute, for it is especially in democracies that the delivery of a public policy success such as economic growth is likely to play an important role in affecting the tenure of leaders.9 This line of inquiry does not imply that nondemocratic leaders are indifferent to national economic growth or that nondemocratic leaders do not sometimes seek to promote such growth. Yet we would suggest that while some autocratic leaders may prefer national economic growth and even use it to retain office, virtually all democratic leaders know that they will rise or fall politically on the basis of their capacity to deliver (or at least be associated with) national economic growth.10 While even nondemocratic leaders may not survive in the face of a precipitous decline in national economic activity (Przeworski and Limongi 1997), they should nevertheless typically be better able than democratic leaders to insulate their narrower minimum winning coalition from the immediate pains associated with trade disruption. By consequence, insofar as trade affects national growth rates, and the latter is vital to the tenure of democratic leaders but is relatively less important to the tenure of nondemocratic leaders, we expect that the prospect of forgoing economic growth as a result of the interruption of normal trading relations will be Page 50 → more salient to democratic as opposed to nondemocratic leaders. Therefore, economic interdependence should be more likely to dampen the desire of democratic state leaders than of nondemocratic leaders to become embroiled in militarized conflicts with trading partners.11 Finally, this line of reasoning recognizes that democracies do not produce perfectly open international markets. In particular, as endogenous tariff theory teaches us (Mayer 1984; Magee, Brock, and Young 1989; Magee and Young 1987; Yang 1995), producer groups that are disadvantaged by economic openness may use democratic procedures to slow or even reverse trade liberalization. Still, while endogenous tariff theory may help us understand why a democracy does not have fully open markets and might elect to raise certain barriers to trade for domestic political reasons, our argument picks up the story one step later than endogenous tariff theory. That is, we suggest that, given that level of trade interconnectedness, democratic leaders will be more likely than nondemocratic leaders to fear being penalized politically if their country suffers reduced efficiency, and reduced prospects for growth, as a result of a prospective militarized dispute that might interrupt the trade the two states currently enjoy.

Some Preliminary Evidence We conducted a preliminary test of this hypothesis relying on the data from Oneal and Russett and from Barbieri. Our analyses essentially replicate their studies, but we add an interaction effect to assess our hypothesis that the impact of interdependence depends on the presence of democracy. We test our hypotheses using a logit estimator developed by King and Zheng (2001) that is specifically designed to analyze rare-events data.12 Our central expectation predicts that the coefficient on the interaction between lower trade dependence and the lower democracy score will be negative. This result would indicate that democracy has a negative impact on the relationship between economic interdependence and the likelihood of military conflict. The results of our analysis are displayed in table 1. Table 1 begins by replicating the results obtained by Oneal and Russett in their 1997 and 1999 articles. Specifically, the first column in this table estimates the impact of interdependence without considering its interactive relationship with joint democracy. In this analysis, the lower level of trade dependence has a negative and statistically significant impact on the probability of a militarized dispute (b = −59.8, p < .05). These results are virtually identical to those previously obtained by Oneal and Russett.13 In the second column of table 1, however, Page 51 → we alter our specification of the relationship between trade and conflict to include an interaction between democracy and trade. These results are consistent with our argument about the contingent effects of trade. The coefficient interaction term is sharply negative and statistically significant (b = -9.2, p < .01). Democracy, then, has a negative impact on the relationship between trade and militarized conflict. Page 52 →

Interestingly, the coefficient for minimum trade dependence alone is positive and statistically significant (b = 70.0, p < .01). Thus for autocratic dyads (lower democracy score = 0), trade actually increases the risk of the onset of militarized conflict. This result was not entirely anticipated by our argument, but neither is the result inconsistent with our logic. Specifically, our examination of leaders' incentive to retain office led us to the hypothesis that interdependence would reduce conflict between democracies. The behavior of authoritarian leaders, however, was harder to anticipate because they can choose whether or not they wish to retain office through public policy successes or the distribution of private benefits. These results would indicate that—as an empirical matter—authoritarian leaders are generally not sensitive to the costs of trade disruption associated with militarized disputes. As a result, trade actually exacerbates conflict between these states as anticipated by realists such as Barbieri. Next we obtained Barbieri's data for trade dependence and replicated the analyses we had performed on Oneal and Russett's data. The results of these analyses are displayed in the third and fourth columns of table 1. As with Oneal and Russett, we began by replicating Barbieri's results without considering the interaction of democracy and trade. The third column of table 1 describes these results. In this case, the impact of the lower level of trade dependence is not substantively or statistically significant (b = -2.8, n.s.). This result is consistent with Barbieri's 1996 and 1998 analyses contending that, at best, trade does not constrain military conflict. In the fourth column of table 1, however, we again introduce the interaction between democracy and trade. When we considered only the independent effects of democracy and trade, the coding rules used by Barbieri and Oneal and Russett led to nearly diametrically opposed conclusions about the impact of economic interdependence. Once we take into account the interaction between democracy and trade, however, the results from the two competing data sets are nearly identical. As was the case with Oneal and Russett, Barbieri's trade data provides strong support for our argument. Specifically, the coefficient on the interaction between the lower democracy score and lower trade dependence is negative and statistically significant (b = –5.1, p < .05). Also mirroring our results from the Oneal and Russett data, the coefficient on the lower trade-dependence score alone is positive and statistically significant (b = 40.9, p < .05). This coefficient indicates that economic interdependence between two autocratic states (lower democracy score = 0) actually increases the probability of a militarized dispute. While the results in table 1 can give us an estimate of the statistical significance of our variables of interest, it is very difficult to gauge their substantive impact directly from the rare-event logit coefficients. Consequently, in figure 1 we present the marginal impact of dyadic trade dependence on the incidence of conflict as the lower democracy score in the dyad changes from 0 to Page 53 → 20.14 Since the results using the Oneal and Russett and Barbieri measures of trade are nearly identical, we present marginal effects for the Oneal and Russett data only. The squares in figure 1 summarize the impact of trade on autocratic dyads (lower democracy score = 0). Under these circumstances, the expectations of Barbieri (1996, 1998) appear to be supported. As the figure indicates, each 0.3 percent increase in trade dependence between autocracies increases the relative risk of a militarized dispute by approximately 25 percent. Overall, a shift across this entire spectrum of trade dependence increases the annual probability of militarized conflict within an autocratic dyad from 0.08 percent to 0.15 percent, a nearly 100 percent increase in the risk of a dispute. The circles in figure 1 describe the impact of trade dependence on conflict in dyads with mixed regime-types (lower democracy score = 10). In this case, increases in trade dependence have a very modest negative impact on the probability of a militarized dispute. An increase in trade from 0 to 0.3 percent of gross domestic product (GDP) leads to a 7 percent reduction in the relative risk of a militarized dispute. Additional increases in trade to 0.6 percent and 0.9 percent of GDP generate 5 percent and 6 percent reductions in the risk of a dispute respectively. Overall, this increase in dyadic trade dependence reduces the Page 54 → probability of a dispute from 0.041 percent to 0.036 percent, a relatively modest 12 percent drop in the risk of a dispute. Within democratic dyads (lower democracy score = 20), however, mutual trade dependence has a substantial pacifying impact. As the triangles in figure 1 indicate, a shift in dyadic trade from a situation in which two democratic states do not trade with one another at all, to one in which the less dependent state enjoys a dyadic

trade volume equal to 0.3 percent of its GDP, reduces the relative risk of a dispute by 29 percent. A continued increase in dyadic trade dependence from 0.3 percent to 0.6 percent reduces the relative risk of a militarized dispute by an additional 27 percent. A further increase in dyadic trade up to 0.9 percent on the part of the less dependent democracy reduces the risk of conflict by an additional 36 percent. Hence, for democratic dyads, this net increase in trade dependence reduces the probability of a militarized dispute from 0.021 percent to 0.007 percent—a 67 percent reduction in the risk of a dispute.

Conclusion Our study indicates that classical liberal scholars were right to draw attention to the constraint that the costs of trade disruption can place on the outbreak of military conflict. However, we also find that sensitivity to these costs depends critically on the structure of a state's domestic regime, as was anticipated by Immanuel Kant. Democratic leaders rely on the support of relatively broad minimum winning coalitions, and as such they must deliver public policy successes such as economic growth; fearful of losing such growth, they are more likely than nondemocratic leaders to be averse to foregoing the benefits of trade, including the capacity of trade to be a motor for aggregate national economic growth. Nondemocratic leaders, relying upon the provision of private goods to a smaller range of key backers, are relatively less likely to be concerned about growth and, therefore, the risk that a militarized international dispute will lead to a rupture of trade and the growth impulse associated with it. Consequently, trade acts as a powerful constraint on conflict within democratic dyads. For autocratic states, however, economic interdependence may actually increase the incidence of military conflict because the entangling of their economies gives them a cause for conflict without introducing commensurate constraints against the use of force. Our results indicate that there is indeed a liberal peace, and it is comprised of at least two distinct influences. First, we find that democratic states are generally less likely to engage in military conflict with one another. This finding holds true even in the absence of interdependence. Second, we find that Page 55 → between democratic states, trade dependence acts as an additional constraint on the outbreak of military conflict.

NOTES 1. While Oneal and Russett (1997) model Kant's argument as if he posited only independent effects for democracy and economic interdependence on the probability of war, their textual interpretation of Kant could be consistent with a reading of Kant that would ascribe an interactive effect between the two. For example, they interpret Kant as suggesting in Perpetual Peace that “peace can be built on a tripod of complementary influences: republican constitutions (i.e., representative democracy), international law and organization, and ‘cosmopolitan law’ (i.e., economic interdependence)” (268, emphasis added). Influences that are complements of one another make each other complete, and in this sense we may say that the magnitude of the impact that each factor identified by Kant will have on the probability of war is conditioned by the presence or absence of the other. Similarly, Oneal and Russett suggest that, based on their reading of Kant, we may expect that “economic interdependence reinforces [domestic] structural constraints and liberal norms by creating transnational ties that encourage accommodation rather than conflict” (269). The idea that economic interdependence reinforces, or accentuates, the effects of democracy and international law would seem to be consistent with the suggestion that there are interactions between the variables that have an impact on the probability of war. We also suggest that while the recent assessment by Russett, Oneal, and Davis of the independent effect of mutual international organizational membership on the probability of militarized conflict between states is helpful, future analysis would benefit from an assessment of the interactive impact of international law and organization with the two variables examined in this chapter. See Russett, Oneal, and Davis 1998 and Oneal and Russett 1999a. 2. Clearly, election rules and procedures do not guarantee that the “median voter” will be represented in every democratic system. For our purposes the important assumption is that democratic winning coalitions will generally be large relative to authoritarian winning coalitions. 3. These goods must be widely consumable, but they need not be “public goods” in the technical sense. That is, the goods need not be nonrival and nonexcludable. For example, the poor may be excluded from the

benefits of economic growth if the leader can build an adequately large coalition of the rich. 4. The concept here is analogous to the use of the term rents in the economics literature. See Lake 1992 for an application of this concept to international conflict. 5. Indeed, insofar as economic growth promotes the development of independent social forces that may come to oppose authoritarian rule, as suggested by modernization theory, nondemocratic leaders might view the curtailment of growth positively! More realistically, nondemocratic leaders may be expected to prefer national economic Page 56 → growth to be sufficiently great that they can extract from the economy benefits sufficient to retain the support of the members of their minimum winning coalition but not so large that it unleashes and empowers social forces that might come to threaten their hold on power. 6. Employing different measurements, Barro (2000, 220–22) finds a somewhat different relationship between regime-type and economic growth, one that tends toward an inverse-U association. That is, initial movement from very low to middle-level democracy scores is associated with an increase in growth, after which further increases are associated with an abatement of national growth performance. 7. Very useful overviews of the theoretical and empirical studies relating economic openness to national economic growth are provided by Edwards (1993, 1998). However, declining economies of scale would eventually curtail the capacity of this specialization in product ranges through trade to raise the rate of economic growth for the country as a whole. For a graphical treatment, see Caves, Frankel, and Jones 1999, 36–39. 8. An early statement of endogenous growth theory was put forward by Romer (1986). Romer (1994) presents an important intellectual history of the development of the endogenous growth research program, and in it he emphasizes the importance of how the expansion of markets, together with government protection of intellectual property (which in turn permits temporary monopoly rents for innovators), increases the incentives for firms to engage in research and development and thereby directly and indirectly increase a nation's skills, thereby enhancing its national rate of growth. For helpful summaries of the development of the different lines of analysis in the endogenous growth research program, see Long and Wong 1997. Grossman and Helpman (1994) provide a superb discussion of how, from the viewpoint of this approach, trade can often, but not always, contribute to improved national growth rates. For reviews of empirical studies that link economic openness to investments in innovations and thus to greater mediumterm economic growth, as well as supplementary statistical tests of the relationship, see Edwards 1998 and Lawrence and Weinstein 1999. 9. That economic growth does have a major impact on leadership tenure in democracies is well documented in political science: indeed, research on American politics and on the politics of other industrialized democracies has indicated that overall macroeconomic performance is the central determinant of leaders' tenure in democratic nations (Fiorina 1978; Kinder and Kiewiet 1981; Kinder, Adams, and Gronke 1989; and Lewis-Beck 1990). 10. If, ex ante, we would identify which authoritarian leaders so prefer public policy successes like economic growth as part of a strategy to extract private goods for themselves or supporters, or to claim governing legitimacy, we would expect international economic interdependence through trade and other mechanisms to have a restraining effect on their use of force similar to that we expect of democratic leaders. 11. In a similar vein, Papayoanou (1996, esp. 66, 77) argues that differences in the way liberal democracies and autocratic regimes respond to economic interdependence help explain Britain's failure to make it clear to Germany well before August 1914 that it Page 57 → would fight the latter if it initiated a war against Belgium and France. Papayoanou argues that the British government was constrained in the years up to 1914 in balancing against a rising German threat because British business interests were deeply concerned that doing so would put their highly profitable economic relationships with Germany into jeopardy. In contrast, Germany's authoritarian regime prevented German economic actors who had a comparable stake in economic relations with Britain (primarily bankers and exporters other than those in heavy industry) from effectively voicing their concerns about a conflict between the two countries. Thus, if societal interests that have a stake in economic ties with a potential adversary also possess access and effective voice, as is the case in liberal democracies, then economic interdependence will be likely to constrain the country's political-military strategy; however, if societal interests with an interest in economic ties with the adversary do not have access and effective voice, as is the case in authoritarian regimes, then economic interdependence will have little impact on the country's political-military strategy.

12. Logit estimation is analogous to linear regression but is an appropriate statistical tool for dichotomous dependent variables. 13. Specifically, these results are nearly identical to those reported in Oneal and Russett 1999, table 2. We were able to replicate their results exactly using traditional logit analysis. The use of the King and Zheng rare-events-corrections procedure yields only very minor changes in the coefficients. 14. All marginal effects were calculated with peace years set at 0, the modal value in the data set.

REFERENCES Angell, Norman. 1913. The Great Illusion: A Study of the Relation of Military Power to National Advantage. New York: G. P. Putnam's Sons. Barbieri, Katherine. 1996. Economic Interdependence: A Path to Peace or a Source of Interstate Conflict? Journal of Peace Research 33:29–49. ______. 1998. International Trade and Conflict: The Debatable Relationship. Paper presented at the 39th annual convention of the International Studies Association, March 18–21, Minneapolis. Barro, Robert. 2000. Democracy and the Rule of Law. In Governing for Prosperity, ed. Bruce Bueno de Mesquita and Hilton L. Root. New Haven: Yale University Press. Bueno de Mesquita, Bruce, James D. Morrow, Randolph M. Siverson, and Alastair Smith. 1999a. An Institutional Explanation of the Democratic Peace . American Political Science Review 93:791–807. ______. 1999b. Policy Failure and Political Survival: The Contribution of Political Institutions. Journal of Conflict Resolution 43:147–61. Caves, Richard, Jeffrey Frankel, and Ronald Jones. 1999. World Trade and Payments. Reading, MA: AddisonWesley. Page 58 → Doyle, Michael. 1997. Ways of War and Peace: Realism, Liberalism, and Socialism. New York: W. W. Norton. Edwards, Sebastian. 1993. Openness, Trade Liberalization, and Growth in Developing Countries. Journal of Economic Literature 31:1358–93. ______. 1998. Openness, Productivity, and Growth: What Do We Really Know? Economic Journal 108:383–98. Fiorina, Morris P. 1978. Retrospective Voting in American National Elections. New Haven: Yale University Press. Grossman, Gene M., and Elhanan Helpman. 1994. Endogenous Innovation in the Theory of Growth. Journal of Economic Perspectives 8 (winter): 23–44. Kant, Immanuel. 1983. Perpetual Peace and Other Essays on Politics, History and Morals. Trans. Ted Humphrey. Indianapolis: Hackett Publishing. Kinder, Donald, Gordon S. Adams, and Paul W. Gronke. 1989. Economics and Politics in the 1984 American Presidential Election. American Journal of Political Science 33:491–516. Kinder, Donald, and D. Roderick Kiewiet. 1981. Sociotropic Politics: The American Case. British Journal of Political Science 11 (June). King, Gary, and Langche Zeng. 2001. Explaining Rare Events in International Relations. International

Organization 55:507. Lake, David. 1992. Powerful Pacifists: Democratic States and War. American Political Science Review 86:24–37. Lawrence, Robert Z., and David E. Weinstein. 1999. Trade and Growth: Import-Led or Export-Led? Evidence from Japan and Korea. Working paper 7264, National Bureau of Economic Research. Available at . Lewis-Beck, Michael S. 1990. Economics and Elections: The Major Western Democracies. Ann Arbor: University of Michigan Press. Long, Ngo Van, and Kar-yiu Wong. 1997. Endogenous Growth and International Trade: A Survey. In Dynamics, Economic Growth, and International Trade, ed. Bjarne S. Jensen and Kar-yiu Wong. Ann Arbor: University of Michigan Press. Magee, Stephen P., William A. Brock, and Leslie Young. 1989. Black Hole Tariffs and Endogenous Policy Theory: Political Economy in General Equilibrium. Cambridge: Cambridge University Press. Magee, Stephen P., and Leslie Young. 1987. Endogenous Protection in the United States, 1900–1984. In U.S. Trade Policy in a Changing World Economy, ed. Robert M. Stern, 145–95. Cambridge: MIT Press. Mayer, Wolfgang. 1984. Endogenous Tariff Formation. American Economic Review 74:970–85. Morse, Edward L. 1970. The Transformation of Foreign Policies: Modernization, Independence, and Externalization. World Politics 22:371–92. Oneal, John R., Frances H. Oneal, Zeev Maoz, and Bruce M. Russett. 1996. The Liberal Peace: Interdependence, Democracy, and International Conflict, 1950–85. Journal of Peace Research 33:11–28. Page 59 → Oneal, John R., and Bruce M. Russett. 1997. The Classical Liberals Were Right: Democracy, Interdependence, and Conflict, 1950–85. International Studies Quarterly 41:267–94. ______. 1999a. Assessing the Liberal Peace with Alternative Specifications: Trade Still Reduces Conflict. Journal of Peace Research 36:423–42. ______. 1999b. The Kantian Peace: The Pacific Benefits of Democracy, Interdependence, and International Organization, 1885–1992. World Politics 52:1–37. Paine, Thomas. 1956. Letter to Abbe Raynal. In The Anglo-American Tradition in Foreign Affairs: Readings from Thomas More to Woodrow Wilson, ed. Arnold Wolfers and Laurence W. Martin, 126–38. New Haven: Yale University Press. Papayoanou, Paul. 1996. Interdependence, Institutions and the Balance of Power: Britain, Germany, and World War I. International Security 20:42–76. Przeworski, Adam, and Fernando Limongi. 1997. Modernization: Theories and Facts. World Politics 49:155–84. Romer, Paul M. 1986. Increasing Returns and Long-Run Growth. Journal of Political Economy 94:1002–37. ______. 1994. The Origins of Endogenous Growth. Journal of Economic Perspectives 8 (winter): 3–22. Russett, Bruce M., and John R. Oneal. 2001. Triangulating Peace: Democracy, Interdependence, and International Organizations. New York: W. W. Norton.

Russett, Bruce M., John R. Oneal, and David R. Davis. 1998. The Third Leg of the Kantian Tripod for Peace: International Organizations and Militarized Disputes, 1950–1985. International Organization 52:441–67. Yang, C. C. 1995. Endogenous Tariff Formation under Representative Democracy: A Probabilistic Voting Model. American Economic Review 85:956–63.

Page 60 →

Internationalization, Coalitions, and Regional Conflict and Cooperation Etel Solingen In their introduction Edward D. Mansfield and Brian M. Pollins identify different actors in the drama of interdependence and conflict, cast at different levels of analysis: supranational, state, and subnational. This chapter explores a subnational category, the nature of domestic coalitions, as an important yet relatively overlooked variable.1 In particular, it explains how internationalization affects domestic coalitions and how, in turn, the nature of resulting coalitions affects conflict and cooperation at the regional level. Different regional orders form as a product of the distribution of different coalitions throughout a region. Although pivoted on coalitions as key actors, the full argument developed here cuts across levels of analysis, as will be clear from the next section. On the independent variable side, for all of the exploding concern with internationalization (and globalization) in the international relations literature, internationalization per se has permeated the analysis of interdependence and conflict only to a limited degree. As a working definition, I shall refer to internationalization as involving the expansion of global markets, institutions, and norms, a process progressively reducing the purely domestic aspects of politics everywhere.2 On the dependent variable side, studies of interdependence and conflict have largely ignored regions as an important category, particularly when one considers the proliferating interest in regions in the overall international relations literature. The first section provides a brief overview of the conceptual framework.3 The second section specifies the temporal and spatial boundedness of claims. The third section outlines possible extensions and the research agenda ahead. Page 61 → The fourth section places the coalitional argument within the literature on interdependence and conflict, highlighting convergences and divergences with classical liberalism.

The Theoretical Framework in Brief: Internationalization, Domestic Coalitions, and Regional Behavior What is the impact of internationalization on the regional behavior of states? The argument is in two steps, beginning with “second-image-reversed” effects, identifying how internationalization affects different domestic constituencies.4 The second step examines how political entrepreneurs, within a given institutional context, logroll across constituencies differently affected by internationalization and hence holding different preferences over its advance. Out of this process emerge three ideal-typical coalitions endorsing different ideal-typical grand strategies. The policies derived from grand strategies then feed back into international (second-image) processes but also affect domestic coalitional cleavages. Regarding the first step, ongoing research suggests that there is no simple formula for estimating the precise domestic political impact of internationalization. We are still—paraphrasing Clausewitz—under the “fog of internationalization.” Our understanding of the political impact of expanding global markets, norms, and institutions remains open-ended. Nonetheless, as a first cut, we may begin by mapping domestic political constituencies according to their hypothesized positions regarding internationalization.5 Internationalization involves increased openness to international markets, capital, investments, and technology but also to an array of political and security regimes, institutions, and values. Political entrepreneurs rely on material and ideal aspects of internationalization and on the available institutional context (democratic or otherwise) to broker coalitions across different constituencies, logrolling or exchanging the mutual rights of partners to seek their most valued preference. Three ideal-typical coalitions tend to form: internationalizing and backlash (of which pure forms are hard to find in the real world) and hybrid.6 Entrepreneurs logrolling internationalizing coalitions aim at attracting the beneficiaries or potential beneficiaries of internationalization, such as mobile capital, export-intensive sectors and firms, highly skilled labor employed in

competitive industries or firms, analysts and professionals oriented toward an open global economic and knowledge/technology system, consumers of imported products, farm workers and small landowners with an eye on higher producer prices for agricultural exports, ethnic and religious groups posed to Page 62 → thrive from greater openness, and bureaucracies entrusted with reform (independent central banks, finance ministries, managers of export-processing zones). Entrepreneurs crafting backlash coalitions attract groups that resist internationalization, including import-competing firms and banks closely tied to the state; state-owned enterprises and banks; unskilled blue-collar and white-collar sectors; state bureaucracies rendered obsolete by reform; segments of the military and its industrial complex; and civic-nationalist, ethnic, and religious movements threatened by material and normative dimensions of internationalization. Entrepreneurs faced with most intractable political landscapes must logroll across diverse constituencies, leading to hybrid coalitions that are much less clear-cut in their composition than either of the two other ideal-types and that can bring together otherwise strange bedfellows. Given their unwieldy nature, these coalitions are also bound to be significantly more afflicted with distributional conflicts.

Coalitions differ in their preferences over domestic and international resource extraction and allocation and in their orientations toward regional and international behavior. Accordingly, each sponsors a grand strategy with synergistic effects across the domestic, regional, and global arenas. Grand strategies are modified by interactive effects with other coalitions in the region, as we shall see shortly. In principle, however, the grand strategy of internationalizing Page 63 → coalitions combines economic liberalization at home; cooperation and stability in the region; and the safeguarding of secure access to global markets, capital, investments, and technology.7 Central to these coalitions' economic program is the primacy accorded to macroeconomic stability and to international competitiveness. Macroeconomic stability reduces uncertainty, encourages savings, and enhances the rate of investment (including foreign). Why are these coalitions cooperative with their neighbors? Conflict-prone postures require the need to uphold them by mobilizing domestic resources for potential military conflict. Such mobilization often contributes to many of the weaknesses burdening the domestic political economy (from the standpoint of internationalizing coalitions) such as unproductive and inflation-inducing military investments and the protection of state-owned enterprises under a cloak of “national security.” Even where the circumstances compel them to invest in defense, internationalizers attempt to avoid overblown military budgets that increase government and payments deficits, raise the cost of capital, curtail savings and productive investment, deplete foreign-exchange coffers, induce overvalued exchange rates and currency instability and unpredictability, and distort the humanpower base. Naturally, these coalitions are also averse to regional instability that thwarts foreign investment. The strategy of backlash coalitions seeks to preserve inward-looking sectors, state industry, and ancillary militaryindustrial allocations; to resist external pressures for economic liberalization and intrusions on sovereignty; and to target internationalizing adversaries at home and abroad. Regional insecurity and competition is a natural side effect at best, and a dominant requirement at worse, of this grand strategy. Regional cooperation threatens backlash coalitions because it compels downsized military allocations and weapons complexes, erodes statist privileges, and devalues nationalist and confessional mythmaking as a political currency. Intransigent and uncertainty-inducing regional policies raise the propensity for conflict and the risks for foreign investors and may trigger denial of aid and technology. However, backlash coalitions thrive with such responses, at least in the short and medium terms, having an inherent affinity with import-substitution; classical populism; active states controlling prices; increasing nominal wages; overvaluing the currency to raise wages and profits in nontradedgoods sectors; protecting state and military enterprises; allocating credit at low interest rates; and dispensing rents to private firms by discriminating against competing imports through tariffs, controls, and multiple exchange rates. The grand strategies of hybrid coalitions bestride elements from the other two and are therefore less coherent and synergistic across the domestic, regional, and international arenas. Page 64 → A grand strategy becomes raison d'état once a certain coalition prevails politically as a function of its resources relative to the opposition and of the institutional context in which it operates. The strength of a ruling coalition

relative to its domestic challengers affects the extent to which its grand strategies are more pure or thinned versions of the ideal-type. A ruling coalition is also constrained externally—by the nature and strength of the coalition it faces across the border—in its ability to implement its favored grand strategy. Interactive effects are portrayed in figure 2, which depicts all possible coalitional combinations for a dyad A and B, differentiating between domestically weak and strong ruling coalitions. Dyads of strong internationalizing coalitions are expected to create more cooperative and peaceful regional orders than either mixed dyads or dyads of strong backlash coalitions. The darker the shaded areas in figure 2, the more cooperative the relationship can be expected to be. Cooperation declines as we move to mixed cells 2 and 3—where internationalizing and backlash coalitions face each other—and even more so in cell 4, which depicts relations between backlash dyads. Figure 2 can also portray the relationship between a coalition ruling state A (e.g., Jordan ca. the 1960s) and the regional coalitional cluster B that surrounds it (e.g., a heavily Nasserite group of Arab states). Taking note of the regional coalitional context, or of the coalitional center of gravity in a given region, provides more complete information about a coalition's likely behavior. This can also be thought of as a regional coalitional balance of power that favors either internationalizing, backlash, or hybrid coalitions. Where internationalizing coalitions prevail, higher and more extensive levels of cooperation can be expected than where backlash or mixed, competing internationalizing and backlash neighbors dominate a given region. Summing up the argument thus far, certain grand strategies and behavioral expectations are hypothesized on the basis of the composition and strength of coalitions both at home and in the region. Figure 3 recapitulates the argument's causal logic. Both coalitional strength at home and the regional coalitional balance of power are conditional or intervening variables, changing the relationship among coalitions, grand strategy, and regional outcomes. It is also possible to think of the policies a given coalition is likely to embrace as the first dependent variable and of the aggregate behavior of a specific regional coalitional cluster as a second dependent variable. However, the regional coalitional balance of power is not a mere resultant of the aggregate nature and strength of individual coalitions but also actively affects the policies, and indeed the very constitution, of domestic coalitions in any given state in the region. I turn to these dynamics next. Page 65 → Clusters of internationalizing coalitions (zones of peace in fig. 2) are characterized by extensive and intensive cooperation, that is, cooperation across various issue-areas and with a deepening level of commitment. In the realm of security, domestic considerations of political survival drive economic rationalization—and military downsizing—as much as external factors. Therefore, there is a virtual built-in guarantee that like-minded internationalizing coalitions will be, ceteris paribus, reluctant to defect through militarized strategies or to exacerbate territorial or ethno-religious disputes. The potential for armed conflict and extensive military buildups threatens the economic and political fundamentals—fiscal conservatism; macroeconomic, political, and regional stability; global access—that an internationalizing strategy requires. Furthermore, the mutually reinforcing domestic and regional incentives to assuage conflict also reduce transaction costs, making it easier to reach agreements and lowering the need to monitor behavior, punish noncompliance, or improve Page 66 → information. Reciprocal incentives to cooperate are highest where internationalizing strategies are more fully in place and their political agents more strongly entrenched, conditions that are of, course, interrelated. Thus, internationalizing coalitions beget the conditions for self-sustained, rather than externally imposed, regional cooperation. These features also make concertlike multilateralism more efficient and dense institutionalization less necessary (but certainly not impossible). This is also the case for economic cooperation, where regional relations are forged for the purpose of boosting the internationalizing model at home, disabling the opposition, and lubricating external ties to the global political economy. The need to harmonize standards and legal and administrative infrastructures expands institutional links and interdependence. This may lead to integrative efforts and trade-creating schemes that emphasize positive regional and global externalities, compatible with “open regionalism.” An absolute increase in regional trade and investment often obtains. However, economic integration is not required for cooperation to endure. Indeed, even

in the absence of dramatic economic benefits from bilateral economic interactions, internationalizing coalitions find it beneficial to sustain cooperation. The underlying logic of cooperation is global access, subordinating varying forms of regional arrangements to that logic. Where the credibility of internationalizing Page 67 → forces is questionable (e.g., in Latin America's Southern Cone in the early 1990s) formal regional arrangements in the direction of freer trade can signal a more believable commitment to extraregional investors. Backlash coalitions thrive on regional competition, self-reliance, military prowess, upholding sovereignty, and national or confessional purity.8 Instead, integrative efforts have the potential of forcing many of these coalitions' natural political constituencies out of business. Regional cooperation weakens their rationale for maintaining high levels of unilateral military preparedness (often absorbing 15 to 25 percent of gross domestic product [GDP], or over three times the world's average). The presence of neighboring backlash coalitions heightens the “balance of threat” for these coalitions no less—and often even more—than the presence of internationalizing coalitions in the region. The personalistic element in these coalitions is an important source of myths and tends to fuel the slippery slope into war zones (fig. 2) with the bloodiest military confrontations. Nuclear weapons are ideal technological allies of these coalitions because such programs compel mammoth scientific, technological, industrial, and bureaucratic investments that are beyond formal budgetary oversight, sometimes even under democratic rule. The actual or imaginary output of this large-scale provider of rents—a nuclear weapon—itself becomes a powerful source of myths. In the realm of economic relations, cooperation between backlash coalitions, such as projecting importsubstituting schemes into the region, results in trade-diverting schemes, with negative externalities for third actors in the region and beyond. Private and state monopolies threatened with competition from regional counterparts frequently resist lowering trade barriers and strive to maintain a regional system of competing, rather than complementary economies. A backlash coalition in control of a large regional power can avail itself of the imperial commercial strategy so fittingly described by Hirschman ([1945] 1980), importing goods for which there are no substitutes at home and those required for its war machine while actively deindustrializing weaker potential competitors and making neighbors dependent on industrial exports in which the hegemon enjoys a monopolistic position. An imperial commercial strategy allows a large backlash regional power to maximize economic profit, military power, and regional influence, all of which sustain and reproduce its ruling backlash coalition. The basis for a stable cooperative framework is absent in “mixed” regions, where the policies of alternative coalitions create negative externalities for, or impose costs on, their rival neighboring coalitions. Mixed regional clusters are not likely to develop extensive trade relations, given the inherent economic closure Page 68 → and adversarial policies of backlash regimes and the primarily global orientation of internationalizing ones. This does not preclude instances of unilateral or mutual dependence on markets and raw materials, where the benefits of trade are carefully channeled to the respective coalitional beneficiaries, backlash or internationalizing. Mixed regions avoid high values at either end of the conflict-cooperation spectrum, yielding zones of more contained conflict than backlash clusters (see fig. 2). Hybrid coalitions, diluted grand strategies, and mixed regional clusters create open-ended regional orders that elude extensive cooperation or warfare. However, in mixed orders, regionally hegemonic coalitions can have a strong effect on the fate of domestic coalitional balances, generally shifting them toward their own coalitional type. Summing up, this framework also outlines the conditions under which its expectations would be under stress.9 This would be the case, for instance, when coalitions composed of backlash constituencies that are relatively strong vis-àvis their domestic opposition embrace and implement an internationalizing grand strategy, particularly in a region dominated by backlash coalitions. Another anomaly would entail the reverse, when strong internationalizing coalitions advocate a backlash strategy in the midst of an internationalizing regional balance of power. The toughest challenges emanate from strong internationalizing clusters waging war and from strong backlash clusters creating and maintaining zones of durable, intensive, and extensive cooperation. The framework has several conceptual advantages. First, on the independent variable side, it relies on a single explanatory category—coalitional type— as a pivot for understanding most relevant considerations of late twentieth-and early twenty-first-century world-time: internationalization, nationalism, ethno-confessionalism, and

the role of international institutions. Second, a reliance on coalitions helps transcend dated level-of-analysis categories by linking global and subnational processes, hinged on coalitional grand strategies, to explain regional outcomes. Indeed, the framework helps resolve some of the intractability involved in defining regions: coalitions define regions, subsuming geographical ontology to the political requirements of their grand strategy. Third, coalitions ground evolving regional relations in a dynamic framework, able to explain both continuity and change. Coalitions develop, blossom, disintegrate, and reconstitute, generating a changing regional context, qualities that advantage the approach over cognitive and structural (neorealist, interdependence, or world-systems) theories of interstate behavior. Fourth, the framework clarifies and qualifies the relative weight of theories such as bureaucratic politics, the military-industrial complex, and electoral cycles in explaining patterns of defense expenditures. Fifth, on the dependent Page 69 → variable side, it helps explain outcomes that span the conflictcooperation spectrum. The nature of coalitions also offers a helpful grip on the sources of state compliance with international institutions and on the puzzle of why certain states, given their coalitional makeup, are more receptive to great power inducements than others. Sixth, the framework can be used to understand both internal (domestic intercoalitional) as well as external (interstate) competition. In that sense, it can be marshaled to explain the roots of civil wars affected by coalitional cleavages related to internationalization and their potential regional implications. It thus rejects reductionist (economistic, culturalist) interpretations of ethnic conflict by exploring the conditions under which ethno-confessional and political-economic cleavages intersect. Studies of ethnic and civil conflict can benefit from paying more attention to the broader regional and international contexts and their implications for coalitional choices. Seventh, it provides a unifying framework for comparing regions, flexible enough to accommodate wide variation in state-society relations and political institutionalization. It transcends regional “exceptionalisms,” drawing together the common political experience of states undergoing different phases of economic and political transformation. Furthermore, it integrates great power theorizing on the security implications of coalitional forms (Snyder 1991) with the balance of the international system, avoiding the need for a separate theory of “Third World” behavior.

Boundedness of Claims The first extensive application of this argument involved a qualitative study of three regions—the Middle East, the Southern Cone of Latin America, and the Korean peninsula—involving dozens of states and successive coalitions.10 This yielded a rather significant number of cases (coalitions, dyads, clusters), involving varied coalitional forms succeeding each other within the same state and varied coalitional forms interacting with each other in a given region. The time frame—1950s to 1990s—allowed for an extensive temporal realm over which the behavior of successive coalitions could be examined. This led to a rather large domain of cases spread over time and space that included a sizeable part of the industrializing world. This domain has been far less favored in international relations theory than the study of great powers, yet it provides a particularly useful ground to examine arguments related to interdependence and vulnerability to global forces. At the level of regions, the selection allowed the possibility of holding important variables constant (e.g., levels of security threat) so that the impact of domestic coalitions could be ascertained with Page 70 → greater confidence. Regional choices also involved cases “least likely” to support coalitional dynamics as the explanatory variable. The analysis was qualitative, relying on process tracing, focused comparisons, and in-depth case studies.11 The findings provided preliminary support for the expected relationship among the composition of ruling coalitions, the nature of grand strategies, and the actual behavior of different coalitions in their regions. Internationalizing coalitions deepened trade openness, expanded exports, attracted foreign investments, restrained military-industrial complexes, eschewed weapons of mass destruction, deferred to international economic and security regimes, and strove for regional cooperative orders that reinforced those objectives.12 Backlash coalitions restricted or reduced trade openness and reliance on exports; curbed foreign investment; built expansive military complexes; developed weapons of mass destruction; challenged international regimes; and exacerbated civic-nationalist, religious, or ethnic differentiation within their region. Values for hybrid coalitions were largely in-between, reflecting intermittent efforts—neither forceful nor coherent—in either direction. The strength of coalitions relative to their domestic and regional rivals affected the effective operationalization of grand strategy, as expected. Where competitors loomed large on the political survival of coalitions, policies were

more diluted. Different mixes of weak and strong coalitions throughout a region appeared to create and reproduce typical regional orders and, conversely, were affected by them. Backlash clusters produced higher levels of military conflict than either internationalizing or mixed clusters, from the Korean War; to successive Arab-Israeli and Indo-Pakistani wars; to Iran-Iraq, Kampuchea and Vietnam, and Somalia and Ethiopia, among others. The logic of backlash strategies, particularly extensive militarization and risky posturing, often pushed backlash entrepreneurs—sometimes reluctantly—into armed conflict. Backlash leaders often read war opportunities too favorably, miscalculating the resolve of neighboring and extraregional actors anathematized by their policies, as Kim Il Sung in the 1950s, Gamal A. Nasser in 1967, Leopoldo Galtieri in 1982, and Saddam Hussein in 1990. Where internationalizing coalitions gained a critical mass they largely avoided militarized conflict and mutually adjusted on lingering territorial and other disputes. Internationalizing clusters (e.g., Latin America's Southern Cone) placed direct challenges to remaining backlash coalitions in their region, often overturning the coalitional balance within residual backlash states. Backlash clusters operated in the same way (as with the dominance of Nasserite-style coalitions in the Arab world in the 1960s). Mixed dyads and clusters (of internationalizing and backlash ruling coalitions in different states) precluded extensive and intensive Page 71 → cooperation but also yielded lower levels of militarized conflict than backlash war zones. With respect to the particular evolution of different regions, the framework provided a persuasive account of four decades of Middle East wars and enduring rivalries in the Arab-Israeli, inter-Arab, and Arab-Iranian arenas. It helped elucidate why regional economic barriers among Arab states never receded, why their integrative political schemes never lasted beyond declarations, why regional institutions such as the Arab Common Market existed largely on paper, and why militarized threats were a constant. It also shed light on the conditions leading to a cooperative Arab-Israeli breakthrough in the early 1990s but a serious reversal since, given backlash responses to the Oslo process and to internationalizing strategies more generally. Coalitional analysis also captured the evolving texture of regional relations in the Southern Cone over five decades. Backlash and hybrid coalitions in Argentina and Brazil presided over an era of guarded regional relations, limited economic exchange, and risky nuclear competition, even where initial conditions—over one hundred years free of war— would have favored deeper cooperation. An entrenched backlash coalition in Argentina spearheaded backlash strategies in the region, including nuclearization, while unleashing military crises, mobilizations, and the Falklands/Malvinas war. An internationalizing revolution in the 1990s ushered in a radically new Argentine strategy including more effective economic integration (Mercado Común del Sur [MERCOSUR]) and denuclearization. Finally, the framework helped explain the outbreak of the Korean War, subsequent shifts away from war in Korean North-South relations, cooperative overtures, and alternative nuclear postures in the 1970s–80s but converging ones in the early 1990s. Some findings, while consistent with the framework, appear counterintuitive for other approaches. First, a state's (or region's) ethnic and religious diversity per se does not seem an efficient indicator of its interstate-conflict proneness. The coalitional pattern available to ethno-religious entrepreneurs, the relative strength of the coalition they join, and the identity and strength of the coalitional neighbor or cluster they face at the borders provide a more proximate indication of likely behavior. Thus, regions ridden with ethnic and religious cleavages—such as the Association of Southeast Asian Nations (ASEAN)—may breed more cooperation under internationalizing rule than more ethnically, linguistically, and religiously homogeneous regions under backlash rule (the Southern Cone and the Middle East for decades). Second, a coalitional perspective proves its utility even for a least-likely dependent variable: nuclear behavior. Whereas the dominant neorealist literature assumes a strong conceptual bond between security dilemmas and nuclear behavior, the Page 72 → evidence points to coalitional dynamics as a more powerful predictor of nuclear policies and actions. Remarkably, of all states (beyond the original five) that have considered a nuclear option in recent decades, not one endorsed denuclearization—fully and effectively—under a ruling backlash coalition. Nuclear efforts in the direction of acquiring nuclear weapons were frequently taken in tandem with backlash policies, from North Korea to Iran, Iraq, Libya, India, and Pakistan. Instead, only internationalizing coalitions undertook effective commitments to denuclearize, from Taiwan and South Korea to Egypt, South Africa, Brazil, and Argentina. A subsequent study analyzed aggregate data from South Asia, East Asia (excluding China), Southeast Asia, the Middle East, and the Southern Cone, involving over ninety-five coalitions in nineteen states between 1948 and

1993.13 The independent variable was the makeup of the domestic coalition (backlash, internationalizing, hybrid). The dependent variable was grand strategy (coalitional policy preferences and behavior, particularly vis-à-vis the region). A second dependent variable was regional orders or outcomes (zones of stable peace, war zones, contained conflict). The three ideal-typical coalitions were found to differ significantly in their actual international behavior. Building on yearly aggregated data on trade openness (TO = imports plus exports as percentage of GDP) for each coalition (N = 572), the analysis identified a mean TO of 88 percent for internationalizing coalitions, 38 for backlash, and 52 for hybrid. Mean internationalizing TO values were more than double those of backlash and less than double those of hybrids.14 Internationalizing coalitions were found to increase their TO by 1.9 percent per year and hybrid ones by 1.1 percent whereas TO for backlash coalitions decreased by 0.23 percent per year on average. Mean yearly export/GDP ratios for internationalizers neared 44 percent, almost three times the backlash mean of 16 percent and nearly double the hybrid mean of 25. Mean changes in exports/GDP reflected an increase of 1.3 percent yearly for internationalizers and 0.45 for hybrids but an average yearly decline of 0.27 for backlash coalitions. Mean yearly foreign direct investment (FDI) inflows for internationalizers was $712 million, slightly less for hybrids ($655M), and only $38M for backlash.15 Notably, internationalizers increased FDI inflows by $110M yearly on average and hybrids by $33M whereas inflows decreased by $25M yearly under backlash coalitions. There are significant differences in mean military expenditures as a percentage of GDP (MILEX/GDP), averaging over twice as much for backlash (9.53 percent) than for internationalizers (4.59), with hybrids in the middle (7.58).16 MILEX/GDP declined by 0.15 percent yearly on average for internationalizers and by .07 percent for hybrids but rose by 0.62 for backlash coalitions. Data on Page 73 → MILEX as a percentage of centralgovernment expenditures reflects a similar pattern. Longitudinal comparisons of successive coalitions of different types in a single state, for all of these variables, support these findings. Regionally, backlash coalitions were found to have initiated over thirty international crises, hybrid coalitions about seventeen, and internationalizing coalitions only two.17 In other words, of all the international crises initiated under this period by the countries under study, backlash coalitions account for 62 percent, hybrids for 34, and internationalizers for 4. Evidence dealing with regional behavior— cooperative/conflictive—also provides support for the hypothesized synergies across all pillars of a “grand strategy,” domestic, regional, and global. The results are particularly significant if one takes into account that the quantitative data set used (the only one effectively available for this purpose) posed a particularly hard test for discerning coalitional effects. Instead, the systematic inclusion of post-1993 data would have arguably sharpened the effects of internationalizing coalitions and policies. Yet differences in coalitional behavior are clear even under a more inauspicious era for internationalizing strategies, prior to 1989. In sum, regional orders can certainly not be reduced to the sole outcome of coalitional balances and profiles, but these provide a critical lead regarding likely levels and forms of regional conflict and cooperation. Regional cooperation, as wars, have more than one source, and internationalizing coalitions may not even be necessary for at least some cooperation to come about. Yet the finding that internationalizing coalitions—particularly strong ones—quite systematically beget intensive (deep) and extensive (in scope) regional cooperation suggests that they tend to create near-sufficient conditions for the emergence of zones of stable peace. Clearly, these are no more than probabilistic expectations that can be falsified. Many of the claims seem valid for the particular world-time characterizing the latter part of the twentieth century and the dawn of the twenty-first. This world-time is epitomized by tensions between a globalizing political economy and backlash challenges to it, between a more institutionalized global political order and the resistance to it, between a pluralist (multicultural) political approach to human diversity and exclusivist ethno-religious counterparts, and between regionally differentiated versus globally homogeneous solutions to the opportunities and predicaments posited by the first three. In particular, Mansfield and Pollins call for greater attention to the influence of international institutions on conflict, a concern directly addressed by this framework. International economic, security, and other normative structures and institutions purposefully, but also unintendedly, influence the domestic coalitional interplay, at times strengthening internationalizing coalitions, at others empowering Page 74 → their rivals. Different international contexts privilege the interests of one coalition over another. The early Cold War era provided a more supportive global structure for backlash coalitions, from economic protection to

militarization and regional conflict, helping to explain the relative scarcity and weakness of internationalizing coalitions at the time. Even some United Nations agencies provided a convenient site for the coordination, elaboration, and diffusion of backlash policies. Toward the 1990s international economic structures and institutions appeared to favor internationalizing coalitions worldwide to an extent never evident before. The new structures of international production and financial intermediation, new terms of trade and investment, and structural-adjustment policies favored internationalizing groups and strengthened the institutions of reform (particularly finance ministries, central banks, export-promotion bureaus, and export-processing zones).18 These same structures and institutions heavily constrained state entrepreneurship and classical populist and military-industrial constituencies and demanded greater budgetary accountability and transparency. At the same time, the new international environment involved new risks, unintended consequences, and severe crises. Structural adjustment, stabilization programs, and economic liberalization implied painful transitions, recessions, unemployment, and reduced subsidies and investments in infrastructure, all of which strengthened resistance to internationalization. Where internationalizing coalitions failed to distribute the costs and benefits of reform more equitably, their political support declined. Shock-style therapy without safety nets or social insurance undermined reform and made backlash politics, nationalism, and ethno-confessional extremism ripe as a solution to the predicaments of internationalization. Notably, where integration into the world economy evolved in tandem with a more egalitarian income distribution, as in Taiwan and South Korea, internationalizers rendered their grand strategy least reversible. Protected markets in the industrialized world pose no less of a challenge for internationalizing coalitions. Although international institutions can play critical roles in sustaining or changing domestic coalitional balances, they are not invariably successful. Sometimes their myopic intervention—as in Southeast Asia most recently— can lead to the unintended consequence of revitalizing backlash forces, as in Indonesia. Yet domestic coalitional competition is not reducible to international manipulation. Coalitions purposefully rely on convenient global structures and institutions but are not wholly defined by them. Notwithstanding the differential domestic conversion of global influences in each country, it is Page 75 → important to recognize how free capital mobility has deepened international effects on domestic coalitional balances in an unprecedented way, rewarding or punishing coalitional policies and behavior—at home, in the region, and vis-àvis the world—virtually overnight.

The Agenda Ahead The coalitional framework accommodates a wide range of methodological preferences, yielding testable propositions that can be examined through aggregate data or qualitative analysis. It allows for a focus on the longitudinal evolution of single countries throughout alternating coalitional forms, the evolution of dyads of states ruled by different coalitional forms over time, comparisons between coalitionally homogeneous and heterogeneous dyads or regional clusters, and various other possibilities. The results of a statistical study were reported earlier. Some have explored the framework's broader applicability to cases in Central America, East Asia, South Africa, Eastern Europe, and even Britain and the United States historically, while others have suggested its potential relevance to understanding Russia, the Balkans, and advanced industrial democracies such as Austria, Denmark, and Japan.19 More generally, Kahler (2000, 672) has suggested that these coalitional forms may also explain varying positions vis-à-vis the legalization of regional institutions or the extent to which regional institutions display heightened obligation, greater precision in rules, and delegation of rule interpretation and enforcement to third parties. In his view, internationalizing coalitions may be more prone to use legalization as a means for regional stability whereas backlash coalitions may be more likely to resist legalization due to the high sovereignty costs or loss of autonomy. Another application examined Southeast Asia, tracing the activities of ASEAN to the coalitional characteristics of its member states. The economic and political crisis that afflicted this region in the late 1990s offered an opportunity to submit hypotheses derived from coalitional analysis to an empirical test. ASEAN as an institution can be regarded as a product of an internationalizing cluster favoring domestic and regional, political and economic, stability and global access. Internationalizing coalitions and strategies are quite vulnerable to domestic and international sources of instability. Furthermore, when threatened by backlash challengers at home and in the

region, their tendency to cooperate regionally can erode. The 1997 crisis introduced the possibility of just such developments, leading to four possible scenarios. Page 76 → 1. Backlash political forces grow stronger, replacing internationalizing coalitions in power and reversing ASEAN's regional cooperative patterns typical of earlier times. 2. Internationalizing coalitions rebound from the worst economic and political debacle in thirty years and stay ASEAN's cooperative course. 3. Backlash coalitions prevail throughout the region, but the quality of regional cooperation is maintained and grows even deeper and wider in scope (encompassing new issue-areas). 4. Strong rebounding internationalizing coalitions reverse their earlier cooperative pattern by, for instance, waging a war. The first two scenarios are compatible with a coalitional account, but the latter two would falsify it or point to anomalies. On a preliminary basis (as of early 2002), it might be argued that despite the mobilization of nationalist and backlash forces, ruling coalitions remained, on the whole, more representative of constituencies accepting internationalization. These coalitions retained ASEAN's fundamentally cooperative relations characteristic of the precrisis era, even if they navigated through serious challenges in bilateral relations and multilateral collective action on issues of economic cooperation, expansion, intervention, and security. However, financial and capital account liberalization has clearly induced greater vulnerability, and the full distributional effects and political implications of the economic crisis may not be evident for some time. No linear or irrevocable progression toward internationalization or regional cooperation can be assumed, nor can scenarios 3 and 4 be discounted. Several important challenges for a better specification of the coalitional argument remain. First and foremost, our collective understanding of the impact of internationalization on domestic constituencies must be improved. Purely international political economy (IPE) approaches often blackbox important agents in the battle for internationalization, such as the military and ethnic groups. As Eichengreen (1998, 355) argues, the assumption that individuals care only about economic welfare may be convenient for factoral and sectoral models of foreign economic policy. However, these fail to address individual and group concerns with security and other values, which makes the mapping of policy to interests and vice versa far more difficult.20 We lack integrated understandings of responses to both material and normative aspects of internationalization that might explain different coalitional landscapes.21 Gourevitch (1986) identified a second difficulty in predicting coalitional formations in second-image-reversed models: leadership and entrepreneurship. Since political entrepreneurs can broker coalitions on the basis of imputed Page 77 → rather than real effects of internationalization, pivoting the analysis on them may be particularly fruitful. It may also help resolve some problems of aggregation (Kahler 1999, 291) and provide a better handle on top-down versus bottom-up interpretations of coalition formation and impact. A third hurdle relates to the correspondence between real-life coalitions and suggested ideal-types, which are too crude to capture all empirical forms fully, and the related problem of estimating coalitional strength.22 Stronger versions of coalitions are more easily identifiable, facilitating the analysis of their behavioral effects.23 Thresholds of coalitional strength enabling the implementation of a given strategy are far harder to estimate. However, even hybrids are associated with specified outcomes, such as largely diluted and unstable grand strategies. Finally, institutions (states, regime-types, electoral and party systems, labor-markets institutions) play important roles in the emergence, composition, strength, grand strategy, and regional behavior of coalitions.24 All three coalitional ideal-types can thrive in nondemocratic and cartelized systems, and all can also benefit from democratization, particularly when they are in the opposition. However, the affinity between backlash populistnationalism and transitional democracy can pose important barriers for internationalizing entrepreneurs.25 On the basis of theories linking levels of economic development or trade openness to the onset of democracy one might expect a certain synergy between democracy and internationalizing coalitions, as Mansfield and Pollins suggest in their introduction.26 Peceny and Stanley (2001, 156) confirm the higher proclivity of Central American internationalizers—relative to their backlash competitors—to adopt democracy. This synergy could be the product

of a particular world-time, however, when international institutions and great powers create strong incentives favoring these twin reforms. The incidence of democratic internationalizers was much lower during the Cold War, but so was the frequency of backlash democracies.

The Argument's Place in the Literature on Interdependence and Conflict How is this framework related to classical liberalism? The framework is generally compatible with some core assumptions of liberal theory but also departs from it in several important respects.27 A most important contribution to this literature, sensitive to political, economic, and historical dynamics, is that of Rosecrance (1986, 1999), who maps the stages in the evolution from territorial to trading and, more recently, to virtual states. Leaders of virtual states “must encourage domestic factors of production to seek out areas of the highest Page 78 → return”; must use monetary and fiscal tools that can secure stable currencies, low interest rates, low inflation, and minimal government deficits; and must develop a regulatory environment that administers commercial interests impartially and protects foreign factors of production. These objectives define a central pillar of the grand strategy of internationalizing coalitions as defined previously. Not many contributions to classical interdependence theory stipulate how trading states come about in response to domestic and international incentives. Beyond adhering to the unspecified notion that prosperity begets peace, economistic theories of integration—as Higgott (1995, 78, 94) notes—have “little or no theory of politics to sustain them” and therefore ignore competing rationalities at the domestic level. Not unsurprisingly, studies on the relationship between trade flows and conflict diverge in their conclusions, linking commerce to more military conflict or less conflict or finding no relationship at all. These studies rarely disaggregate the domestic arena to examine competing interests and preferences. There are important exceptions. Snyder (1991) resorts to the impact of industrialization, cartelization, logrolling, and myth-making to provide a domestic logic for imperial (over)expansion. Rosecrance and Stein (1993) examine an array of domestic influences on great powers' grand strategy, including interest groups, social ideas, the character of constitutions, and economic constraints.28 Most coalitional analyses of grand strategy have been largely circumscribed to the great powers, perhaps understating their broader applicability. The coalitional perspective advanced in this chapter departs from the classical liberal tradition and from several more recent formulations in some important ways. 1. It rejects the universally pacifying, cooperative effect of commerce assumed by interdependence theory.29 The framework assumes no linear, inevitable, or irrevocable progression toward liberalization, internationalization, or regional cooperation. The global political economy and its cycles induce contingencies in domestic politics that make predictions on coalitional outcomes quite tentative.30 Accordingly, backlash and confessional forces could be more resilient than “liberal optimists” at the dawn of the twenty-first century may recognize (Venezuela-style reversals, Belarus, Indonesia?). Indeed the coalitional framework highlights the vulnerabilities (Solingen 1998, 47, 284–89), not just the advantages, embedded in a coalition's efforts to internationalize. Hence, internationalizing coalitions must pay close attention to compensating mechanisms designed to minimize adjustment costs and enhance the longer-term benefits from internationalization. Page 79 → 2. At the same time, rather than assuming a purely economistic aggregate calculus of costs and gains from trade, war, and cooperation, a coalitional perspective examines the domestic political foundations of a coalition's regional and global policies. It thus underscores internal political opportunities, often ignored in liberal theory at the expense of a focus on states' external “vulnerabilities.” Furthermore, coalitional entrepreneurs are motivated not only by extant but also by future opportunities and risks. 3. The coalitional perspective does not require that states become “fully modern industrial nations” before the incidence of war declines, only that strong—particularly redistributive —internationalizing coalitions overpower their challengers.31 Indeed, extant levels of trade openness, a traditional focus of commercial liberalism, are not a

wholly reliable indicator for predicting coalitional behavior. The inception of an internationalizing coalition, even at low levels of international economic openness, may launch a grand strategy that anticipates synergies across domestic and regional stability and global access. 4. Coalitional analysis does not dwell on the extent of economic bilateral/regional interdependence, the preferred causal variable, particularly in the empirical literature. The approach here emphasizes not dyadic interdependence but reliance on the world economy for markets, assistance, raw materials, technology, capital, and political support. In other words, coalitional grand strategies are responsive to nation-to-system or internationally oriented interdependence.32 Regional stability and a peaceful environment are important requirements to attract foreign investment, but the size and content of dyadic exchanges (trade, investments) matter less than how each state (ruling coalition) defines its relationship to the global political economy. 5. Regarding the demand for regional institutions, the preceding point makes it clear that this coalitional framework expects internationalizing regions to be generally cooperative but does not require regions to integrate economically or politically. Expectations about conflict and cooperation do not hinge on institutionalized regional interdependence, even if the extent of such interdependence increases as internationalizing coalitions prevail in a region (e.g., East Asia). The European Union (EU) path is not inherent or necessary. Internationalist dyads /clusters are not required to be enmeshed in formal bilateral or regional preferential trade or security agreements, but internationalizing strategies are compatible with regional institutional frameworks attuned to the strategy's global pillars. 6. As for the supply effect of regional institutions, neoliberal institutionalism, building largely on the experience of the EU, assigns institutions an important role in the promotion of cooperation, stability, and peace. A coalitional perspective Page 80 → involves a far more contingent understanding of their effects, expected to be mediated by the identity and strength of competing coalitions in the region. The latter are anticipated to play a more fundamental role in the texture of regional relations than extant institutional infrastructures. On the one hand, regional institutionalization in the Arab Middle East did not translate into cooperation. Institutions were rather a reflection of the (largely backlash) domestic coalitions that ruled the Arab world for decades. Another region with a crowded institutional space—Latin America—reveals more cooperative opportunities wasted than undertaken. Only a dramatic coalitional reversal in the 1990s infused old institutions with new life and created fresh, effective ones in both economics and security. The absence of war in this region precedes institutional frameworks by nearly a hundred years, devoiding institutions from any explanatory role in this regard either. On the other hand, the emergence of cooperation in the Arab-Israeli arena and in the Korean peninsula in the early 1990s can be much more readily traced to coalitional developments than to any institutions facilitating cooperation. Incipient institutions lagged after cooperation and, in the former case, collapsed with the breakdown of feeble internationalizing coalitions. In sum, a coalitional perspective is essential in understanding why certain institutions come into being, in whose interest they operate, when they are allowed to play a significant role, and how they can lag behind—rather than lead—a cooperative praxis.

NOTES I am grateful to Brian M. Pollins, Edward D. Mansfield, and other participants at the Mershon Center Conference on Interdependence and Conflict (September 14–17, 2000), as well as to participants in international relations seminars at Stanford/CISAC, the University of California at Berkeley, the University of Washington, the University of Virginia, the University of Wisconsin-Madison, Georgetown University, Cornell University, University of Utah, Singapore's Nanyang University, and the University of Southern California for their helpful comments. I would also like to acknowledge Matthew Evangelista, Miriam Elman, Stephan Haggard, David Lake, and John Odell for useful suggestions. 1. Most research on coalitions in international political economy does not address broader (noneconomic) international conflict. For a landmark analysis of coalitions and grand strategy, see Snyder 1991. 2. That more and more aspects of domestic politics are exposed to external phenomena does not necessarily imply that domestic responses are uniform. Internationalization involves reduced barriers to international

flows of goods, capital, and ideas but Page 81 → not necessarily global convergence (a term closer to “globalization”), at least not in the short to medium terms. 3. For a fuller version, see Solingen 1998. 4. See Gourevitch 1978. 5. For a more elaborate effort at mapping constituencies, see Solingen 2001b. 6. The suffix in “internationalizing “ indicates a process, a path, an approximation but never quite the idealtype. 7. Internationalizing coalitions can retain some state intervention and industrial policy, but they do allow the expansion of private capital—local and international—far more significantly than backlash coalitions. 8. On how states characterized by small export sectors are more prone to choose war, see Domke 1988. 9. See Solingen 1998, 262. 10. See Solingen 1998. For an earlier version focusing on a more discrete dependent variable, see Solingen 1994. 11. Cases included South and North Korea, Brazil, Argentina, Israel, the Palestinian Liberation Organization/Palestinian Authority, Egypt, Syria, Iran, Iraq, and Jordan, and less inclusive treatments of Lebanon, Chile, Taiwan, Saudi Arabia, the Gulf sheikhdoms, Morocco, and Tunisia. 12. On the relationship between openness to capital flows and decline in militarized trade disputes, see Gartzke, Li, and Boehmer 2001. On trade openness and reduced use of force, see Oneal and Russett 1997, 281. 13. See Solingen 2001b. 14. The difference in means across coalitions is significant at the .01 level. A test of multiple comparisons shows mean differences as significant at the < .001 level. Only the mean TO difference between hybrid and backlash is statistically insignificant. Data is from World Tables (1991, 1995); the Statistical Yearbook (United Nations 1999a); World Development Indicators (World Bank 1998, 310); and World Development Reports (World Bank 1991, 1992, 1993, 1994, 1995, 1996, 1997). 15. F = 4.075, significant at the (½ r – c) implies that d < (½ r + c). Thus, offers (½ r – c) < d < (½ r + c) are mutually preferred to a contest. For war to occur in this simple framework, states must prefer fighting to higher payoffs from negotiated settlements, and so war does not occur in this model. Since war has positive costs, and since the contest itself is assumed not to yield any consumption value, the resources available to distribute are invariably less after contests than prior to fighting. Similarly, adding some additional benefit to maintaining the peace (or, alternately, adding an additional cost for fighting) typically does not serve to modify dispute behavior (certainly not if there exists no motive for fighting to begin with).6 Suppose that A and B obtain a trade benefit (n i, where n > 0 and i ε [A,B]), so that A's payoff for peace is (r – d + n A) and B obtains (d + n B) if the two states refrain from fighting. If the two states experience a contest, assume that the benefit of interdependence is forfeit.7 Here A's optimal offer Page 101 → simply subsumes B's trade benefit into a more extractive demand (½ r – c – n B), so that A again can make an offer that B just prefers to fighting. The effect of opportunity costs for fighting linked to trade is simply to allow A to redistribute a larger portion of the mutual benefit toward itself (and away from B). There is no increase in the probability of peace (the probability of fighting remains at zero in the game). The opportunity costs associated with economic interdependence are generally unlikely to have any effect on whether states fight, but interdependence can have an impact on dispute behavior if it alters informational conditions. Indeed, economic interdependence can increase the likelihood of contests if interdependence generates additional uncertainty. Suppose that there are two types of state B, a high-value interdependence state (nBhigh) and a low-value state (nBlow, where nBhigh > nBlow). Suppose also that high-value and low-value types of B are distributed according to some function P, so that the probability of observing nBhigh equals p and the odds of experiencing nBlow equals (1 – p). Finally, assume that B is privately informed about its type (“high” or “low”). Since A does not know what B is just willing to accept, there is now a risk that A will make an offer that B sees as inadequate. While B can inform A about its type, Bs that place a high value on interdependence attempt to claim

otherwise, since the revelation of this information leads A to extract more in its demand. Thus, both types of state B have incentives to claim to have relatively little value for the benefit of interdependence. Because of this “pooling” of types, B cannot expect A to believe the equilibrium claims made by its competitor. A's optimal mixed strategy in this game is to demand (½ r – c – nBhigh) with probability p and (½ r – c – nBlow) with probability (1 – p). However, this means that A makes an extractive offer and encounters a state B with a low value for interdependence with probability px (1 – p). If there is a positive probability of experiencing each type of player B, then there must also be a positive probability of a contest between players. Thus, by allowing economic interdependence to introduce uncertainty, we have a situation where interdependence can generate costly contests. Yet it is less likely that uncertainty is introduced by interdependence than that economic linkages are a major avenue facilitating the revelation of information. Uncertainty is probably as ubiquitous a feature of international affairs as is warfare. States were drawn to fight in all epochs because of the inability to bargain effectively when opponents concealed, or were not able to credibly reveal, the actual value of strategic variables. The traditional options in world politics have largely been limited to acquiescing to bluffs or resorting to military violence to force the revelation of genuine states of affairs. Interdependence Page 102 → allows states to use common economic linkages as a substitute for disputes to credibly communicate resolve through costly signaling. A key insight about the causes of war suggested by the bargaining approach is that many, if not most, contests result from asymmetric information among actors with incentives to compete. The effect of uncertainty in games of international competition is to allow the state making a demand to err in estimating what its opponent will accept in lieu of a contest. Weak or unresolved states that wish to benefit from concessionary demands have incentives to attempt to deceive competitors about their willingness or ability to compete (to bluff) whenever opponents are uncertain about the true state of strategic conditions. By exaggerating their willingness or ability to compete, the leaders of states may be able to obtain preferred settlements to those they are likely to obtain through honesty. This encourages unresolved or weak states to claim to be strong or resolved unless or until these states are faced with costly or risky contests. War and other forms of military violence thus serve to reduce uncertainty. Competition must occasionally resort to force to distinguish resolved or capable competitors from less resolved opponents. So how can interdependence lead to peace? Several studies offer an interpretation of the informational effects of interdependence consistent with the bargaining approach (Gartzke, Li, and Boehmer 2001; Gartzke and Li 2002; Morrow 1999). Valuable linkages between states can serve as the common basis for costly signaling. If states fight because they have trouble identifying mutually acceptable bargains short of a contest, then factors that allow states to reduce uncertainty should more often allow interdependent states to avoid the need for military violence. Imagine again that two states compete over some outcome. As before, imagine that A makes an offer to B, which B can accept or reject (leading to war). This time, however, instead of two types of state B with different benefits for interdependence (nBhigh and nBlow), suppose that types of state B are differentiated based on the costs they face in fighting. Imagine that states differ in their subjective experience of the rigors of war, so that states B may have high war costs (chigh) or low costs for fighting (clow, where chigh > clow). Again, the probability that B has high war costs equals p while the probability that B has low war costs equals the converse (1 – p). With no interdependence, A offers B (½ r – c high) with probability p and (½ r – c low) with probability (1 – p). With interdependence, A offers B (½ r – c high – n B) with probability p and (½ r – c low – n B) with probability (1 – p). Notice that the two sets of demands parallel each other. The probability of fighting is unaffected by the costs or benefits faced by actors directly, but the prospects of war are affected by uncertainty. State B refuses A's Page 103 → offer if its costs for fighting are low and A makes a demand that presumes that B has high war costs. Since the benefits of peace change, but not uncertainty, there is no change in fighting. There are several ways to overcome the problem of cheap talk, the situation when actors that might settle for a less generous offer seek to “pool” or hide themselves among actors that are unwilling to accept less generous bargains. Fearon (1995) argues that liberal democracy leads to “domestic audiences” that punish leaders who make aggressive claims and then back down in conflicts. Schultz (1998, 1999) makes a similar claim involving domestic

opposition groups. Talk can also be made more credible by linking statements to actions that have up front (ex ante) costs. Interdependent and economically integrated states face the prospect of economic losses if the leaders of these states pursue competitive political goals. Costly signaling through economic interdependence suggests that it is not the possession of economic linkages, but the differential willingness of competitors to relinquish these benefits, that leads to relative peace. States that are willing to relinquish their trade benefits in a competition demonstrate resolve in almost the same way as states that are willing to fight. Interdependent states can “fight” through their economic linkages, so that they sometimes are able to avoid more costly military action. Suppose that B has an initial opportunity to demolish or diminish the trade linkages between it and state A. If B expects A to offer (½ r – chigh – nB), and if B is type clow, then B weakly prefers to sever trade benefits. This is because B type clow expects to fight if it receives the less generous offer (since d = ½ r – chigh – nB → UBlow of peace = ½ r – chigh < UBlow of war = ½ r – clow). Here B type clow anticipates receiving no trade benefits in any case, so it can use the destruction of these benefits to signal its type to A. The B type chigh anticipates accepting any offer from A (since d = ½ r – chigh – nB → UBhigh of peace = ½ r – chigh = UBhigh of war = ½ r – chigh), and so B type chigh has no incentive to damage its trade benefits. Here A prefers to offer all opponents demands that they will minimally prefer to fighting. If B type clow reveals itself through the costly signal, A prefers to offer B type clow a more accommodating demand (d = ½ r – clow, since UA of peace = ½ r + clow > UA of war = ½ r – c). This implies that costly signaling through interdependence can reduce the probability of disputes (since UBlow of peace = ½ r – clow = UBlow of war = ½ r – clow, B weakly prefers to accept). While the details of a proof are complex, the basic insight is perhaps simpler. An explanation for the impact of interdependence on peace must address the causes of war. If war is largely informational, then one must search for informational effects of interdependence and economic integration. Interestingly, Page 104 → one of the features of markets that is often highlighted by economists is that markets efficiently aggregate information. It might be controversial to claim that markets promote peace by anticipating war, but the ability of markets to respond to changes in risk and their impact on the welfare of political elites suggests that markets better allow leaders to communicate credibly about relative resolve, reducing the need to resort to military violence. I next turn to a discussion of several extensions of this logic and of approaches to testing.

The Moral of This Story: Conflict Should Not Be Seen as Synonymous with Contests There is a strong tendency in the literature on international relations to conflate conflict among actors with conflictual behavior or contests. Lumping the two together makes it difficult for researchers to distinguish the impact of economic factors on conflict (affecting the terms of settlements) or on contests (influencing the probability of disputes or wars). Fearon (1995) follows Blainey (1973) in arguing that a failure to distinguish conflict from contests means that most explanations for why states fight are not capable of motivating the behavior of interest. As Morrow (1999) and Gartzke, Li, and Boehmer (2001) demonstrate, interdependence is generally likely only to influence contests by altering uncertainty through costly signaling.8 Yet material aspects of interdependence can influence conflicting state interests. The relative valuation of beneficial linkages should be an important contributor to what each state can expect to achieve in bargaining. States that are asymmetrically dependent may be expected to carve out agreements that asymmetrically benefit the less dependent state.9 The perspective of dependency theory and other arguments concerned with distributional and normative issues is not entirely incompatible with the informational perspective. If one is interested in the nature of settlements of conflicts of interest or in what agreements obtain, then variables measuring relative bargaining power are critical. Once one seeks to explain which conflicts become contests, however, the distribution of bargaining power is substantially less relevant. What then matters is the informational status of the interaction between actors and whether states are able to identify mutually preferred bargains ex ante. Here, signaling through interdependence can benefit all parties, even asymmetrically dependent ones (though the reduction in uncertainty will be limited by asymmetry). Differentiating between the impact of interdependence on conflicts of interest and on contests (which reflect but are not determined by conflicts of interest) appears to be a critical step in better understanding the effect of markets on Page 105 → global politics. The evidence marshaled in favor of the opportunity-cost argument then

appears assailable, and optimism about peace depends on the informational, rather than material, consequences of integration. This implies the need for an exploration of markets to better understand which aspects of markets most inform politics.

Interdependence Is Not Just about Trade A second tendency of the literature is to equate interdependence with trade. Flows of goods and services are only one component of economic interdependence. Stocks and flows of capital, in particular, are at the core of any prospering economic system. Basic accounting rules require that net trade equal net capital flows, but this does not tell us much. While goods and services transactions must be canceled by an equivalent transaction in currency, the vast majority of currency transactions are conducted for other ends. A huge quantity of money circles the globe each day in the form of loans, investments, and speculations. While speculating might at first appear to be the least desirable of these efforts, speculative currency markets serve to divest investors, corporations, and governments of unwanted risk. All of these elements of economics are lost in most approaches to evaluating interdependence and interstate conflict. Indeed, the informational focus of costly signaling suggests that trade may potentially be the least important component of the economic and political processes linking and affecting the dispute behavior of states. A costly-signaling theory of liberal peace suggests that states can communicate credibly short of war through economic linkages. States that are willing to risk economic harm in the pursuit of political objectives inform competitors about relative resolve. Political shocks in turn interfere in markets to the extent that (a) markets are affected by political behavior and (b) states are unable to restrain markets in times of crisis. The linkages represented by modern capital markets are far more extensive and more sensitive to risks implied by political shocks. The effect of political shocks on money and securities markets is more immediate (the price of money goes up) and more pervasive (domestic markets may only be modestly affected by disruptions in trade but all aspects of an economy depend on the price of money) than trade. Perhaps most important of all, signaling occurs autonomously. States need take no action to block their economic relationships. Instead, market actors responding egoistically to economic threats posed by political competition credibly and reliably impose costs on political entities. States that pursue actions that threaten economic returns are punished by markets. Leaders who are willing to endure economic consequences Page 106 → of political competition differentiate themselves from others who are less resolved.10 States can learn about private political information by watching the markets. Liberal economics also involves limiting the ability of the state to intervene in markets. If a state is vulnerable to the market and if the market can respond, but is prevented from doing so by government intervention, then the signaling effect of market reactions is lost. States have incentives to intervene to halt or avoid negative economic consequences brought about by shocks from political competition. If a given state has extensive institutional and informal restrictions on such intervention, then it is less likely that the state will intervene and more likely that the adverse reactions of the market will be informative to political observers. Leaders in such a state are likely to become aware that their actions are limited in regard to the market and that acting aggressively in the international arena can result in negative economic consequences. Thus, liberal economic institutions and traditions have the effect of tying the hands of leaders, making it more difficult to avoid negative economic consequences of political shocks. The omission of capital and monetary policy coordination and restraint variables from studies of interdependence may not be problematic if these variables just mimic the effects of trade. There is evidence, however, that while these variables correlate with trade, capital and monetary variables also have important independent effects. Gartzke, Li, and Boehmer (2001) show that the inclusion of monetary policy coordination and net capital flow variables significantly predict dispute behavior while trade does not. Further exploration of the empirical relationships is called for, but it seems reasonable to assert that quantitative research attempting to link interdependence with peace must consider other major aspects of what makes states interdependent.

What Is the Relationship between Interdependence and Globalization? A final way of extending analysis of interdependence is to explore global economic integration. Two separate literatures have developed addressing linkages between interstate economics and political conflict.

Interdependence is basically the study of how economic ties (mostly trade) influence interstate dispute behavior. For lack of a better definition, globalization is interested in how growing integration of world markets affects domestic and world politics. The economic-interdependence literature generally appears to offer a more optimistic message about the consequences of economic linkages. Students of interdependence have also gone further in mastering some of the technical and Page 107 → methodological issues involved in studying this aspect of international political economy. Research on globalization might benefit from greater exposure to the concepts and findings of the interdependence community. At the same time, students of interdependence have tended to ignore systemic, domestic, and dynamic aspects of political economy. Research on globalization has paid more attention to distinctions between state and market, leader, citizen, and firm. Globalization research is careful to recognize that states fight and firms trade. Studies of globalization also treat linkages between states and markets as both a temporal and extradyadic process. Studies of globalization often adopt unsophisticated views of why states fight, either relying on realist concepts of “stability” or attempting to extend dependency arguments about dissatisfaction. Quan Li and I have begun to apply the costly-signaling approach to the dynamic conception of a systemic process offered by globalization research (Gartzke and Li 2002). Work to date suggests reasons for greater optimism about political consequences of globalization than exists in the literature. The two communities appear to be studying the same subject, though with different concepts, units of analysis, and often different conclusions. The next logical step for studies of interdependence is to begin to map effects and implications out beyond the dyad, across time, and to new aspects of political behavior. It may therefore be appropriate to begin the arbitrage process, borrowing strengths from each community in an attempt to build more robust common conceptual frameworks. This suggests that at least part of future study of interdependence and peace involves integrating insights and addressing shortcomings of globalization research.

Conclusion The idea in this chapter was to pose concerns with the current literature and to offer some possible avenues for future study. I have only briefly touched on a number of issues affecting current and future analysis of interdependence and peace. At the very least, I hope the chapter is provocative enough to elicit reactions (and action) from readers. Knowledge is delayed when we agree to ignore inadequacies and inconsistencies in current arguments. I hope I have pointed to a few inadequacies and suggested ways to improve on what exists today. That being said, it is worth remembering that the relationship between liberal economics and peace is one of the most important and promising areas of research in international relations. Much of traditional international relations is challenged by the evolution of a post–Cold War world in which the study of Page 108 → war, deterrence, and balancing behavior appears less relevant. This is not the case with interdependence. Indeed, as proponents have shown, the advent of more integrated economies appears to coincide with changes in the international system. It has to be considered as likely at this point that economic integration encourages peace. Here, I have emphasized that traditional interpretations of why interdependence should promote a reduction in dispute behavior are likely incorrect. Instead, interdependence, or economic integration, facilitates reductions in the frequency of disputes between certain states by making it easier for these states to signal, reducing the need to experience military violence. While the international system remains one of self-help, there is room to consider the benefits of linkages that promote nonviolent competition and the peaceful resolution of conflict. Such a view is no longer utopian. If the current results are modest and not without controversy, they are also quite promising. The charge to researchers is to refine our understanding of how interdependence promotes peace, to identify more fully the nature and scope of linkages between states, and to determine how to promote further development of beneficial effects.

NOTES I thank Edward D. Mansfield, Brian M. Pollins, and the participants in the Conference on Interdependence

and Conflict, the Mershon Center, the Ohio State University, for their comments and suggestions. 1. For a review of the literature, see McMillan 1997; and Mansfield and Pollins this volume. Some argue that interdependence can increase the prospect of militarized conflict (Barbieri 1996, this volume). See Wagner 1988 for a critique of the dependency approach to interdependence. Gartzke and Jo (2001) find that interdependence is associated with more nonmilitarized conflict behavior and fewer militarized disputes. 2. In an influential article, Oneal and Russett (1997) argue that “the classical liberals were right” and provide evidence that liberal trade, along with democracy, contributes to peace. Here, I argue that the classical liberals were not right, only lucky. 3. Any attempt to summarize a body of thinking as broad and deep as classical liberal theory naturally leads to distortions. My intent here is only to reflect the more common interpretations present in the current literature on interdependence. 4. Considerable recent interest has focused on improving the measurement of interdependence (see Polachek, Robst, and Chang 1999; Crescenzi 2001). Such efforts are desirable, but since most hypotheses involve comparative statics claims, new data is likely to be informative only if results alter the sign value of predicted coefficients. Results from better estimation (such as the nonlinear findings of Beck 2000) encourage additional theoretical and empirical efforts. 5. The model assumes risk neutrality. Risk aversion widens the range of mutually Page 109 → acceptable bargains, while risk acceptance narrows the Pareto space. This means that highly risk-acceptant actors can prefer fighting to any settlement an opponent is willing to make. However, this also implies that such actors are unwilling to settle for bargained settlements at later stages of the game (war becomes total). While possible, there are empirically relatively few contests involving total war. Modeling variation in international conflict behavior using risk propensity is also problematic because no metric typically exists on which to measure risk (see O'Neill 2001). 6. Differential costs can affect dispute behavior if, under a given set of conditions, a state prefers any offer to fighting. These “boundary” conditions occur when the cost of fighting is higher than the value of any demand from an opponent or when actors' interests are so similar that no contest is justified. Notice that it is not sufficient that costs exceed benefits for some demands. 7. One need not assume that all of the benefit of interdependence is lost in fighting; this is just a convenient simplification in the model. Alternately, one can interpret ni as a net benefit of trade. 8. Another possibility is that interdependent states are less likely to fight because they have similar interests (Papayoanou 1999; Solingen 1998). Gartzke (1998) examines this possibility, challenging the notion that liberal democracies are less likely to fight but still supporting the association between economic interdependence and peace. Oneal and Russett (1999b) challenge the notion that the democratic peace is due to similar interests, suggesting that interdependence and democracy determine similar interstate interests. Gartzke (2000) shows that, even allowing for this possibility, states' interests continue to overpower the effect of liberal politics. Gartzke, Li, and Boehmer (2001) include controls for the similarity of state interests and shows that a broader set of measures of economic interdependence continues to inhibit disputes. 9. Wagner (1988) provides an essential caveat. The claim that bargaining power is fungible (that it may be used to advance political or economic interests) should not be confused with arguments that imply that superior bargaining power can be used to realize both political and economic gains. Advances on one dimension must be paid for in terms of opportunity costs for preferred settlements in the other dimension; asymmetric dependence offers options, not omnipotence. 10. Market forces appeared to be informative during the 2001 “spy plane incident” between the United States and the People's Republic of China. President George W. Bush initially reacted to Chinese seizure of the U.S. aircraft forcefully but adopted a more conciliatory tone when it became apparent that his attitude was unsettling to U.S. stock markets.

REFERENCES Barbieri, Katherine. 1996. Economic Interdependence: A Path to Peace or a Source of Interstate Conflict? Journal

of Peace Research 33:29–49. Beck, Nathaniel. 2000. Trade and Conflict in the Cold War Era: An Empirical Analysis Using Directed Dyads. Typescript, University of California, San Diego. Page 110 → Blainey, Geoffrey. 1973. The Causes of War. New York: Free Press. Crescenzi, Mark. 2001. Economic Exit, Interdependence, and Conflict. Typescript, University of North Carolina. Fearon, James D. 1995. Rationalist Explanations for War. International Organization 49:379–414. Gartzke, Erik. 1998. Kant We All Just Get Along? Opportunity, Willingness, and the Origins of the Democratic Peace. American Journal of Political Science 42:1–27. _____. 1999. War Is in the Error Term. International Organization 53:567–87. _____. 2000. Preferences and the Democratic Peace. International Studies Quarterly 44:191–212. Gartzke, Erik, and Dong-Joon Jo. 2001. Tipping the Scale: Signaling and the Democratic Peace. Typescript, Columbia University and Pennsylvania State University. Gartzke, Erik, and Quan Li. 2002. War, Peace, and the Invisible Hand: Positive Political Externalities of Economic Globalization. Typescript, Columbia University and Pennsylvania State University. Gartzke, Erik, Quan Li, and Charles Boehmer. 2001. Investing in the Peace: Economic Interdependence and International Conflict. International Organization 55:391–438. McMillan, Susan M. 1997. Interdependence and Conflict. Mershon International Studies Review 41:33–58. Morrow, James D. 1999. How Could Trade Affect Conflict? Journal of Peace Research 36:481–89. O'Neill, Barry. 2001. Risk Aversion in International Relations Theory. International Studies Quarterly 45:617–40. Oneal, John R., and Bruce Russett. 1997. The Classical Liberals Were Right: Democracy, Interdependence, and Conflict, 1950–1985. International Studies Quarterly 41:267–93. Papayoanou, Paul. 1999. Power Ties: Economic Interdependence and War. Ann Arbor: University of Michigan Press. Polachek, Solomon R., John Robst, and Yuan-Ching Chang. 1999. Liberalism and Interdependence: Extending the Trade—Conflict Model. Journal of Peace Research 36:405–22. Schultz, Kenneth A. 1998. Domestic Opposition and Signaling in International Crises. American Political Science Review 94:829–44. _____. 1999. Do Domestic Institutions Constrain or Inform? Contrasting Two Institutional Perspectives on Democracy and War. International Organization 53:233–66. Solingen, Etel. 1998. Regional Orders at Century's Dawn: Global and Domestic Influences on Grand Strategy. Princeton: Princeton University Press. Wagner, R. Harrison. 1988. Economic Interdependence, Bargaining Power, and Political Influence. International Organization 41:461–83.

_____. 2000. Bargaining and War. American Journal of Political Science 44:469–84.

Page 111 →

Trade and Conflict: Uncertainty, Strategic Signaling, and Interstate Disputes Arthur A. Stein

Arguments in Search of Data, Findings in Search of a Logic That trade reduces conflict between states has been a tenet of liberal thought since its articulation in the late eighteenth and early nineteenth centuries. Classical liberals provided several lines of reasoning, however, and the historical record of the last two centuries undercuts some of their assertions, even as it provides a great deal of data with which to assess their central claim and their competing antecedent links. To this day, the relationship between trade and conflict remains mired in a host of conflicting logics and empirical results, and scholars have turned to arguing about data and specification rather than dealing with core analytic problems.1 This chapter delineates some truths about trade and conflict that pose problems for many arguments made in the literature. Typically, trade is treated as an independent variable that reduces the incidence of conflict. Yet trade is itself endogenous to political calculations and decisions. That is, interstate cooperation and conflict affect trade. Second, most formulations presume trade to facilitate conflict avoidance. Yet we know that trade generates conflict and that states use trade as an instrument of coercion. Thus, blithe assertions that trade reduces conflict ignore both the cooperative bases of trade and the tensions and conflicts created by it. They also ignore that trade is a tool of statecraft, used both as carrot and stick. This chapter then constructs an alternative argument about trade and conflict, one that both incorporates what is known about trade and roots it in Page 112 → a broader theory of conflict. In contrast, most extant logics linking trade and conflict depend simply on a presumption of relative cost, arguing implicitly that trade increases the costs associated with conflict and thus should reduce its incidence at the margins.2 The line of reasoning developed here embeds the impact of trade in a broader theory of the causes of conflict and the conditions for cooperation. Yet the chapter suggests remaining flies in the ointment for this argument, as well, and discusses why finding any relationship between trade and conflict is inherently problematic. The reasoning developed here is that the coercive potential of trade is the basis of any empirical finding that it reduces conflict. Trade reduces the escalation of political disputes and thus the incidence of militarized disputes because trade both increases the costs of conflict and gives states a set of coercive instruments with which to signal their commitment in any political dispute. It thus reduces both the occurrence of political crises and the need for militarized actions once they arise. The chapter begins by delineating the endogeneity of trade by laying out its cooperative bases. Next, the chapter explains that linking trade with reduced conflict is problematic because trade both generates disputes and is used as a coercive instrument of statecraft. Third, the chapter develops a signaling argument that is consistent with these realities of trade and with a general theory of conflict. This section also discusses the problems in developing a strategic-interaction argument linking trade and conflict. Fourth, the chapter discusses the different units of analysis in trade and conflict and the need for an analytic link between society and state in order to have a complete theory.

Cooperative Bases of Trade The first problem with the argument that trade reduces conflict is that trade is not exogenous. Rather, trade itself reflects interstate cooperation and conflict. Intergovernmental agreement is a prerequisite for trade. Moreover, governments sometimes encourage trade with specific countries for political purposes and use a variety of levers to affect trade levels. Finally, private traders incorporate political considerations into their decisions. For all these reasons, trade levels already reflect and embody cooperative relations between states, and this must affect any

assessment of the independent effect of trade upon cooperation and conflict. Trade Agreements Trade presupposes cooperation between countries. Excepting smuggling, trade between countries requires some minimal degree of interstate cooperation— Page 113 → that of trade agreements between countries. In the last two centuries, little trade has developed without such agreements, which resulted from political decisions of governments. Classical bilateral trade agreements allowed trade in specific goods at some nonprohibitory tariff rate. This made possible some trade in a mercantilist world with many trade prohibitions. But the classical liberals railed against such trade agreements as mercantilist in character, allowing both trade barriers and trade discrimination. The mechanism used to liberalize trade on a nondiscriminatory basis was the insertion of a most-favored-nation (MFN) clause in bilateral trade agreements. In this way, states assured one another that any tariff reductions they extended to others in any subsequent trade agreements would automatically be extended to any nation with mostfavored-nation status. The use of MFN clauses led to the adoption of two-tier tariff schedules. The high, typically prohibitory, tariff schedule applied to all nations with whom there was no MFN agreement. The lower tariff rates applied to those with which the nation had negotiated trade agreements with MFN clauses.3 MFN clauses were thus a mechanism by which to allow trade within a club (a subset of nations linked by trade agreements with MFN clauses) while effectively excluding it with nonmembers.4 Thus, trade in most items, even non-strategic ones, with the Soviet Union and China during the Cold War was not viable without an MFN agreement.5 In addition to discriminating among states by signing trade agreements and by the extension of MFN status, states also negotiate specific preferences for specific countries in specific categories of goods. Governments can and do negotiate country-specific preferences intended to channel trade toward specific politically more important countries.6 Hence, trade flows, at least to some degree, reflect the political interests of countries. Channeling Trade Trade is undertaken by private actors, and although governments can choke it off by erecting barriers they have a more difficult time making it flow. Government policy is thus asymmetric. It can bring trade to a halt, but it cannot determine trade. Yet governments have some instruments at their disposal to direct trade.7 Governments can open the tap by reducing trade barriers, selectively opening the tap in some sectors and not others and for some countries and not others. They sign trade agreements with some countries and not others, and they distinguish between least and most favored nations. And although they cannot determine trade levels themselves, they can encourage and channel them. Page 114 → States channel trade toward specific countries, for example, by subsidizing some exchanges and not others. Countries often provide trade financing and subsidized insurance for some international transactions.8 States channel trade to specific other countries for political reasons (Skålnes 2000), including the need to solidify alliances as well as to signal commitment.9 They are interested in solidifying alliances, for example, by both economically strengthening their allies and signaling to prospective enemies their commitments to defend their allies and clients. In the deterrence literature, the empirical findings suggest that trade ties do indeed deter (Huth and Russett 1984). Finally, trade is itself affected by interstate cooperation and conflict even without direct governmental intervention. Traders as well as foreign investors prefer stability and eschew conflict. Although spot markets may be immune from such considerations, long-term commercial links depend on continuity and stability. Indeed, liberalizing (if not liberal) politicians, aware that investment, commerce, and wealth depend upon international

tranquility, work to resolve existing foreign conflicts as part of their domestic strategy for economic growth and development (Solingen 1998). Conversely, countries that are rivals or enemies purposely constrict trade between themselves. Such relationships embody closure, sanctions, embargoes, and the like. During the Cold War, even as the United States liberalized trade with its allies, it simultaneously controlled trade with the Soviet Union (Stein 1984). In summary, trade levels already reflect political relations. The role of governments in determining trade flows must be reflected in any attempt to infer the political consequences of trade. Any statistical relationship between trade levels and conflict may reflect not the conflict-reducing implications of trade but the political underpinnings of trade in the first place. Any assessment of the political consequences of trade must deal with its political roots.

Trade, Conflict, and Coercion The second fly in the ointment for the argument that trade reduces conflict is that trade often creates conflict. States also use trade as an instrument of coercion, both by threatening trade disruption and imposing trade sanctions. The argument that trade reduces conflict by making it more costly necessarily implies that trade reduces the incidence of all disputes that carry the cost of reduced trade. This argument is problematic. First, such a line of reasoning should generate robust empirical results across an array of data sets of international conflict. All kinds of conflict, not just militarized interstate disputes Page 115 → (MIDs), should be reduced given this simple causal formulation. There is no analytic logic to limiting the domain of effect simply to MIDs. Second, the existence of trade disputes and trade wars poses a general problem for this simple logic. That trade raises the costs of conflict has not prevented states from purposely disrupting trade as part of their disputes with one another. Trade as a Cause of Conflict Trade relationships embody and generate conflict. Growing levels of international commerce are the basis for trade disputes. Initially, political conflicts over trade emerged when countries had to renegotiate tariff-reduction agreements or when they dumped underpriced goods. As tariffs have declined dramatically and their levels need not be regularly renegotiated, conflict has emerged over a whole array of incongruent domestic practices that are the basis of charges of unfair trade (Stein 1993). The historical record is replete with trade wars (Conybeare 1987) and trade disputes, and these arise only among countries with extensive commercial ties that are, typically, politically close. It is striking that during the Cold War, the United States and its allies engaged in numerous commercial disputes even as they strove to sustain a military alliance of great strategic importance. Thus, despite overwhelming strategic imperatives and close political ties, trade nevertheless became the basis of conflict.10 The critical implication is that trade disputes presuppose trade. Trade disputes do not arise between nations that do not trade with one another. And the growth of trade between countries generates trade frictions and disputes. Trade as a Coercive Instrument of Statecraft Trade has, moreover, been used as an instrument of coercion. Again, the historical record is filled with cases of economic sanctions of various kinds undertaken in pursuit of political rather than economic objectives (i.e., as opposed to trade wars undertaken to deal with solely commercial conflicts). Countries with extensive economic links have been prepared to cut such ties to compel a change in others' policies. The very existence of trade provides states with a tool they would not otherwise have when disputes arise. They can and do impose economic sanctions of various kinds to get others to shift course. Further, the prospect of sanctions affects state decisions. When markets are not deep and competitive, states can

become quite dependent on particular Page 116 → buyers and sellers and vulnerable to the exercise of trade coercion. Indeed, the centrality of certain goods to an economy can lead governments actively to prevent the development of trade dependence. The United States pursues a policy of energy independence despite the fact that the states on which it is dependent are political allies. It also reacts adversely to politically induced price fluctuations absent any effort at political extortion. In short, economic sanctions, a form of interstate conflict, presuppose the existence of a prior trading relationship. Clearly, trade generates neither such levels of cooperation nor such aversion to conflict as to preclude either commercial disputes or the use of economic coercion for political objectives. The use of trade sanctions as instruments of coercion implies that trade does not preclude conflict and does not necessarily lead to cooperation. Moreover, in such cases, states are prepared to absorb the costs of trade disruption and to do so in a political dispute that has yet to be militarized. Trade can be both a source of conflict between states and a casualty of coercive diplomacy in interstate disputes. Trade in no way guarantees a pacific world. The challenge for any coherent and complete theory of the political consequences of trade is to incorporate these realities of trade into a causal logic linking trade and conflict. One implication of the foregoing discussion is the need to specify clearly the dependent variable. One reason for conflicting hypotheses and empirical results is an inadequate specification of the dependent variable, whether trade reduces all forms of conflict or only militarized disputes. Hypotheses linking trade and conflict simply have to be precise both in the selection of a dependent variable and in its link to the underlying analytic argument.

Unraveling the Trade-Conflict Nexus The challenge for ascertaining the link between trade and conflict is to embed it in a broader theoretical framework in a way that is consistent with what is known about trade. First, the impact of trade on interstate conflict should relate to a general theory of conflict. Second, it should recognize that trade generates conflicts of interest and that trade sanctions are used as coercive measures in interstate disputes. Such an argument is developed here, and it provides an explanation for how trade may simultaneously increase disputes between nations while providing them a means to resolve them at a lower level of interstate conflict. In the analytic logic provided subsequently, regardless of whether trade reduces the propensity for political disputes among states, it nevertheless reduces the chances that political conflicts will escalate to the level of militarized disputes. It Page 117 → does this precisely because of the ability to use trade coercively to signal resolve and commitment in an interstate conflict. This would also explain why empirical studies using different data sets come to conflicting empirical conclusions. Trade as a Political Signal In any strategic-choice setting, costly conflicts should not occur. When conflicts are costly to all parties, actors are invariably better off moving to the postconflict outcome and not incurring the costs of conflict. A classic economic problem is that of explaining strikes. After all, workers and employers are interested in continuing their relationship but want to renegotiate the terms of the employment relationship. A strike is costly to both, and both are better off going directly to the outcome that would be negotiated given a strike but doing so without incurring the costs of a strike. Similarly, states would be better off going directly to a postwar settlement without incurring the intervening costs of conflict. Wars and strikes thus pose comparable problems to a model of strategic choice built on rational calculation. The explanation for the occurrence of costly conflict is that actors have private information about their resolve and have no credible way to signal that to others short of incurring the costs of conflict. What emerges is a view of military engagements as rooted in informational asymmetries and incomplete information in which the dynamics of bargaining and even conflict are about demonstrating resolve and signaling commitment.11 Such militarization can be avoided through costly signaling.

In such a setting, trade plays an important role in interstate relations. Assume that conflicts of interest arise between states irrespective of their trading relationship. Minimally, we know from both the existence of trade disputes and the use of trade sanctions that states that trade with each other still get into conflict. Additionally, conflicts of interest arise between states that do not trade with each other. Indeed, the absence of trade in such cases may itself reflect an underlying political hostility that may lead to periodic crises. When conflicts of interest arise, states look at the range of instruments they have to signal their concerns and the intensity of their preferences. Moreover, in such settings, talk is cheap, and costly signaling is one mechanism to evince commitment and resolve. In relationships in which there is some trade, economic sanctions are an intermediate step between mere diplomacy (typically symbolic and verbal) and military measures. States often resort to economic sanctions prior to military actions because they are less confrontational and do not run the risk of accidental war that militarization may carry. In conflicts of interest in which there are no economic links, states have little Page 118 → choice between mere diplomacy and military steps. Military measures short of war are then coded as militarized interstate disputes. But where economic measures are available whether used or merely threatened, they will be sufficient in some cases to signal resolve and avert a more militarized dispute.12 Moreover, even when economic measures are not sufficient, if the crisis continues for enough time, the turn to the military would occur at a time of reduced trade between the parties. The prospect of trade disruption as a signaling device would then explain the existence of an inverse relationship between trade and conflict, one based on an alternative logic to that typically offered in the literature. Higher levels of trade are associated with lower levels of militarized disputes because trade provides a mechanism for costly signaling short of military measures. In some cases, lower levels of trade are associated with militarized disputes both because the parties have little or no trade and because, in other cases, the use of economic sanctions has substantially reduced the trade by the time of the emergence of the militarized dispute. Trade then does not reduce conflict; it provides an instrument of statecraft short of military action. It is the very coercive potential of trade sanctions that exists in states with commercial links that provides the opportunity to avoid militarized disputes.13 This argument integrates the trade-conflict finding in a broader argument about the basis of war. That is, militarization and war occur because of private information and an inability credibly to signal one's interests and the intensity with which they are held. Trade provides a mechanism for sending a costly signal and revealing private information short of using the military. This argument can be assessed by using more than the militarized dispute data set as the dependent variable and linking it with broader international conflict and cooperation data sets, especially ones that capture economic coercion. Moreover, the argument could be assessed by treating the costly signal itself as a variable and using information on the trade relationship to measure the costliness of steps taken or threatened. The opportunity costs of closure with respect to a specific trading partner would capture the costliness of the signal of a prospective or actual trade embargo. Trade as Signal and the Failure of Economic Sanctions The problem with this argument is that the utility of economic sanctions by states in their political disputes is unclear. Indeed, there is a large literature suggesting Page 119 → that economic sanctions are often ineffective instruments of political coercion. But even as scholars of economic sanctions argue that they are often politically ineffective, scholars in the trade-conflict literature argue that trade reduces conflict. Somehow, the fear of an economic break is sufficient to reduce interstate conflict, but the actual use of economic sanctions is ineffective. There is something problematic about arguing that the anticipation of trade disruption and its costs leads states to eschew conflict and military measures even while the actual use of economic sanctions is held to be ineffective. Indeed, the reasons why sanctions are ineffective pose problems for the trade-conflict literature. Sanctions are

ineffective because sanctioning states rarely have market power, and competing buyers and sellers undercut the effectiveness of unilateral sanctions. But this should prove true for assessments of trade relationships and concerns about trade disruption, as well. In a competitive marketplace, absent market power by specific actors in specific goods, trade should have no political consequences at all. Political calculations can be made without thinking about trade. The disruption of bilateral links by any interstate dispute simply leads to trade diversion.14 Trade diversion undercuts not only the effectiveness of bilateral sanctions but also the costliness of the signal they send. In a competitive market, alternative buyers and sellers are available to both parties whose trade links have been severed. The costliness of trade signals (as well as the viability of sanctions) are determined by the nature of market power and the costs of adjustment. This suggests that trade ties must be disaggregated and assessed industry by industry. Market power may exist in some sectors, such as oil, and not in others. The sectors in which market power exists may or may not be of strategic importance, they may or may not be highly inelastic, and their use may or may not be widespread. The important point is that sanctions constitute costs to the sanctioning state. One can even imagine cases in which the sanctioned state experiences little pain but the sanctioning state does and in doing so sends a message costly to itself but not to the sanctioned state. As a side benefit, the use of trade sanctions as signals suggests a different basis for assessing the effectiveness of sanctions. Sanctions are effective not only when they lead to policy change. First, sanctions can send costly signals that may deter third parties and prevent escalation. Second, sanctions may make it possible to avoid escalation to a militarized dispute. Thus, even when Page 120 → sanctions do not function as costly signals that lead to a ready resolution of the dispute, they can ameliorate conflict by providing an acceptable basis for waging a dispute without militarizing it.15 Trade, Strategic Interaction, and the Problem of Anticipated Reaction Another problem with a costly-signaling argument of trade and conflict is that the trade links between states are known and strategic interaction is inherently built on the logic of anticipated reaction. The arena of international relations is quintessentially one of strategic interaction. States make decisions in interaction with specific others whose choices they are interested in affecting and whose responses must be anticipated and incorporated in decisions. International conflict and cooperation, especially in bilateral relations, are the product of a strategic calculus. Assessing the impact of trade on international conflict requires incorporating trade into such a calculus. The key insight of formal work in strategic interaction is that actors calculate and anticipate reactions to their strategic choices. Actors anticipate others' actions and reactions, and steps both taken and avoided reflect a calculus of expectations. The logic of anticipated reaction creates major problems for empirical work and assessing the implications of what is actually observed. Trade is known and observable to the parties prior to any conflict. Conflict exists in the shadow of trade.16 A state calculating the initiation of some conflict anticipates its trade partners' possible reactions and its own response and so on. The trade costs absorbed by the initiator may make it less willing to initiate a dispute. But the initiator is also aware that the trade costs of conflict will also make the responder more reluctant to sustain the dispute, and this emboldens the initiator. The trade link thus has the simultaneous affect of dissuading and emboldening an initiator to a dispute, and there is no basis for inferring which effect is systematically stronger (Morrow 1999, this volume). The result should be that trade links have no net affect on the initiation of disputes. The strategic-interaction view of conflict is that it is driven by information asymmetry between the parties, information available to one party that is not known by the other and cannot be credibly revealed short of conflict itself. Conflict is thus a product of uncertainty, and what reduces conflict works through its impact on uncertainty. The amount of trade between countries is known, as is its composition. In addition, states know the market conditions for different products and industries and thus the availability of alternative buyers and sellers. Any

decision to become involved in conflict already reflects what is known about trade and anticipates the reactions of others. Conflict is Page 121 → thus in the error term.17 In such circumstances, any empirical relationship between trade and conflict simply captures the uncertainty in the relationship. Trade may reduce conflict because it captures the degree of uncertainty in the relationship: higher levels of trade are associated with greater certainty in the relationship between states, and lower levels of trade are associated with greater uncertainty.

Wealth, the State, and Society The foregoing discussion has largely been built on a unitary-actor view of states, although it recognizes that trade and conflict constitute choices by different actors. But a complete model linking trade and conflict must deal with the disjuncture in units of analysis between the phenomena in question. Trade is, in most places and at most times, carried out by private actors, firms, and individuals and reflects societal supply and demand conditions. Conflict is an interstate phenomenon, carried out by governments. Any link between conflict and trade must, therefore, confront the relationship between state and society and how societal interactions become translated into state actions.18 The relationship between state and society as regards trade is asymmetric. States control cross-border flows and make trade possible. Getting rid of barriers does not determine flows, however. As argued previously, states do channel and shape the nature of cross-border flows through a host of public policies. Yet states vary in their involvement in the private economy: some states control enterprises, others do not, and even liberal states vary tremendously in their degree of intervention in the private economy. States thus vary in the extent to which they determine the flow of goods and services between them, with few cases at either extreme of complete control of trade or complete lack of control of trade.19 Moreover, states cannot be assumed to reflect the interests of private commercial actors, nor can private actors be presumed to reflect the interests of the state. This is implicit in the use of economic sanctions by governments. That states need to impose sanctions implies that private traders do not by themselves eschew commercial transactions with states with whom their government has political disagreements. States not only impose sanctions but also monitor them and punish cheaters. This implies that, at least to some degree, private actors are willing to exchange with citizens of other countries even in the context of political disputes. Private traders may follow the flag, but once there they do not necessarily modulate their commercial activities to the ups and downs of political relations. Page 122 → The disconnect between the interests of traders and those of governments also implies that the preferences of those engaged in foreign commerce do not necessarily get smoothly translated into public policy. Just as governments override commercial concerns in imposing sanctions, they also override commercial considerations by getting into political disputes with states with whom they engage in substantial commerce. This disjuncture between state and society must itself be addressed in any argument linking trade and conflict. A comparable problem exists as regards the sensitivity of states to material concerns. The political consequences of trade presume governmental concern with material conditions. Trade disruptions and their social and economic consequences are presumed to have political effects. Here, too, there is variation across states over both time and space. A marginal-cost argument implies that the relationship between trade and conflict should not be a historical constant but should vary as a function of the importance of wealth considerations in calculations of the national interest. Modern governments, autocratic as well as democratic ones, are more sensitive to the implications of public policy for national wealth. In contrast, nineteenth-century democratic politicians were insulated from such pressures because of the belief in a natural cycle of boom and bust, which government was powerless to affect. And at any time, variations in political systems make some governments more impervious to fluctuations in material wealth.20 Yet in the final analysis, most international relations scholars would concur that security dominates material wealth as a concern for states, with wealth growing in importance as security is assured.

That the trade-conflict relationship depends, to some degree, on presumed state-society links suggests variation in the relationship over time, across political systems, and across circumstances. Not surprisingly, the estimated relationship is sensitive to different temporal domains and the inclusion or exclusion of different states.

Trade and Conflict The relationship between trade and conflict is beset by flies in the analytic ointments. The search for any relationship will have to incorporate the realities of trade: that it reflects state policy as well as being used as an instrument of coercion, that even as it can be turned off by states its levels and composition reflect societal interests and not necessarily governmental ones, that its political implications have varied across political systems and with time, and that market power varies across actors and sectors. More broadly, what is arged in the trade and conflict literature must be consistent with findings on the implications Page 123 → of trade for deterrence and with those on economic sanctions. In addition, any relationship between trade and conflict will perforce have to depend on a logic of strategic interaction and will thus have to confront the problem of anticipated reaction and the fact that conflict in such models derives from informational asymmetries and uncertainty. Finally, to get out of the current morass of incompatible findings that are specific to particular databases, time periods, coding rules, and specifications and controls, statistical estimation will have to reflect the analytics, and the facts, and be carried out at a less highly aggregated level.

NOTES My thanks to all the participants in the September 2000 Conference on Trade and Conflict held at the Mershon Center, the Ohio State University. My thanks to Jana Von Stein for research assistance and to the UCLA Academic Senate for research support. 1. For my review of the literature, combining a historical with an analytic overview, see Stein 1993. 2. Even this marginal-cost argument implies that the relationship between trade and conflict should not be a historical constant but rather vary as a function of the importance of wealth considerations in calculations of the national interest (see the discussion following). For a discussion of the problem with the relative-cost argument, see Gartzke (this volume). 3. I discuss the importance of the MFN clause as an element of institutional design in Stein 1984, 1990; and in Rosecrance and Stein 2001. 4. For a discussion of international politics as involving overlapping clubs, see Rosecrance and Stein 2001. 5. The United States, for example, which has negotiated away its tariffs in many categories, still applies the astronomical Smoot-Hawley tariff rates to nations without MFN status. Indeed, all the agreements to reduce U.S. tariffs were relative to the Smoot-Hawley rates; thus, without MFN, trade in most items was not commercially viable. 6. Skålnes (2000) demonstrates how preferential arrangements reflect political need. For the impact of trade preferences on trade, see Mansfield and Bronson 1997. 7. That trade reflects government support and encouragement is discussed by Mastanduno (this volume). 8. The U.S. government, for example, has an export-import bank for such purposes. 9. For macroassessments of the political bases of trade ties, see Gowa 1989, 1994; and Gowa and Mansfield 1993. 10. Although the discussion here is about trade disputes between major economic powers, this point extends to trade disputes between countries with asymmetric trade Page 124 → relationships. States have used trade policy to foster economic dependence and thus political control. The classic work analyzing this is by Hirschman ([1945] 1980). 11. I have discussed this in Stein 1982, 1990. For a recent formal treatment, see Fearon 1995. For a similar argument, see Gartzke (this volume). 12. The signaling argument about trade and conflict is made by Papayoanou (1996, 1997, 1999). The argument developed here is the converse of his. He argues that economic interdependence generates domestic interests that, operating through domestic institutions, undercut the ability of status quo states to deter revisionists. The argument developed here is a statist one and emphasizes the coercive signaling

potential of trade sanctions. 13. A comparable argument is made by Rosecrance (1999) regarding foreign investment. Rather than the classical liberal argument that suggests that foreign direct investment (FDI) creates mutual interests in avoiding conflict, Rosecrance argues that FDI creates hostages and thus assures interstate cooperation. 14. Just as economic integration can lead to trade diversion without trade creation, so economic sanctions can lead to trade diversion without any trade destruction. 15. There is a domestic politics variant of this argument in which sanctions are effective in that they allow political elites to respond to domestic pressures to express a country's displeasure with others without having to take steps that risk military confrontation. Sanctions are often used in cases in which they are bound to fail. They are typically used to compel rather than deter, to convince another state to undo an action already taken. Cases of successful compellence short of war are hard to come by, even using military measures short of war. Thus, it is not surprising that sanctions do not succeed; they are used when deterrence has already failed. It is also not surprising that they are used in such cases as an alternative to military steps, which might lead to escalation in a context in which any action short of war is unlikely to generate the desired policy change. 16. This play of words is intended to get across the point that trade is known and that all interactions occur in the knowledge of trade relations and the costs of their rupture. 17. This play on the title of an article (Gartzke 1999) is meant to get across the point that it is conflict in general, and not just war, that should be in the error term in any rational model in which conflict is costly; anything known should already be incorporated in a bargaining calculus that should result in the avoidance of conflict. 18. I develop this at length in Stein 1993. This aspect of the trade-conflict relationship is developed in Papayoanou 1999; Skålnes 2000; and Simmons (this volume). Hess (this volume) builds an argument based on one particular state-society linkage. The problem also bedevils Marxist-Leninist arguments, which then rely on the argument that politicians are the ruling committee of the bourgeoisie. 19. Statistical estimation of the trade-conflict relationship should therefore include state-society relations as a variable rather than treat them as fixed. 20. This is also a reason why sanctions have been deemed to fail. Nowadays there is Page 125 →a recognition that sanctions hurt societies without comparably hurting governments and without affecting government policy.

REFERENCES Conybeare, John A. C. 1987. Trade Wars: The Theory and Practice of International Commercial Rivalry. New York: Columbia University Press. Fearon, James D. 1995. Rationalist Explanations for War. International Organization 49:379–414. Gartzke, Eric. 1999. War Is in the Error Term. International Organization 53:567–87. Gowa, Joanne. 1989. Bipolarity, Multipolarity, and Free Trade. American Political Science Review 83:1245–56. _____. 1994. Allies, Adversaries, and International Trade. Princeton: Princeton University Press. Gowa, Joanne, and Edward D. Mansfield. 1993. Power Politics and International Trade. American Political Science Review 87:408–20. Hirschman, Albert O. [1945] 1980. National Power and the Structure of Foreign Trade. Reprint, Berkeley: University of California Press. Huth, Paul, and Bruce Russett. 1984. What Makes Deterrence Work? Cases from 1900 to 1980. World Politics 36:496–526.

Mansfield, Edward D., and Rachel Bronson. 1997. Alliances, Preferential Trading Arrangements, and International Trade. American Political Science Review 91:94–107. Morrow, James D. 1999. How Could Trade Affect Conflict? Journal of Peace Research 36:481–89. Papayoanou, Paul A. 1996. Interdependence, Institutions and the Balance of Power: Britain, Germany, and World War I. International Security 20:42–76. _____. 1997. Economic Interdependence and the Balance of Power. International Studies Quarterly 41:113–40. _____. 1999. Power Ties: Economic Interdependence, Balancing, and War. Ann Arbor: University of Michigan Press. Rosecrance, Richard. 1999. The Rise of the Virtual State: Wealth and Power in the Coming Century. New York: Basic Books. Rosecrance, Richard, and Arthur A. Stein. 2001. The Theory of Overlapping Clubs. In The New Great Power Coalition, ed. Richard Rosecrance, 221–34. Carnegie Commission on Preventing Deadly Conflict. Lanham, MD: Rowman and Littlefield. Skålnes, Lars S. 2000. Politics, Markets, and Grand Strategy: Foreign Economic Policies as Strategic Instruments. Ann Arbor: University of Michigan Press. Solingen, Etel. 1998. Regional Orders at Century's Dawn: Global and Domestic Influences on Grand Strategy. Princeton Studies in International History and Politics. Princeton: Princeton University Press. Stein, Arthur A. 1982. When Misperception Matters. World Politics 34:505–26. Page 126 → _____. 1984. The Hegemon's Dilemma: Great Britain, the United States, and the International Economic Order. International Organization 38:355–86. _____. 1990. Why Nations Cooperate: Circumstance and Choice in International Relations. Ithaca: Cornell University Press. _____. 1993. Governments, Economic Interdependence, and International Cooperation. In Behavior, Society, and International Conflict, ed. Philip Tetlock, Jo Husbands, Robert Jervis, Paul Stern, and Charles Tilly, 241–324. New York: Oxford University Press, for the National Research Council of the National Academy of Sciences.

Page 127 →

Economic Interdependence, Opportunity Costs, and Peace Jack S. Levy The study of the relationship between economic interdependence and international conflict constitutes one of the most vibrant research programs in the international relations field today. It has been propelled by the success of democratic peace theory; by liberals' hopes of finding comparably strong patterns linking economic interdependence to peace and thereby expanding the foundations of the liberal peace; by the realist counterattack; and by the fact that the trade-promotes-peace hypothesis reinforces the current American foreign policy agenda. Much of the literature on economic interdependence and conflict has focused on the question of the net effect of economic interdependence, and of trade in particular, on international conflict. Are high levels of economic interdependence between states associated with a decrease in the probability of militarized interstate disputes (MIDs) or wars, and, if so, is this relationship a causal one? While there are extensive debates over the proper research designs for investigating this question, and while some empirical studies find that trade is associated with international conflict (Barbieri 1996, 2002), most studies conclude that trade is associated with peace, both at the dyadic and systemic levels (Polachek 1980; Mansfield 1994; Oneal and Russett 1997, 1999; Russett and Oneal 2001).1 There is less agreement, however, about the strength of this empirical association, whether it represents a causal relationship, the underlying causal mechanisms driving the relationship, or the conditions under which it holds. I have two aims in this chapter. One is to argue that some important causal mechanisms posited to lead from economic interdependence to peace are logically incomplete, and another is to go beyond the current focus on the net Page 128 → effects of interdependence on militarized conflict and to identify some of the additional testable implications deriving from the trade-promotes-peace hypothesis. Although it is important to assess the net effects of economic interdependence on international conflict, the preoccupation with this question has inhibited theoretical progress. Research programs often advance through the generation and testing of additional “novel facts” (Lakatos 1970) or observable implications of their central hypotheses (King, Keohane, and Verba 1994), and we can gain leverage on our theories of economic interdependence and conflict by identifying and testing their additional implications.2 Some of these additional testable implications of the theory involve the relative frequencies of occurrence of various outcomes and are best tested through large-N quantitative methods. Other predictions concern the perceptions, expectations, and calculations of political leaders, and these hypotheses are often best explored through qualitative methods. Given the predominance of quantitative approaches in the literature and in this volume, I give special emphasis in this chapter to the potential contributions of case-study analyses, but the key point is that the study of economic interdependence and international conflict is best served by the use of multiple methodological approaches.

The Economic-Opportunity-Cost Hypothesis Proponents of the trade-promotes-peace proposition identify several interrelated theoretical arguments in support of their hypothesis but give greatest emphasis to the economic-opportunity-cost argument based on the benefits of specialization and efficiency.3 Because trade exploits comparative advantages and generates economic benefits for both parties, the anticipation that war will disrupt trade and lead to a loss or reduction of the gains from trade creates incentives for political leaders to avoid taking actions that are likely to lead to war against key trading partners (Polachek 1980; Doyle 1997; Russett and Oneal 2001). This opportunity-cost mechanism is reinforced at the domestic level. Trade increases the influence of economic groups who benefit most from trade and who have incentives to pressure the government to maintain a peaceful environment for trade (Rogowski 1989; Solingen 1998). Lower levels of trade reduce the economic opportunity costs of war and reduce economic incentives for political leaders to avoid war.

Realists and mercantilists question the liberal economic theory of peace. They argue that any pacifying effects of economic interdependence are always small relative to the effects of the distribution of military power and related strategic considerations at the dyadic and systemic levels, so that economic Page 129 → interdependence has a negligible impact on international conflict (Gilpin 1987; Ripsman and Blanchard 1996–97). Some realists (and Marxists as well) go beyond this to argue that under certain conditions economic interdependence, particularly if it is asymmetrical interdependence, may actually increase the probability of militarized disputes or war—by creating new issues of conflict between states, by tempting the stronger state to exploit its advantages for coercive purposes (which can lead to conflict spirals), or by other means (Hirschman [1945] 1980; Barbieri 1996). Historically, close trade ties have sometimes led to trade rivalries and trade wars, which sometimes escalate to war (Conybeare 1987; Levy and Ali 1998). The relative impact of economic and other variables on international conflict is in part an empirical question, of course, and one that after years of neglect became the focus of an intensive inquiry by the mid-1990s. While the majority of empirical studies find that economic interdependence tends to reduce the impact of international conflict, the empirical debate is far from resolved.4 Different studies use different indicators of both interdependence and international conflict, different spatial and temporal domains, different control variables, and different statistical techniques.5 While there has been a lot of attention to the different indicators scholars use to measure trade interdependence, there has been less attention to the conceptualization and operationalization of the dependent variable (but see Mansfield and Pollins this volume). Many proponents of the trade-promotes-peace hypothesis argue that economic interdependence reduces the frequency of war, whereas others make the more sweeping claim that interdependence reduces the frequency not only of war but also of militarized disputes and other forms of militarized competition between states. Cosmopolitans like Thomas Paine, for example, argued that commerce promotes a general harmony of interests that reduces low-level as well as higher-level conflicts between states. Immanuel Kant, on the other hand, spoke in more limited terms about the dampening effects of interdependence on war. He emphasized the role of selfinterested business leaders in using their influence to prevent economically costly wars but argued that business leaders would have fewer incentives to try to stop lower-level disputes with fewer direct and immediate costs to their own interests (Walker 2000).6 With Kant generally taken as the spokesman for liberal international theory, it is perhaps not surprising that most theoretical arguments about the pacifying effects of economic interdependence are framed in terms of reducing war, not reducing lesser forms of disputes between states. The important question is not which formulation of the trade-promotes- Page 130 → peace hypothesis is more common, but whether the operational indicator used to test the hypothesis is congruent with its theoretical formulation, and here we often find a gap between theoretical arguments and the research designs constructed to test them. Most of the empirical literature uses indicators of conflict that include militarized interstate disputes and other uses of force short of war without giving adequate attention to the question of whether the trade-promotespeace hypothesis makes predictions about low-intensity conflicts of the kind that make up much of the militarized interstate dispute data set.7 Most MIDs are quite mild: a third last less than a week; over two-thirds involve no battle-related deaths; and only 4 percent escalate to war (Jones, Bremer, and Singer 1996). These are not the kinds of events that Kant imagined might be suppressed by extensive economic ties between societies. Those who operationalize the dependent variables in terms of MIDs need to be much more explicit about the theoretical justification of that choice and the causal mechanisms driving their theoretical propositions. We cannot simply assume that the same causal linkages work for MIDs and for wars, that if trade suppresses wars it will also suppress less costly disputes.8 The question of when disputes escalate to war also raises the possibility of some interesting strategic dynamics. If high levels of economic interdependence reduce the probability that a dispute will escalate to war because of the economic opportunity costs of war, and if political leaders understand this, it is conceivable that political authorities will be less inhibited in initiating disputes for coercive purposes, knowing that the risks of escalation are lower.9 Concerns about inadvertent escalation, on the other hand, might suppress low-level disputes as well. What we need is a more complete specification of the causal linkages between economic interdependence and disputes, interdependence and war, and the separate linkages between disputes and

war. We should also note that the empirical strength of the relationship between trade and conflict, and particularly between trade and war, appears to be fairly modest and leaves much of the variance in the presence or absence of conflict to be explained by other variables. While Oneal and Russett (1999) find that democratic dyads and economic interdependence each have significant dampening effects on international conflict, they also find that capability ratios, territorial contiguity, and major power status have statistically significant impacts on the presence or absence of militarized conflict. These findings suggest that interdependence, democracy, and other variables associated with liberal international theory cannot alone explain variations in international conflict and that such variables need to be combined in a broader theory of conflict.10 Indeed, Oneal and Russett (1999, 3) frame their recent work in terms of a “conceptual Page 131 → synthesis of Kantian and realist theories.” This increases the explanatory power of the model, but at some cost to parsimony and to any claim for a distinctively “liberal” theory of international conflict. The role of other variables in explaining international conflict also suggests that it is misleading to argue that economic interdependence actually deters war, in the sense that a certain threshold of interdependence is a sufficient condition for peace. It makes more sense to treat economic interdependence as a contributory cause of war and to construct the hypothesis in probabilistic terms: the greater the degree of economic interdependence, the lower the probability of war or other forms of militarized international conflict. War can still occur despite the pacifying effects of economic interdependence, if strategic or possibly domestic pressures for war are sufficiently strong. Though the focus of most of the theoretical literature and all of the empirical literature is on current trade, the logic of the commercial liberal theory of peace and its key economic-opportunity-cost proposition implies that it is expectations of future trade that really matter. This idea is the cornerstone of Copeland's (1996) “theory of trade expectations.” He argues that if trade between two states is substantial, but is expected to decline significantly in the future, the economic opportunity costs of war are substantially lower than if trade is expected to continue at current levels. Indeed, if states expect trade to decline, and if they expect to lose access to vital goods in the process, they may be tempted, in some situations, to resort to military force to seize those goods.11 Similarly, expectations of future trade can have a dampening effect on future conflict even when current trade levels are low.12 Expectations of future trade are not necessarily exogenous, however, and political and economic leaders concerned about future trade often take actions to insure against the loss of trade and possibly to increase it. This is explicit in Mansfield and Pevehouse's (2000, 778–81) theoretical explanation of the beneficial effects of preferential trading agreements (PTAs) and their dampening effects on international conflict. Mansfield and Pevehouse talk about “the future stream of gains” from participation in a PTA and emphasize the role of PTAs in maintaining a state's access to key international markets and also in providing insurance against the possibility of protectionist measures by their trading partners in the future.

Some Analytic Problems The economic-opportunity-cost hypothesis of commercial liberalism states that economic interdependence promotes peace because of the fear that the Page 132 → outbreak of war or other forms of militarized conflict will disrupt trade and other economic relationships. It is possible, however, that the causal arrow runs in the opposite direction or is reciprocal. Blainey (1988) criticizes the Manchester liberals on the grounds that peace creates the conditions that facilitate the development of interdependence, not the reverse, and Pollins (1989) and Gowa (1994) each argue that trade follows the flag, so the causal arrow goes from conflict and politics to economic interdependence and not the other way around. The liberal economic-opportunity-cost hypothesis itself states that trade promotes peace precisely because war inhibits trade, so that the relationship is fundamentally reciprocal. While scholars in principle recognize the reciprocal relationship between economic interdependence and conflict,

most empirical research focuses only on one unidirectional relationship or the other. Scholars have only recently begun to systematically examine the reciprocal influences of economic interdependence on conflict and of conflict on interdependence (Reuveny and Kang 1996; Reuveny 2001). This is an important step forward, but one limitation of most of these studies is that they are generally restricted to the post–World War II period (where events data measurements are ample). These studies also say little about the relationship between economic interdependence and war, particularly major war, given the absence of the latter in the last half-century. Empirical studies of the reciprocal relationship between interdependence and peace need to be extended to a broader temporal domain that incorporates greater variation in the intensity of international conflict, including great power wars of the past.13 The observed association between economic interdependence and peace may also be the result of other factors that simultaneously affect both variables.14 For example, states with common interests tend to trade with each other (Morrow, Siverson, and Tabares 1998), and their common interests generate fewer issues to fight about, so the association between trade and peace may be spurious and explained in part or in full by the commonality of interests, quite independently of trade. A related possibility is that because there is more trade between allies than between adversaries (Gowa 1994), and because allies are less likely to go to war with each other (Ray 1990), alliances may account for part of the association between trade and peace. Finally, hegemonic-stability theorists argue that one of the primary conditions facilitating trade is the existence of a hegemon that is able and willing to maintain a stable political economy. They also argue that hegemony promotes peace (Keohane 1984; Gilpin 1981).15 If so, the association between trade and peace might be Page 133 → spurious and explained primarily by the degree of hegemony over the political economy.16 There is another respect in which theorists have not fully specified the causal linkages leading to peace in the economic-opportunity-cost variation of commercial liberalism. The widely repeated argument that the fear of the loss of the gains from trade helps to deter political leaders from taking actions that might lead to war is logically incomplete because it attempts to explain a dyadic outcome (peace/war) with state-level variables (state costbenefit calculations) and ignores the questions of the strategies states use to advance their interests and the processes through which they interact with their partners/adversaries. The desire to preserve peace and its associated gains from trade may very well lead two trading partners to each refrain from belligerent actions when a conflict of vital interests arises. It is also possible, however, that one side might conclude that its partner/adversary is so eager for peace and so fearful of war that by increasing its own demands the first side can secure additional concessions without an appreciable risk of war and its associated economic costs. Or, a state content with the status quo may nonetheless fear that any concessions it makes in the name of peace might only be perceived as a sign of weakness and induce the other side to increase its demands. In the absence of additional information about each side's expectations regarding the economic benefits of trade and the economic opportunity costs of war, as well as each side's domestic sensitivity to those costs and willingness to take risks, not to mention the role of other variables, it is impossible to say for sure whether the conflict of interests will be resolved peacefully or whether it might lead to militarized conflict. Guided by insights from game-theoretic approaches, scholars have recently begun to emphasize the theoretical indeterminacy of the monadic economic-opportunity-cost hypothesis and have attempted to incorporate a theory of strategic interaction to help explain the potentially pacific effects of economic interdependence (Morrow 1999; Gartzke, Li, and Boehmer 2001). These scholars build on the argument that in the context of full information there are settlements that each side prefers to violence and that explaining militarized conflict requires explaining how uncertainty, private information, and incentives to misrepresent that information preclude agreement on mutually beneficial and efficient settlements (Fearon 1995). They also build on the logic of signaling games to suggest that actors can reduce uncertainty through costly signals of their preferences and intentions. The basic argument is that trade and other forms of economic interdependence Page 134 → can promote peace by providing opportunities for states to send credible signals of their resolve without resorting to riskier military actions as instruments of signaling (Morrow 1999; Gartzke, Li, and Boehmer 2001). Higher levels of interdependence lead to a broader menu of options for signaling and thus make signaling more efficient.17 Cutting

back trade or financial flows is costly and helps to distinguish between actors making genuine demands and those who are bluffing in a crisis. This reduces each actor's uncertainty about the intentions and resolve of the other, increases the likelihood that adversaries can find an outcome that each prefers to violence, and reduces the dangers of crisis escalation driven by misperceptions.18 Because the hypothesized deterrent effects of the economic opportunity costs of war, and therefore the impact of economic interdependence on peace within a dyad, may be indeterminate in the absence of an explanation of the nature of strategic interaction or bargaining between trading partners, an important task for future research is to explore the nature of these bargaining processes. Game-theoretic models provide one clear way of doing this. Institutional approaches (game theoretic or otherwise) can also be useful, as demonstrated in Mansfield and Pevehouse's (2000, 781) discussion of the role of preferential trading arrangements in providing a forum for negotiation and bargaining, providing information and reducing uncertainty, and establishing norms of reciprocity.

Other Implications of the Economic-Opportunity-Cost Hypothesis While most tests of the economic-opportunity-cost hypothesis focus on its prediction that economic interdependence reduces the likelihood of international conflict, it is useful to examine some of the proposition's other testable implications. This will extend the analytic power of the hypothesis and provide additional ways of testing it. One key premise of the economic-opportunity-cost hypothesis is that dyadic trade will be significantly reduced or eliminated once war breaks out between trading partners. Evidence contrary to this prediction would raise questions about the validity of the general hypothesis and hence the primary causal mechanism in the liberal economic theory of peace and war.19 If a substantial amount of trade is expected to continue during wartime, then the anticipated economic opportunity costs of war, and hence the economic incentives for avoiding war, will be much lower. There is a long history of “trading with the enemy” that goes back many Page 135 → centuries (Levy and Barbieri 2001). The Dutch were particularly well known for this practice, and certainly no Dutch leader or his international adversary in the sixteenth or seventeenth century would refrain from war on the assumption that trade with the enemy would cease during wartime. Most of the evidence of trading with the enemy is anecdotal and drawn from histories of particular periods, however, and there have been few if any systematic and theoretically driven studies of this phenomenon, at least until recently. In the last few years scholars have begun more systematic investigations of the impact of war on trade. In their study of seven minor power dyads, for example, Barbieri and Levy (1999) found that in most cases war has no statistically significant effect on dyadic trade. Anderton and Carter (2001), while not challenging these particular findings, questioned whether they could be generalized to other cases. Anderton and Carter extended the analysis to major powers and found evidence of significant declines in dyadic trade in a number of major power dyads going back to 1870. While Anderton and Carter raise serious questions about the extent of trading with the enemy, there are enough examples of this phenomenon to warrant future research on this question. It is an important question in itself, and it has critical implications for the economic-opportunity-cost variant of the trade-promotespeace hypothesis.20 It is clear that trade between adversaries sometimes declines during war while at other times it does not, and the total number of cases analyzed is still too small to reach any definitive conclusion. A resolution of this debate on the impact of war on trade will require further analyses based on additional cases, with attention not only to the aggregate effects of war on trade but also to the particular conditions under which trade declines significantly during war and the conditions under which such trade continues at substantial levels. It would also be useful to explore whether the extent of the continuation of trade during war has declined over time as a function of the growth of the state's ability to monitor and sanction such trade; the growth of nationalism and patriotic norms against trading with the enemy; and the institution of the income tax, which reduced the state's dependence on revenues from trade.21 Scholars should also broaden the focus beyond trade and analyze the extent to which

financial flows continue between wartime adversaries. There are also questions regarding what kinds of trade continue, in what kinds of goods, between what kinds of adversaries, in what kinds of wars, and with what impact on states' war efforts and domestic economies. Another question concerns the expectations of political and business leaders regarding the impact of war on trade. Do firms want to continue trade in search of profits or to cut back trade in an anticipation of increased transport and insurance Page 136 → costs? Are governments willing to permit trade to continue, either because of pressures from key economic groups or because of more general fears that the continuation of trade with the enemy is necessary for economic stability? Or are they driven by strategic concerns or patriotic pressures to prohibit trade? Most important for our purposes, how do these considerations affect political leaders' decisions on war or peace? 22 The use of case-study methods to explore the decision-making processes of political leaders during important international crises would help answer some of these questions. Another question raised by the assumption that war diminishes trade, one that has important implications for how we test that hypothesis, is how quickly dyadic trade resumes after the termination of war. Although some scholars assume that trade will resume immediately after cessation of wars, and take evidence of such resumption as supporting the war-impedes-trade hypothesis (Anderton and Carter 2001), liberal theory offers no clear predictions with regard to trade after the ending of hostilities (Barbieri and Levy 1999, 2001). It is certainly possible that the cessation of hostilities will lead to a rapid recovery of prewar trade by removing economic and political impediments to free trade. Postwar trade may even exceed prewar levels, if prewar trade was partially depressed by the perceived risks of a coming war. On the other hand, the economic-opportunity-cost argument might suggest that business and political leaders fear not only the loss of trade during war but also a disruption of the trading relationship and a slow recovery of trade in the years after the end of war. The vision of a community of shared interests created and reinforced by trade (Deutsch et al. 1957) may be shattered by war and be very slow to recover in its aftermath. The same is true of preferential trading arrangements (Mansfield and Pevehouse 2000) and other institutions that facilitate economic interchange. In addition, trade may be lost to third parties and not easily recovered, depending on both the structure of trade and on economic ideology.23 A delay in the recovery of trade is also influenced by the political context surrounding the end of formal hostilities. If the underlying conflicts between states are not resolved by war—and they often are not, particularly in wars between enduring rivals—the occurrence of war may only enhance the perceived risks of a subsequent war and keep trade depressed. Economic restraints designed to prevent the economic and military resurgence of the defeated adversary (such as those imposed on Germany after World War I) may further depress trade. Alternatively, the victor in war might impose conditions that facilitate an increase in trade, as illustrated by the United States in its relationships with Germany and Japan after World War II. Or, the motivation for war in the first place may be to open up markets, resources, or trade routes. In Page 137 → short, the resumption of trade following war is critically dependent on the political context in which the war ends (Barbieri and Levy 2001). Given the ambiguity of the predictions of the economic-opportunity-cost hypothesis as to the rate of recovery of trade after the termination of war, it would be premature to use trade data in the aftermath of war as evidence for or against the hypothesis. We need to specify the implications of the economic-opportunity-cost hypothesis more precisely before that hypothesis can be tested empirically. This discussion of the termination of war raises another issue as well. Almost all of the theoretical and empirical literature on economic interdependence and war looks at the impact of interdependence on the outbreak of war. It would also be useful to examine the impact of interdependence on the termination of war. The logic of the economic-opportunity-cost argument implies that the hope of recovering the gains from trade should be one factor motivating political leaders to end a war. In fact, there may be theoretical grounds for expecting that economic calculations have a greater impact on decisions to seek an end to war than they do on decisions regarding the possible initiation of war.24 Recall Blainey's (1988) argument, refined by Fearon (1995) and others, that disagreements about relative power

are the central cause of war, that war serves to eliminate uncertainty about the dyadic balance of military power between adversaries, and that peace comes when adversaries agree on the dyadic balance of power and therefore on the likely outcome of the war. The onset of war is normally characterized not only by uncertainty about the outcome of war but also by enormous uncertainty about the economic costs and opportunity costs of war, the public's reaction to those costs, and its political and diplomatic ramifications. Just as the course of war reduces uncertainty about relative military power, it also reduces uncertainty about the economic costs of war and hence reduces misperceptions about relative resolve. This increases the likelihood of adversaries finding negotiated solutions that they each prefer to a continuation of a costly war. I suspect there are a number of cases in which leaders who were contemplating war concluded that the economic costs of war would be tolerable, decided on war in part for that reason, came to realize that the economic costs of war (and the public's adverse reaction to those costs) were much greater than they had anticipated, and sought to end war in part for economic reasons. One example comes from the War of the Spanish Armada (1585–1604) between England and Spain. While there is some evidence that Elizabeth and possibly Philip II attempted to assess the likely economic costs of war (Parker 1998, 356), Page 138 → strategic considerations clearly dominated economic ones in their decisions for war. But the economic costs of war, and particularly the costs resulting from restrictions on trade between the two countries, turned out to be enormous. The impact on merchants' revenues and on daily life led to substantial and creative efforts to circumvent state restrictions on trade; to widespread warweariness in both England and Spain; and to public pleas for peace, at least in Spain. By the end of the 1590s Elizabeth and Philip were each “swimming against a strong-running economic tide,” and the expectation that peace might bring unimpeded trade was “one of the most persuasive arguments” in favor of ending the war in the late 1590s (Croft 1989, 296–99).25 The idea of war reducing uncertainty about the economic costs and opportunity costs of war raises the interesting question of how leaders learn, or update, their beliefs about those costs. This is quite complex, because these costs are endogenous to state strategies regarding the extent to which trade with the enemy and neutrals should be prohibited and to the course of the war. If political leaders allow trade with the enemy to continue, the economic opportunity costs of war will be minimized, but if they impose embargoes and other economic sanctions, those costs of war will be increased. Those decisions will affect the domestic support for the war effort and for state policies regarding trading with the enemy—support by key economic groups; by organized political opposition; and by the public, which may be motivated both by concerns about the economic costs of war and by the symbolic appeals of patriotism. These changing levels of domestic support in turn affect state policies with regard to trading with the enemy. This suggests that state policies with respect to trading with the enemy are the product of a series of decisions rather than a one-time decision and that these policies change during the course of the war as a function of changes in the military, economic, and domestic political context. One example comes from the United States in the War of 1812. After blocking several efforts by President Madison to cut off trade with Britain, in December 1813 Congress passed a sweeping embargo aimed at stopping the flow of supplies to British armies in Canada and Europe and to British fleets in American waters, tightening the nonimportation system, preventing the British from fraudulently using neutral flags, and stopping the use of ransoming as a cover for illegal trade. Although the embargo reduced trade with the enemy, it also reduced U.S. government revenues, cut into prosperity, and generated a swell of protests, particularly in New England, and Congress was forced to back off from some of the harsher provisions (Hickey 1989; Levy and Barbieri 2001). The analysis of the evolution Page 139 → of state policies on trading with the enemy, and the reaction to those policies by key economic groups and the public, could be profitably explored through longitudinal case-study analyses.

Third-Party Considerations Explanations for the trade-promotes-peace hypothesis, whether based on the logic of economic opportunity costs or signaling games, are basically dyadic-level explanations.26 Proponents of this hypothesis usually ignore the systemic context in which trade and financial flows take place, particularly the role of diplomatic alignments and alliances. This is a critical omission, because the logic underlying the economic-opportunity-cost and signaling-

game models does not necessarily imply that economic interdependence always promotes peace when third parties are involved. At the dyadic level, economic interdependence between A and B may help to reduce the probability of war between A and B. But if B threatens C, and if A simultaneously has trade ties with B and an interest in maintaining C's security against B, A's economic ties with B may prevent A from attempting to deter B's attack on C or at least raise the threshold of threat before which A attempts to deter B. As a result, economic interdependence between A and B may increase the probability of war between B and C (and possibly even draw in A). Whereas the 1914 case is often seen as a violation of the liberal trade-promotes-peace hypothesis because war occurred despite high levels of interdependence, the logic of economic opportunity costs and signaling combined with the logic of diplomatic alignments helps to explain that outcome.27 Britain's failure to make a formal commitment to join France and Russia if they were attacked by Germany, which many historians argue was a critical factor leading to the German decision for war (Fischer 1967), is explained in part by the extensive economic ties between Britain and Germany prior to the war and by Britain's hesitation to alienate its key trading partner by attempting to restrain that partner's strategic ambitions (Papayoanou 1999).28 In the context of third parties and of alliances in particular, the dyadic economic-opportunity-cost and signaling models are incomplete. The theoretical implications of those models must be further developed by incorporating the interaction effects between economic interdependence and diplomatic alignments and alliances. Until that is done, behavior in nondyadic cases such as World War I cannot be used to confirm or disconfirm dyadic models of economic interdependence and conflict. Page 140 →

Conclusion I have argued that the logic of the opportunity-cost model is essentially monadic in nature and that for this reason it is incapable of explaining the impact of economic interdependence on war or other forms of international conflict, which are dyadic outcomes. The model is also incapable of explaining the impact of economic interdependence on the likelihood of war in a multiactor system in which the actors in question also have important security ties with other actors. The economic-opportunity-cost model requires a theory of bargaining that explains how trading partners interact in situations involving serious conflicts of interests, and the model must also incorporate the interaction effects between economic interdependence and diplomatic alignments if it is to explain the impact of economic interdependence in multiactor systems involving conflicting security interests. These limitations, along with other ideas developed in this chapter, suggest that a satisfactory theory of economic interdependence and conflict must incorporate variables traditionally associated with both the liberal and realist traditions in international relations theory. Such a theory must incorporate concerns about the opportunity costs associated with the loss of trade, the influence of multiple domestic actors who have interests in increasing trade and the political power to influence state decisions, the role of governments in discouraging or encouraging trade, and the outcome of state-societal bargaining.29 The theory must also incorporate adversaries' conflicting strategic interests, the security externalities of trade, the impact of economic interdependence on the dynamics of alliance behavior, and the systemic security consequences of the loss of trade during war. Thus we need to move away from the tendency to frame the debate over interdependence and conflict in terms of the paradigmatic clash between liberalism and realism. While the “paradigm wars” probably contributed to the growing interest in the question of the relationship between economic interdependence and international conflict, it has become more and more clear that neither liberal nor realist theories by themselves can provide a satisfactory analysis of this relationship. Scholars have increasingly begun to recognize the limitations of traditional liberal and realist paradigms for the analysis of the relationship between economic interdependence and war. Papayoanou (1999), for example, builds on realist theories of balancing behavior, liberal theories of the dyadic and domestic factors linking economic

interdependence to peace, and the logic of signaling games to construct a general theory of great power behavior. Russett and Page 141 → Oneal (2001) call for a conceptual synthesis of Kantian and realist theories and incorporate variables from each theoretical tradition in their statistical models. Game-theoretic models of interdependence and conflict have demonstrated the analytic limitations of traditional formulations of the liberal economic-opportunity-cost hypothesis, recast that hypothesis in terms of models of signaling in the context of strategic uncertainty, and in the process built on insights from the literature on the causes of war. The result in each case is a significant increase in the explanatory power of our theories of economic interdependence and international conflict and several useful frameworks for further theoretical and empirical research on these questions.

NOTES I want to thank Erik Gartzke and the editors of this volume for their helpful comments on an earlier draft of this chapter. 1. There may be some important differences in the impact of trade flows and financial flows on international conflict, and scholars need to give more attention to the distinctive impact of financial interdependence (Gartzke, Li, and Boehmer 2001). But most of the existing theoretical literature uses “trade” interchangeably with economic interdependence, and for convenience I refer to the “trade-promotes-peace” hypothesis in its broader sense. 2. For an example of the role of the identification and testing of new theoretical implications in the development of the democratic-peace research program see Russett and Starr 2000. 3. For surveys of alternative explanations of the trade-promotes-peace phenomenon see Stein 1993; Doyle 1997; Barbieri and Schneider 1999; Russett and Oneal 2001; and Mansfield and Pollins (this volume). 4. For reviews of the empirical literature and some of the controversies, see McMillan 1997; Barbieri and Schneider 1999; and Mansfield and Pollins (this volume). 5. On the debate over the appropriate statistical techniques see Lake and Gourevitch 2001. 6. Similarly, the Manchester liberals also spoke primarily of the role of economic interdependence in reducing the likelihood of war in international politics (Blainey 1988). For evidence on interdependent dyads having more nonmilitarized disputes, see Gartzke 2002b. 7. Significant exceptions are Mansfield 1994; and Barbieri 1996, 2002. For a quick summary see table 2 in Barbieri and Schneider (1999, 395) and also Mansfield and Pollins (this volume). 8. The few studies that have looked at both war and MIDs find that any (negative) association between economic interdependence and broader measures of conflict breaks down when the dependent variable is operationalized in terms of war (Barbieri 1996). Page 142 → 9. This is a version of the old stability-instability paradox at different levels of conflict (Snyder 1965). See also Morrow's discussion of signaling-game models in this volume. 10. Oneal and Russett (1999) also incorporate the effects of international institutions, but their measure of the number of joint memberships in intergovernmental organizations is not sufficiently discriminating to explain the impact of institutions on international conflict. More useful in this regard is Mansfield and Pevehouse's (2000) study of the impact of preferential trading arrangements (PTAs), which include freetrade areas, common markets, customs unions, and other economic-related institutions. Mansfield and Pevehouse (2000) find (for the 1950–85 period) that for states in the same PTA, there is a strong negative relationship between trade flows and militarized disputes, whereas for states that do not belong to the same PTA there is only a weak relationship between trade flows and disputes. 11. Copeland (1996) applies this argument to the period leading up to World War I, which is generally viewed as the classic anomaly in the liberal theory of peace because of the high levels of interdependence before the war (Waltz 1979). Copeland argues that despite high levels of trade, German leaders expected trade to decline and that these expectations led them to resort to force to secure long-term access to markets and raw materials. 12. Oneal and Russett (1997) capture part of this idea when they include a measure of the trend in economic interdependence as well as its level.

13. Another interesting question here concerns the impact of trade on the duration of war (Barbieri and Bremer 1998), which in turn affects the impact of war on trade. 14. This section builds on Levy 2002a. 15. Scholars rarely define hegemony with much precision, and the ambiguities are compounded by the different conceptions of hegemony in the security and political-economy literatures and by the failure of scholars to acknowledge those differences (Levy 2002a). In addition, scholars rarely fully develop the causal linkages from hegemony to peace. The implicit assumption is that the hegemon is a liberal hegemon that is committed to a liberal world economy and also possesses sufficient military power to deter leading challengers and maintain the peace among weaker states. These attributes may have been shared by the United States in the late twentieth century and perhaps by Britain in the mid–nineteenth century, but they are not always congruent. The Dutch, for example, dominated the world economy in the early seventeenth century but lacked the attributes of a military leader who could enforce the peace. 16. Recent empirical studies (Oneal and Russett 1999; Russett and Oneal 2001; Mansfield and Pevehouse 2000) have included a control variable to capture the degree of economic hegemony in the system. In focusing on the impact (presumably negative) of hegemony on conflict, however, these studies neglect the positive impact of hegemony on trade flows and thus fail to deal with the possibility that the observed negative relationship between trade and conflict is spurious. Similarly, well-designed quantitative studies incorporate alliances into the model to capture the effects of alliances on trade and war. 17. An interesting question here is whether different forms of economic interdependence Page 143 → (e.g., trade, capital, or monetary) have different utilities for purposes of sending credible signals. 18. For more detailed discussions of different signaling models of economic interdependence and conflict, see the chapters by Erik Gartzke, James D. Morrow, and Arthur A. Stein in this volume. Morrow suggests a useful test between alternative signaling models, based in part on when signaling should take place (before or after a conflict becomes militarized). Morrow suggests an empirical model based on events data, which provides a more nuanced set of categories than the militarized interstate dispute data set. This approach might be usefully supplemented by using case studies to trace the process of interactions between partners /adversaries over the evolution of the dispute. On recent treatments of the strengths and limitations of casestudy methods see George and Bennett (forthcoming), Brady and Collier (forthcoming), and Levy (2002b). 19. Liberals assume that dyadic trade drops significantly with the onset of war between trading partners but usually are not always explicit about why this occurs. If trade between adversaries stops primarily because state leaders prohibit it for fear that the adversary will reap a disproportionate share of gains from trade and convert those gains into military power (the realist relative-gains argument), the distinctively “liberal” character of the liberal economic theory of war comes into question. Arguments based on the impact of increased transportation and insurance costs on incentives to trade would be more consistent with fundamental liberal assumptions. 20. Realist theories of economic interdependence and war also predict that trade between adversaries will be significantly reduced during wartime. Thus evidence of substantial trading with the enemy would not help to empirically distinguish between liberal and realist theories of economic interdependence and international conflict. Evidence bearing on alternative explanations for the trading-with-the-enemy phenomenon and the conditions under which it is most likely to occur might give more support to one paradigm than to the other. 21. For evidence of trading with the enemy in recent wars in Bosnia and in Chechnya see Judah 1997 and Lieven 1998, respectively. 22. These questions suggest the need for a theory of the state and of state-societal relations, as Beth Simmons emphasizes (this volume). 23. An example of economic ideology reinforcing the idea that it might be difficult to win back a trading partner that has shifted to other markets and suppliers comes from eighteenth-century Britain and the concept of “channels” of trade. In the words of one contemporary analyst, “When trade is once lost, it will be too late by a mistimed care, easily to retrieve it again, for the currents of trade, like those of waters, make themselves channels, out of which they are afterwards as hard to be diverted as rivers that have worn themselves deep within their banks” (quoted in Pares 1963, 405). 24. Selection effects associated with the onset of war between interdependent dyads should affect the conduct and termination of war. Given the economic opportunity costs of war, only the most highly resolved states will be willing to fight (Gartzke 2002a). As a result, we would expect wars between

economically interdependent dyads to be Page 144 →more intense or longer than wars between less interdependent dyads. A signaling model would predict that greater interdependence leads to shorter wars. 25. The war did not end until 1604, which raises questions regarding the impact of economic factors relative to other variables in the termination of war. In the case of England, economic arguments for peace were validated by the boom in Anglo-Spanish trade after the end of the war in 1604, which became the “single most important feature” of England's commercial expansion in the early seventeenth century (Croft 1989, 297). 26. An important exception is Dorussen (1999), who constructs a multiactor model and finds that the pacifying effects of trade diminish rapidly with a larger number of actors. 27. Copeland (1996) argues that key actors believed that economic interdependence would decline in the future, reducing the economic-opportunity costs of war. 28. The linkages between security and political economy are further reinforced by the fact that variations in the strength of the Franco-Russian alliance against Germany in the decade leading up to World War I (a loosening in alliance ties after 1905 and an increase after 1911) was more highly correlated with the variation in economic ties between each of those states and Germany than with changes in the balance of power and perceptions of the German threat (Papayoanou 1999, chap. 4). In strategic terms, the weakening of Russia after its 1904 defeat by Japan should have contributed to a further tightening of the FrancoRussian alliance. 29. Some domestic groups may have incentives to restrict trade.

REFERENCES Anderton, Charles H., and John R. Carter. 2001. The Impact of War on Trade: An Interrupted Times-Series Study. Journal of Peace Research 38:445–57. Barbieri, Katherine. 1996. Economic Interdependence: A Path to Peace or a Source of Interstate Conflict? Journal of Peace Research 33:29–49. _____. 2002. The Liberal Illusion: Does Trade Promote Peace? Ann Arbor: University of Michigan Press. Barbieri, Katherine, and Stuart A. Bremer. 1998. Economic Interdependence and Dispute Escalation. Manuscript. Barbieri, Katherine, and Jack S. Levy. 1999. Sleeping with the Enemy: Trade between Adversaries during Wartime. Journal of Peace Research 36:463–79. _____. 2001. Does War Impede Trade? A Response to Anderton and Carter. Journal of Peace Research 38:619–24. Barbieri, Katherine, and Stuart A. Bremer. 1998. Economic Interdependence and Dispute Escalation. Manuscript. Barbieri, Katherine, and Gerald Schneider. 1999. Globalization and Peace: Assessing New Directions in the Study of Trade and Conflict. Journal of Peace Research 36:387–404. Blainey, Geoffrey. 1988. The Causes of War. 3d ed. New York: Free Press. Page 145 → Brady, Henry E., and David Collier. Forthcoming. Rethinking Social Inquiry: Diverse Tools, Shared Standards. Berkeley: Policy Press, University of California; Lanham, MD: Rowman and Littlefield. Conybeare, John A. C. 1987. Trade Wars: The Theory and Practice of International Commercial Rivalry. New York: Columbia University Press. Copeland, Dale C. 1996. Economic Interdependence and War: A Theory of Trade Expectations. International

Security 20:5–41. Croft, Pauline. 1989. Trading with the Enemy, 1585–1604. The Historical Journal 32:281–302. Deutsch, Karl W., Sidney Burrel, Robert Kann, Maurice Lee, Martin Lichterman, Raymond Lindgren, Francis Loewenheim, and Richard van Wagenen. 1957. Political Community and the North Atlantic Area. Princeton: Princeton University Press. Dorussen, Han. 1999. Balance of Power Revisited: A Multi-Country Model of Trade and Conflict. Journal of Peace Research 36:443–62. Doyle, Michael W. 1997. Ways of War and Peace: Realism, Liberalism, and Socialism. New York: W. W. Norton. Fearon, James D. 1995. Rationalist Explanations for War. International Organization 49:379–414. Fischer, Fritz. 1967. Germany's Aims in the First World War. New York: W. W. Norton. Gartzke, Erik. 2002a. Burning Bridges or Building Bonfires: Selection, Signaling, and the Liberal Peace. Manuscript. _____. 2002b. Interdependence Really Is Complex: Assessing the Liberal Peace Using Events Data. Manuscript. Gartzke, Erik, Quan Li, and Charles Boehmer. 2001. Investing in the Peace: Economic Interdependence and International Conflict. International Organization 55:391–438. George, Alexander L., and Andrew Bennett. Forthcoming. Case Studies and Theory Development. Cambridge, MA: MIT Press. Gilpin, Robert. 1981. War and Change in World Politics. Cambridge: Cambridge University Press. _____. 1987. The Political Economy of International Relations. Princeton: Princeton University Press. Gowa, Joanne. 1994. Allies, Adversaries, and International Trade. Princeton: Princeton University Press. Hickey, Donald R. 1989. The War of 1812: A Forgotten Conflict. Chicago: University of Illinois Press. Hirschman, Albert O. [1945] 1980. National Power and the Structure of Foreign Trade. Reprint, Berkeley: University of California Press. Jones, Daniel M., Stuart A. Bremer, and J. David Singer. 1996. Militarized Interstate Disputes, 1816–1992: Rationale, Coding Rules, and Empirical Patterns. Conflict Management and Peace Science 15:163–213. Judah, Tim. 1997. The Serbs: History, Myth, and the Destruction of Yugoslavia. New Haven: Yale University Press. Page 146 → Keohane, Robert O. 1984. After Hegemony: Cooperation and Discord in the World Political Economy. Princeton: Princeton University Press. King, Gary, Robert O. Keohane, and Sidney Verba. 1994. Designing Social Inquiry. Princeton: Princeton University Press. Lakatos, Imre. 1970. Falsification and the Methodology of Scientific Research Programmes. In Criticism and the Growth of Knowledge, ed. Imre Lakatos and Alan Musgrave, 91–196. New York: Cambridge University Press.

Lake, David A., and Peter Gourevitch. 2001. Symposium: Research Design and Method in International Relations. International Organization 55:439–507. Levy, Jack S. 2002a. The Study of War and Peace. In Handbook of International Relations, ed. Walter Carlsnaes, Thomas Risse, and Beth A. Simmons, 350–68. London: Sage. _____. 2002b. Qualitative Methods in International Relations. In Millennial Reflections on International Studies, ed. Michael Brecher and Frank P. Harvey, 432–54. Ann Arbor: University of Michigan Press. Levy, Jack S., and Salvatore Ali. 1998. From Commercial Competition to Strategic Rivalry to War: The Evolution of the Anglo-Dutch Rivalry, 1609–1652. In The Dynamics of Enduring Rivalries, ed. Paul F. Diehl, 29–63. Urbana/Champaign: University of Illinois Press. Levy, Jack S., and Katherine Barbieri. 2001. Trading with the Enemy during Wartime: Theoretical Explanations and Historical Evidence. Manuscript. Lieven, Anatol. 1998. Chechnya: Tombstone of Russian Power. New Haven: Yale University Press. McMillan, Susan M. 1997. Interdependence and Conflict. Mershon International Studies Review 41:33–58. Mansfield, Edward D. 1994. Power, Trade, and War. Princeton: Princeton University Press. Mansfield, Edward D., and Jon C. Pevehouse. 2000. Trade Blocs, Trade Flows, and International Conflict. International Organization 54:775–808. Morrow, James D. 1999. How Could Trade Affect Conflict? Journal of Peace Research 36:481–89. Morrow, James D., Randolph M. Siverson, and Tressa E. Tabares. 1998. The Political Determinants of Trade: The Major Powers, 1907–1990. American Political Science Review 92:649–61. Oneal, John R., and Bruce M. Russett. 1997. The Classical Liberals Were Right: Democracy, Interdependence, and Conflict, 1950–1985. International Studies Quarterly 41:267–93. _____. 1999. The Kantian Peace: The Pacific Benefits of Democracy, Interdependence, and International Organization, 1885–1992. World Politics 52:1–37. Papayoanou, Paul A. 1999. Power Ties: Economic Interdependence and War. Ann Arbor: University of Michigan Press. Pares, Richard. 1963. War and Trade in the West Indies. London: Frank Cass. Page 147 → Parker, Geoffrey. 1998. The Grand Strategy of Philip II. New Haven: Yale University Press. Polachek, Solomon W. 1980. Conflict and Trade. Journal of Conflict Resolution 24:57–78. Pollins, Brian M. 1989. Does Trade Still Follow the Flag? American Political Science Review 83:465–80. Ray, James Lee. 1990. Friends as Foes: International Conflict and Wars between Formal Allies. In Prisoners of War? ed. Charles S. Gochman and Alan Ned Sabrosky, 73–91. Lexington, MA: D. C. Heath. Reuveny, Rafael. 2001. Bilateral Import, Export, and Conflict/Cooperation Simultaneity. International Studies Quarterly 45:31–58. Reuveny, Rafael, and Hejoon Kang. 1996. International Trade, Political Conflict/Cooperation and Granger Causality. American Journal of Political Science 40:943–70.

Ripsman, Norrin M., and Jean-Marc F. Blanchard. 1996–97. Commercial Liberalism under Fire: Evidence from 1914 and 1936. Security Studies 6:4–50. Rogowski, Ronald. 1989. Commerce and Coalitions: How Trade Affects Domestic Political Alignments. Princeton: Princeton University Press. Russett, Bruce M., and John R. Oneal. 2001. Triangulating Peace: Democracy, Interdependence, and International Organizations. New York: W. W. Norton. Russett, Bruce M., and Harvey Starr. 2000. From the Democratic Peace to Kantian Peace: Democracy and Conflict in the International System. In Handbook of War Studies II, ed. Manus I. Midlarsky, 93–128. Ann Arbor: University of Michigan Press. Snyder, Glenn. 1965. The Balance of Power and the Balance of Terror. In The Balance of Power, ed. Paul Seabury, 185–201. San Francisco: Chandler. Solingen, Etel. 1998. Regional Orders at Century's Dawn: Global and Domestic Influences on Grand Strategy. Princeton: Princeton University Press. Stein, Arthur A. 1993. Governments, Economic Interdependence, and International Cooperation. In Behavior, Society, and International Conflict, vol. 3, ed. Philip Tet-lock, Jo Husbands, Robert Jervis, Paul Stern, and Charles Tilly, 241–324. New York: Oxford University Press. Walker, Thomas. C. 2000. Peace, Rivalry, War: A Theoretical and Empirical Study of International Conflict. Ph.D. diss., Rutgers University. Waltz, Kenneth N. 1979. Theory of International Politics. Reading, MA: Addison-Wesley.

Page 148 →

Liberal Hopes with No Guarantees Gregory D. Hess One learns in undergraduate economics about the gains from international trade. Countries differ in their endowments, resources, factors of production, and so on and therefore can gain by specializing in the production of specific goods so that through trade both countries can be made better off. This is probably one of the central tenets of economics. Of course, a change in the trade regime between two countries creates winners and losers within each economy. However, the benefits of international trade are such that there is enough, in principle, so that all the winners can be compensated for their losses while still leaving a gain to some members of society. Economics can be pretty silent, however, on the specifics of how the losers from free trade are actually compensated for their losses. Such compensation schemes could include job retraining, subsidies to declining or expanding industries, tax abatements, and so on. In the United States these compensation schemes are often limited since these programs are expensive; they must be paid for with taxes that distort economic decisions; and they water down the market signals that encourage workers to reallocate to new industries, occupations, and/or regions. Moreover, aggrieved groups may be powerful enough to bring about tariffs and other trade sanctions that may limit the gains from trade as well as invite retaliatory actions by trading partners. In a sense, multiple equilibrium levels of trade and welfare may persist even though they are dominated by free trade. Hence, from a political-economy perspective, while free trade may provide a tremendous opportunity for countries to make their citizens better off, there is no guarantee that it will do so. Institutions, carefully crafted, can help societies coordinate on these better trade outcomes. Current international trade Page 149 → institutions, while imperfect, have helped to facilitate an increase in world trade throughout the postwar period. As pointed out by Douglass C. North (1990, 34), “How well institutions solve the problems of coordination and production is determined by the motivation of players (their utility function), the complexity of the environment, and the ability of the players to decipher and order the environment (measurement and enforcement).” We classical liberals can continue to hope that trade between countries will continue to grow, though there are no guarantees that our institutions are up to the task. This logic of considering institutions that will help societies coordinate on Pareto-preferred outcomes pervades the theoretical work that I have done with my coauthor Athanasios Orphanides with regard to the relationship between war and regime type (Hess and Orphanides 2001b). In our work we explore the Kantian hypothesis that if the world were populated solely by democracies, then a state of perpetual peace would necessarily prevail.1 We also discuss some theoretical aspects of institutions that can help to facilitate a more peaceful world. In this short chapter, I would like to describe and elaborate on these issues, as well as consider the role that trade can play in fostering a more peaceful environment. But before considering institutions, one must first consider the individual behavior of state leaders who can make the decision to engage in conflict.2 In its simplest form, the model in our work assumes that all countries' leaders are democratically elected and treats conflict as always ex ante costly to consumers but potentially useful to incumbent leaders whose domestic economic performances are lagging.3 But, one might ask, how can wars be helpful to leaders, ex ante costly to citizens, and nevertheless sustainable with rational voters? The modeling idea for how an individual country behaves is as follows. With some probability, regardless of a president's term of office and circumstances, an unavoidable conflict will confront a country. The model also assumes that leaders of a country are characterized by both their economic-handling abilities, which are perfectly observable to all citizens, and by their war-handling abilities, which are only revealed to the leader and citizens if in fact a war takes place. The preferences of a leader also differ from those of the citizens since leaders enjoy the rents from being in office and would prefer to keep enjoying these rents, which requires them to get reelected. Hence, if the economy is doing poorly and an unavoidable war has not occurred, then the democratic leader of a country may initiate an avoidable war. In doing so, he or she would be gambling that he or she will be revealed to be so good at

handling war (i.e., minimizing the costs of war) that voters will now prefer to keep him or her even with his or her bad economic-handling characteristic. Would voters go Page 150 → along with this? Actually, a rational, forwardlooking voter would go along with it; having leaders with good war-handling abilities is important since in the future there is a constant probability that an unavoidable war will occur. The two keys to the story of how an individual democratic country behaves are the leader's desire for officespecific rents—what we label “selfishness”— and the extent to which the probability of future unavoidable war allows the leader to create additional avoidable conflicts to gamble for resurrection. Both elements must be present in order for rational diversionary conflicts to be undertaken by democratically elected governments. But while selfish leaders are here to stay, one wonders from where this underlying threat of unavoidable wars derives. The key to understanding this aspect of the model is to realize that while each individual country is small and does not by itself affect the overall level of unavoidable war, when all countries behave this way they together create the supply of unavoidable conflict.4 An interesting implication of this analysis is that multiple equilibrium levels of worldwide conflict can occur. Most importantly, however, the Kantian equilibrium always exists. To see this point, notice that if the probability of unavoidable war is zero, then democratic leaders with poor reelection prospects have no reason to create avoidable wars since the public will not place value on good war-handling abilities if there are no future avoidable wars. This equilibrium gives liberals hope, especially if institutions can aid in coordinating on this Kantian equilibrium. The problem is that other equilibrium frequencies of worldwide conflict may exist as well. Three contributing factors can be identified. First, the more selfish and office driven democratic leaders are, the more prone they are to take advantage of any diversionary war opportunities. Of course, if more leaders behave this way, then such diversionary behavior will be self-fulfilling, since in a war-prone environment, voters place a high weight on leaders with good war-handling abilities. The second reason why the Kantian equilibrium is not unique is that the world is not populated just by democracies but rather is also populated by non-democratic leaders who have the possibility of engaging in appropriative war opportunities. We believe that wars have some ex ante economic gains to offer nondemocratic leaders as they can appropriate the benefits of conflict while deflecting the costs of conflict to the citizenry. This possibility, in contrast, is not available to democratic leaders, as the costs and benefits of conflict are likely to be more evenly distributed across leaders and citizens. The final contributing factor to why conflict may persist throughout even a fully democratic world is whether the expected opportunity cost of conflict is Page 151 → large. Large? It's not a typo. The intuition is as follows. Recall that a diversionary underpinning to economic conflict is possible if war-handling skills are valued. In this model, the larger the expected losses from conflict are, the greater the gains are from finding a leader who is good at minimizing the cost of conflict. Hence, as the demand for good war-handling leaders rises, leaders will more often start (i.e., “supply”) diversionary conflicts to improve their reelection chances. Unfortunately, that greater economic opportunity costs can lead to more conflict does not bode well for trade as a facilitator of peace. Why? First, consider the case where bad outcomes in a war result in the loss of current trade opportunities that could have enhanced welfare. Forgone beneficial trade opportunities raise the expected costs of conflict, which in turn raises the value that society places on leaders with good war-handling abilities. Accordingly, leaders will seek to create more diversionary-conflict opportunities if the loss in trade rises for leaders who are not very good at handling wars. The finding that larger economic opportunity costs could make conflict more likely is an interesting twist on the “economic-opportunity-cost” approach to conflict reviewed in the chapters by Jack S. Levy (this volume) and James D. Morrow (this volume). In particular, these authors point out that while the traditional economicopportunity-cost argument suggests that larger economic losses from conflict should reduce the likelihood of conflict (since there is more to lose), larger losses could also make potential adversaries less willing to make

concessions in a bargaining setting. The net effect on conflict would then be indeterminate. While Morrow explores this literature in his chapter, note that in our nonbargaining framework, where engaging in conflict provides an important signaling role only to domestic citizens about the leader's war-handling ability, the relationship between economic opportunity cost and the likelihood of conflict is further complicated. The reason is simple. While more costly wars make citizens want to face fewer of them (the traditional economic-opportunitycost argument), selfish leaders (who actually decide whether or not to engage in a conflict) can gain more by “gambling for resurrection” if they reveal good war-handling ability. Therefore, somewhat paradoxically, this diversionary-conflict channel makes foreign trade raise the probability of conflict. Hence, in the first model scenario, the presence of trade may raise the worldwide frequency of conflict, not lower it. So we are back to the world of multiple equilibria, where peace can be sustained but so too can conflict. Let's consider a second case, namely, that countries that fight wars with one another can lose out both on current trade opportunities with one another and on Page 152 → future trade opportunities with one another. Perhaps this norm can be brought about by tariffs or trade sanctions. Will this norm help the world coordinate on a more peaceful equilibrium? Given our framework, this is not likely. Why? First, the fact is that the world provides many potential trading partners, and hence the loss of trade with one country can be overcome by finding another trading partner whose goods are closely substitutable. Second, finding alternative trading partners gets easier as time progresses and as transportation costs continue to decline and inputs become more elastic in the production process. Thus, trade sanctions between two countries will not be a credible disciplining device to stop diversionary conflict, as the countries can adopt alternative trading partners and traded goods.5 There is still not enough hope for classic liberals to cling to. However, what if countries coordinated to isolate or punish countries that started conflicts? Would this help the world to coordinate on a more peaceful equilibrium? It turns out that this can work. Why? The reason is that, in this model, an individual country is “small” in the economic sense, which means that it takes as given the environment that it is working in. Hence, an individual country creates a strategic externality when it decides to initiate a diversionary conflict, since by making the world more war-prone it increases the demand by citizen voters in all countries for leaders with good war-handling skills. Of course, making a leader or country internalize this negative externality provides the only hope of credibly altering its behavior. Indeed, institutions can be designed and implemented to bring about such a desired result. Consider the following institutional scenario that Orphanides and I put forth to help implement the democratic peace (Hess and Orphanides 2001b). Let's suppose that a fraction of all leaders forms an institutional alliance such that they agree not to initiate a conflict with any country and also agree to share in the costs of a retaliatory conflict in the case that any one of them is attacked. Hence if a nonalliance country attacks a country that is part of the alliance, then they will be attacked in the following period with probability one and forced to lose further welfare. That the promise for the alliance to retaliate is credible will be true if the alliance is relatively large so that the cost per member of the attack is relatively small, compared to the large benefit of no longer worrying about being attacked. In contrast, if a nonalliance country attacks another nonalliance country, then while the country that has been attacked may threaten to retaliate, the fact of the matter is that such retaliation is not credible. First, conflict is very costly to citizens (as nonalliance members will have no one to share the costs with). Second, the country cannot be guaranteed that another nonalliance country Page 153 → will not attack them in the future. Hence, the costs of retaliation are higher and the benefits are smaller for nonalliance countries than for alliance ones. These factors taken together, a nonalliance country will prefer to attack another nonalliance country rather than an alliance country. Indeed, a well-functioning, credible alliance can keep the probability of conflict at zero for its members, whereas the propensity for conflict among nonallied countries likely will not be zero. In essence, while an alliance can make a single country internalize the strategic externalities from conflict, a single nonallied country cannot. Can trade play a part in facilitating institutions in order to reduce the propensity for conflict? Indeed it can. Recall

from the earlier discussion that if reduced current trade raises the expected costs of conflict if a leader is bad at handling conflict, then trade may make the world less peaceful since it raises the demand by the electorate for leaders with good war-handling skills. However, if a fraction of countries creates a credible institution that forms both a trade and a military alliance, then these countries can enjoy the economic and welfare benefits of trade and more easily fund a joint retaliation in case of being attacked. Indeed, strong trade ties will make a military alliance more formidable and better able to sustain a perpetual peace within the alliance as the gains from maintaining the joint alliance are increased by trade. In essence, a collective security arrangement is the alliance structure that our theory indicates would be helpful for keeping aloft our liberal hopes. This type of arrangement has been forcefully advocated and defended by Kupchan and Kupchan (1995). In particular, they argue that to the extent that nationalism and domestic politics are important determinants of a state's reasons for considering a conflict, as opposed to, say, an impending and dramatic security concern, a collective security arrangement will temper the desire to engage in conflict. However, while collective arrangements may be impossible to legally formalize in a world with sovereign states, “other institutional formulations, such as concerts, that rely on looser and more informal regulation of balancing” (Kupchan and Kupchan 1995, 53) will likely provide a practical and flexible approach to their implementation. Again, while evolving institutions provide no guarantee for realizing our liberal hopes, they likely provide the only possibility that they can be realized. Of course, even if one were to believe that collective security arrangements would be beneficial for sustaining peace, a number of important questions beg for further exploration. First, how large should this military alliance be? Indeed, this question hinges on the extent to which countries need to share other common interests (language, culture, etc.) in order to forge a stronger union. The larger the institutional alliance, the harder it may be to forge a consensus Page 154 → (or for that matter unanimity) on important matters. Second, if multiple alliances form, will they compete with one another to such an extent that sustainable conflict reemerges as a worldwide phenomenon? On these important questions, the analysis I have described is currently silent. The final question I have in mind is, how similar or dissimilar should the production ability of the alliance members be? For instance, should just oil producers form a trade and military alliance, or should alliances be formed whose members by themselves can benefit from the gains from trade? On this question, I think the paradigm outlined earlier can help us begin to think about these issues. On the one hand, it would appear that the more economically and militarily self-sufficient the alliance is, the better able it will be to both finance a collective defense and benefit from a lasting peace. On the other hand, if countries that monopolize a strategic trade input (e.g., oil) have tremendous economic clout and form a type of trade association (like the Organization of Petroleum Exporting Countries [OPEC]), then the creation of a military alliance against outsiders may be unnecessary. Of course, sustaining their monopoly position may be difficult if each country in the trade association tries to increase its own market share (as has happened numerous times in OPEC's history). Moreover, trade associations' members may engage in conflict with one another, as OPEC members have in a number of instances (e.g., Iran and Iraq in the 1980s; Kuwait and Iraq in the early 1990s). In summary, a thorough consideration of the theoretical issues underpinning the role that trade has on lessening conflict is overdue. Key issues involve the proper specification of trade linkages across countries and the strategic formation of both military and trade institutions that can foster trade and perpetuate peace. While we can be optimistic that better institutional design can lead to a more peaceful world, understanding both what motivates leaders and the environment in which these leaders must make decisions is crucial. We can still hope to get a better handle on these issues, though we have no guarantees.

NOTES I would like to thank Edward D. Mansfield and Brian M. Pollins for very helpful comments on an earlier draft of this chapter. This chapter was prepared while the author was an academic consultant to the Federal Reserve Bank of Cleveland. The opinions expressed in this chapter are those of the author and should not be

taken as opinions of the Federal Reserve Bank of Cleveland or of the Federal Reserve System. 1. This chapter builds off of our earlier work (Hess and Orphanides 1995, 2001a, 2001b). Page 155 → 2. We only consider external conflicts in our “rational diversionary” theory. Of course, internal conflict between regions within a country and the linkages with the role of trade within a country are also topics of interest. 3. Of course, a more sophisticated treatment of both the nature of and the demands on the state is a crucial part of the relationship among regime-type, conflict, and trade (see Simmons, this volume). We present some formal theoretical results in Hess and Orphanides 2001b. 4. This is similar to the assumption, often used in economics, that each country is small. In other words, each country behaves as if it were insignificant, but when all identical countries behave this way they together affect the worldwide level of unavoidable conflict. In some sense there is a strong sense of anonymity in this model that abstracts from the “natural” or “historical” enemies or friends that we observe in neighboring countries. Keith Poole once labeled this model the smorgasbord approach to international relations: countries always have an infinite and anonymous available menu of other possible countries with which to engage in conflict. 5. Of course, as pointed out by Simmons (this volume), the actual nature of the goods traded may be very important for the linkage between commerce and conflict. For instance, an impact on bilateral trade for goods that are very price inelastic will alter countries' expected welfare levels much differently than if the goods were very price elastic.

REFERENCES Hess, Gregory D., and Athanasios Orphanides. 1995. War Politics: An Economic, Rational Voter Framework. American Economic Review 85:828–46. _____. 2001a. Economic Conditions, Elections, and the Magnitude of Foreign Conflicts. Journal of Public Economics 80:121–40. _____. 2001b. War and Democracy. Journal of Political Economy 109:776–810. Kupchan, Charles A., and Clifford A. Kupchan. 1995. The Promise of Collective Security. International Security 20:52–61. North, Douglass C. 1990. Institutions, Institutional Change and Economic Performance: Cambridge: Cambridge University Press. Page 156 →

Page 157 →

Economic Statecraft Page 158 →

Page 159 →

Violence and Disease: Trade as a Suppressor of Conflict when Suppressors Matter Bruce Russett This chapter is divided into three related sections. It begins with an attempt to clarify and resolve some disputes about the evidence concerning the relationship between trade and conflict. This is the most technically dense part of the chapter, but its basic point is that those of us interested in theoretical or policy questions about trade and conflict must insure that the analyses we use as evidence take into account issues of measurement and statistical analysis that cannot be left solely to technical specialists. It contends that, if the appropriate technical decisions are taken, the broad statement that trade reduces international conflict is sound—and also that in some of the technical debates the answer is not very different whether one takes one side or the other. Then, guided by the first discussion, the chapter offers a conceptual formulation, adapted from the medical literature on disease, about trade as a suppressor of violent conflict in the face of serious inducements to violence. Finally it applies the evidence and the new conceptualization to a major policy problem—the future of relations between China and the United States. That relationship will increasingly be in need of powerful suppressors to conflict, and getting the policy right will test the quality of theory and evidence on virtually all the issues discussed in this book.

Some Apparent Controversies about the Relation between Trade and Conflict Comparing Studies I believe that the evidence that trade reduces conflict forms a strong and robust generalization. Edward D. Mansfield and Brian M. Pollins's introduction to Page 160 → this volume is a serious and fair-minded effort to lay out the convergences and divergences of the literature. Its presence allows me to focus on some particular issues without attempting comprehensiveness. I suspect, however, that it and some other contributions overstate the divergence in the literature, once some critical questions about conceptualization (and, for productive conceptualization, also measurement) are addressed. I will focus on those issues. In doing so it is important to resist the formulation “on the one hand A shows, but on the other hand B disagrees” without confronting the reasons for discrepant results. Not all studies are created equal in quality. One must weigh the balance of evidence and also give greater weight to more recent work if it seems to address deficiencies or ambiguities in earlier research. Moreover, it is easy to draw the wrong inference from a poorly or incompletely specified model. Sometimes the differences in results or interpretations arise from genuine conceptual or theoretical differences. Too often, however, they stem from incomplete theorizing or a poor fit of theory to the model that purports to estimate it. More rigorous thinking—in dialogue with other researchers—shows that some of the alleged problems are really nonissues. Finally, it is essential that all data used in published analyses are made available at the time of publication. We simply cannot know anything if we have no opportunity to replicate each other's findings. When these considerations are taken into account, the broad generalization that interdependence diminishes violent conflict holds. Following are some of the issues that in my view need clarification or future attention. What Is the “Right” Measure of Trade? John Oneal and I (Russett and Oneal 2001, and our earlier articles) contend that for spatially and temporally broad analyses the best general measure of state- or system-level interdependence is total trade/gross domestic product (GDP). For dyads, or pairs of states, we believe the appropriate theoretically derived measure is the value of trade within the dyad to the GDP for each dyad member. The trade/GDP ratio captures, in a crude way to be sure, the relative importance of trade not only to the economy but, ceteris paribus, its relative weight in the political system. Virtually all the broad theoretical hypotheses about the effect of commerce on conflict—about rational actors in the domestic political system, about communication and signals, about building identities—implicitly or explicitly

assume that the effects are stronger the larger the relative share of international commerce in the domestic economy of the state. And for the dyad, the most important datum is the trade/GDP ratio of the less dependent state, on the theoretically driven assumption that this less dependent Page 161 → state is the “weak link” in avoiding conflict because it is less constrained by trade from initiating or escalating a dispute. Without controlling for the widely varying size of each economy, the raw value of trade alone tells us little about its potential to restrain conflict. Asymmetries A focus on the weak link should not blind us to the problem of asymmetric dependence—how much the ratio for the weak link differs from that for the stronger link. Trade asymmetry has been an important concern from mercantilist times and in the modern era from Albert Hirschman ([1945] 1980) onward. But we (Oneal and Russett 1999a, 1999b; Russett and Oneal 2001, chap. 4) do not find that asymmetric trade systematically changes the risk of conflict. Overwhelmingly, it is the size of the weak link that matters. This somewhat surprising result looks very robust, at least as a broad generalization and baseline finding. It may be, however, that in enough highly asymmetric cases the small, highly dependent, and constrained partner may avoid any action that could provoke the large, less dependent, and less constrained partner. This could be true in less developed countries where both the economy and the political system are dominated by elites profiting from extractive industries or plantation agriculture. The conflict-damping effect in such cases might on average balance the conflict-inducing effects produced when more revolutionary elites dominate the system. Some case evidence (e.g., events in Cuba, Guatemala, Iran, Nicaragua) would support this hypothesis, and more systematic research is definitely in order. Again, this says there is no systematic general relationship between asymmetric trade and violent conflict but leaves open the possibility that particular kinds of cases may cancel each other out. Better theory about additional conditioning variables, inclusion of those variables in the large-scale analyses, and attention to case studies may illuminate the matter. Mediating or Confounding Variables Handling potential third-variable problems is crucial to adequate conceptualization. An example concerns geographical proximity. The bottom line is that a proper specification of the trade-conflict problem must always include adequate measures of distance or proximity. Proximity promotes conflict, providing both opportunity and disputed issues. We all know this. But proximity also promotes trade, by reducing transport costs and time in transit. We know this, too, and it is built into all the gravity models to explain trade patterns. Consequently, in any bivariate analysis trade is positively related to conflict. If we fail to control for proximity we may get silly answers and inferences—that is, trade Page 162 → causes conflict or conflict causes trade. Samples limited to “politically relevant dyads” reduce this problem by including only contiguous dyads or great powers (for whom, in some sense, all states are within reach). Still, the strongest— and most correctly specified—relationship emerges when one controls, even among politically relevant dyads, for an interval measure of distance. And for politically nonrelevant dyads, without distance one will get the silly answer. I believe Oneal and Russett 1999a settled this; Oneal and Russett 1999b and Russett and Oneal 2001 confirm it. It is not just a matter of taste. Other complex interactions may affect the relation between trade and conflict. For instance, there may be an interaction between trade and level of development, and another between democracy and development, in affecting conflict. Mousseau (2000) and Hegre (2000) think so, at the level of the lowest 10 percent of the world's states on the GDP per capita scale. Mousseau and especially Hegre are cautious in making this claim, but it needs to be further investigated. If supported it would not invalidate the generalization but would condition it. Mansfield and Pevehouse (2000) have found an interactive effect between trade and participation in a preferential trade area, perhaps with international organizations helping to settle disputes short of violence. Other kinds of interactions can be hypothesized from careful theory and should be investigated. Measurement: What to Do about Zeros for Trade Once more, this problem combines theoretical issues with measurement ones. The standard data sources on dyadic

trade report many entries as missing data. But that cannot be taken at face value. Many times it just means the value of trade between two states (perhaps small and/or distant from each other) was really zero or so close to zero as not to matter. This is particularly likely when no trade is reported for dyads in conflict. When states enforce tight restrictions on trade with adversaries—as across the old Cold War divide or between Israel and its Arab neighbors—zeroes can appropriately be imputed. Lack of reported trade between Iraq and Israel, or North Korea and the United States, does not mean missing data. To treat it as missing data would mean dropping many of the pairs in conflict—that is, introducing selection bias on the dependent variable, potentially reducing the apparent effect of trade on conflict and vice versa. Our solution was, where one or both states have reporting obligations to the International Monetary Fund (IMF), and when neither state reports exports or imports to or from the other, to treat it as a true zero, or close enough to a true zero. If neither state was an IMF member, however, the Page 163 → conservative solution is to code the data as missing. Nonetheless, in large-scale analyses this turns out to be an issue of limited empirical importance.1 Fortunately, statistical routines like AMELIA (King et al. 2001) for multiple imputation of missing values now can improve on these somewhat ad hoc rules. Finally, John R. Oneal shows in his chapter (this volume) that a new data set compiled by Kristian Gleditsch (2002) eliminates the missing-data problem— and reconfirms our previous findings. Other Economic Variables The focus of this review on trade does not mean that other variables (e.g., flows and stocks of investment, the elasticities for price and substitution of trade in various commodities) are unimportant. The focus derives rather from the fact that such information is much less generally available and, where available, is only for states and times (e.g., great powers, industrialized democracies in recent decades) that represent a biased sample. These biases are potentially serious, so one must be very careful about generalizing. Small and unrepresentative samples can suggest a great deal about process, causal mechanisms, and compounding or confounding effects of third variables. But broad generalizations expressing probabilities require measures that cover a long time period and a large and diverse sample—close to the universe of dyads, or of politically relevant dyads. Limitation of data coverage (e.g., no elasticities for substitutes) does lose us some important information, but much of the time in large-N studies its implication is primarily that trade/GDP is a noisy measure. If we nevertheless find a strong relationship between dyadic trade/GDP and conflict despite the noise—as we have—it makes that basic relationship more plausible rather than less. Then one can use more limited samples with the other variables needed to refine or condition the generalizations, though usually not to throw the generalizations out. The Dependent Variable Most studies have relied primarily or solely on militarized international dispute (MID) data as the dependent variable. The MID data do suffer from serious limitations of conceptualization and measurement. It is important to look just at wars as well as at all MIDs and, as a compromise between the two, at fatal disputes. So far we find very similar patterns for wars and fatal disputes and, if anything, stronger results for fatal disputes than for all MIDs (Russett, Oneal, and Cox 2000; Oneal this volume). Depending on one's theory, it may be important to distinguish between influences on the initiation of disputes and Page 164 → influences on their continuation, though neither we nor Bennett and Stam (2000) have found it makes much empirical difference save perhaps when asking which member of the dyad was the initiator. For some questions of war and violence events data like those from the Conflict and Peace Data Bank (COPDAB) can be useful, if analyzed in the context of careful theory. The full COPDAB scale covers a wide range of events from close cooperation through various forms of conflict. Like several contributions to this book (e.g., those by Erik Gartzke, James D. Morrow, and Arthur A. Stein), many microtheories of behavior do not expect the relationship between trade and conflict to be the same throughout a range of conflict from purely diplomatic disputes up through various levels of violence. Cooperation is not simply the obverse of conflict—interacting states will have some of both. We may reasonably expect trade to reduce military conflict, but not necessarily all manifestations of conflict short of violence. Even closely cooperating economically interdependent states inevitably experience some conflicts, especially over the conditions under which their goods and services are permitted to enter the other's market. But such conflicts need not imply any readiness to resort to military violence

in order to resolve them. Some theories of signaling do suggest that diplomatic threats to restrict trade can, by indicating resolve, help to resolve those conflicts short of any escalation to violence (Gartzke, Li, and Boehmer 2001). If that were commonly true, we might expect merely diplomatic conflicts (fairly low on events data scales) to be especially common even among those interdependent states that remain at peace with each other. Other theories and recent evidence, however, do not support the idea that trade-related threats often imply a readiness to use military force. Using COPDAB supplemented by events data, we do find that the dispute-preventing effects of trade are a little weaker for nonmilitarized diplomatic disputes than for MIDs. Nevertheless, even these diplomatic disputes are significantly rarer in dyads that are economically interdependent than for otherwise comparable noninterdependent dyads (Kinsella and Russett 2002). Problems in Analyzing Time Series Time series are essential for many substantive and theoretical questions, but they raise their own problems. One is whether to control for lagged dependent variables, for example, the effect of previous disputes on current disputes or previous trade on current trade. This methodological question too cannot be divorced from conceptual and theoretical concerns. Previous disputes will predict to current disputes. In part this may be because the legacy of previous disputes Page 165 → truly causes current ones. But usually the same influences (e.g., distance, autocracy, little trade) that caused the present dispute also caused past disputes. Any imputation of causality to lagged dependent variables thus may be largely illusory. So including past values as a “control” will sop up most of the variance attributable to the theoretically interesting variables. Christopher Achen (2000) raises this matter forcefully. He shows that if serious serial correlation is present, a combination of bias from the lagged dependent variable and that of a key exogenous variable brings big interpretive trouble. This is relevant to our concerns, since both trade and conflict are heavily trended. A lagged dependent variable will not only pick up effects that should be attributed to excluded variables but can also “steal” the effect of independent variables that are included in the equation, thus dominating the regression results even when it has little true explanatory power. This is part of the larger question of how to control for temporal dependence. I try to defer to others on complex technical issues that are still in development and dispute. Some of the disputes may have great consequence for the empirical results on trade and conflict, but often they turn out to be much ado about little. With good specification and high-quality data over a wide range of time and space the differences appear miminal.2 Like measurement issues, seemingly technical issues almost always depend on careful theoretical specification, and if the results do diverge, a decision about which to accept must depend on how well they fit our most selfconsciously developed theories. Concluding This Section Many other questions about trade and conflict remain at issue. They include questions about the relative strength of reciprocal influence (subject to good tests with distributed lag analysis [Oneal, Russett, and Berbaum 2003], vector autoregression models, and perhaps simultaneous equations), the effect of third variables and complex interactions, and better theory and theory-testing on the processes and mechanisms behind the empirically observed relationships. But overall, with appropriate conceptualization from which should derive appropriate methods and measurements, on broad and long time-series data the general conclusion that trade reduces the probability of violent conflict is, in my view, strong and robust—like the conflict-reducing effect of democracy.

Inducements and Suppressors to Conflict I now turn to a more theoretical discussion of various influences on the risk of conflict. In our book (Russett and Oneal 2001, chap. 3), as well as in our earlier Page 166 → writing, we employed an analogy between epidemiologists' statistical analysis of large-scale databases to identify risk factors for disease and quantitative international relations researchers' comparable use of large-scale databases to identify risk factors contributing to or mitigating violent international conflict. Using such an approach to the experience of the period 1885–1992 we

found substantial and significant reductions in risk associated with the hypothesized influences. Among those (Russett and Oneal 2001, 171) were familiar realist variables such as an alliance and a wide imbalance of power between two states (a 40 percent and a 36 percent reduction in risk, respectively). In addition, we found similar strong effects when both states were democratic, were economically interdependent, and shared membership in many international organizations (33, 43, and 24 percent reductions in risk) and a more than 70 percent reduction in the risk of conflict when all three of what we call these Kantian influences were present. We also discussed theoretical reasons why each of these influences should matter individually and, for the Kantian variables, collectively as part of an interacting system. But while we spoke of risk factors in terms of incentives and constraints on conflict, we lacked an explicit unifying statement tying them all together. The epidemiological analogy actually suggests a more unified perspective. In his recent book, How Scientists Explain Disease, Paul Thagard (1999) distinguishes between environmental inducements of mutation on the one hand and bodily suppressors of harmful mutations on the other. It is the combination of these two kinds of influences that accounts for the progression from simple exposure to the development of many diseases in acute form. This insight offers promise for understanding international relations as well, if we divide our list of influences on the occurrence of military conflict into two such groups as they apply to pairs of states. Among the well-known inducements to conflict are major power status, proximity, contiguity, and a near-equal power ratio.3 The presence of one or more of these conditions in any pair of states can be considered a nearnecessary condition for a “mutation” or event: a serious diplomatic dispute. Without them, the two states are in large degree politically irrelevant to each other for purposes of violent conflict: they have few incentives to fight or opportunities to exert military power against each other. But with them, they have the potential to develop political and diplomatic disputes that may under certain conditions escalate into a MID involving the threat or use of military force. The other variables are more usefully considered as potential suppressors, the weakness of one or more of which allows the diplomatic dispute “mutation” to become a disease (MID). They include the Kantian influences of joint Page 167 → democracy, high mutual trade relative to GDP, economies generally open to trade with many nations, and many shared intergovernmental organization (IGO) memberships. To these should be added being allied.4 All of these influences are usually analyzed as additive terms, but it may prove more theoretically productive to consider the two sets of influences— inducements and suppressors—as two separate vectors, each in turn composed of a string of additive terms. The disease, a militarized international dispute, becomes the product of these two additive vectors linked by a multiplicative term. Conceptually, the implication is that there would be no MID without one or more inducements and the weakness or absence of one or more suppressors. Having a mediating but as yet unobserved variable is not problematic in a conceptualization, and it may be possible to move toward an adequate empirical indicator. Measures of serious but not yet militarized diplomatic disputes have been used in several analyses, and an appropriate way to do so for these questions is as follows. COPDAB and related data sets contain a wealth of information on low-level conflicts as well as militarized ones. Using these scales, one can identify many diplomatic disputes as manifesting “strong verbal hostility” but not, in the great majority of cases, escalating further to become militarized disputes. Diplomatic disputes could be expected to arise frequently in dyads that are geographically proximate or contain a major power, for example, but such disputes' escalation into MIDs with the threat or use of military force would typically happen only if the suppressors to escalation were weak. This formulation allows us to consider MIDs as arising from two conditions that are jointly necessary (inducement and weak suppressors), and within each vector the weakness or absence of any one of the suppressors could be a strong and necessary (but not sufficient without at least one in the other vector) condition of the emergence of an MID. This new conceptualization may be more theoretically satisfying and produce different empirical results than the previous all-additive formulation. It also makes more theoretical use of the distinction between politically relevant dyads and others; engages the conceptual distinction between opportunity (inducements) and willingness (the absence of suppressors) for conflict; and draws on recent attention to problems of selection bias, signaling, and the use of two-stage models to address those problems (Kinsella and Russett 2002).

Additionally, it helps elucidate the interaction among distance, conflict, and trade—a problem that, as noted earlier, has confused some discussions. Geographical proximity is strongly related to conflict, providing both opportunity and incentives. In the absence of proximity between homelands, contiguity, the presence of a major power in the pair, and balanced power also facilitate violence, Page 168 → but proximity alone is a powerful influence. Proximity also facilitates trade. Thus proximity promotes trade—a suppressor of violent conflict—but also directly induces conflict. Unless one controls for this direct influence of proximity on conflict, trade could appear (spuriously) to be a cause of violent conflict rather than a suppressor.

The Role of Trade in Suppressing and Managing Sino-American Conflicts The previous formulation can help clarify one of the major debates about policy for future international relations. In an anarchic international system, regional rivals (as in the Middle East) and major powers will have many potential inducements (opportunity) for conflicts of interest and diplomatic disputes, a potential likely to be aggravated by relative equality of power. With the continuing presence of these inducements well recognized by realist theory, it is all the more important for largely liberal suppressors of escalation (willingness to fight) to be strong and numerous and for theorists and policymakers to identify what suppressors may be available and effective. In debates about the future of Sino-American relations, liberals often advocate trade with China as a way to promote peaceful relations. By contrast, realists typically argue that expanding trade with China will only strengthen it for a coming confrontation with the West. Since realists view mutual trade relationships as at best irrelevant to the risk of militarized conflict—and at worst a cause of discord in a world where the concern for relative power and benefit rules—they see closer trading ties only as strengthening China's economic, military, and technological base for an eventual violent conflict, with no peace-inducing benefits. Liberals, by contrast, view mutual trade as a powerful incentive to manage and suppress the escalation of inevitable diplomatic disputes. The potential for a major clash between the United States and China in coming decades is a real one, the most likely and most dangerous great power confrontation. Its possible magnitude compels the most careful analysis. Not only do the two countries have substantial issues in dispute, of which the status of Taiwan is the most obvious. In addition, differential rates of growth between the United States and China (in the 1990s the rates in GDP growth were approximately 3 percent and 10 percent, respectively) make it likely that a perilous era of “power transition” (Organski and Kugler 1980; Kugler and Lemke 2000) will emerge toward the end of the first quarter of this century. The dangers inherent in a power transition are well recognized. For these two states, the major power status of each, and the relative equality of power between them, Page 169 → will serve as serious inducements to conflict despite the distance separating them. Realists' agreement with that assessment leads them to wish to minimize trade so as to minimize Chinese power. American efforts to limit trade are not likely to achieve much in a world with so many other states ready to jump in to replace American commerce with China. This lesson emerges from experience with extended boycotts and sanctions. Unless there is a widespread sense of threat and institutionalized cooperation among allies, the pursuit of private national interests severely limits the coherence of collective action to maintain a common front (Martin 1992). Economic sanctions have been applied in many circumstances for many purposes, and one need not accept the most skeptical views of their effectiveness to see their limitations (Hufbauer, Schott, and Elliott 1990; Pape 1997). Assessments even of the American-led sanctions effort during the Cold War differ widely in judgments of their success (Spero and Hart 1997, chap. 10; Baldwin 1985, chap. 9; Funigiello 1988). They did limit Soviet access to Western military technology, but their effect on the overall Soviet economy is more disputable, and the trade restrictions were loosened over time. They were tightest during the first two decades of the Cold War—the very time when the Soviet Union was growing most rapidly. Unless there should develop a collective interpretation of Western relations with China as a new Cold War no less threatening than the old one, if the United States reduces its trade with China, Europe and Japan are likely to pick up the difference, even in much strategic technology. Russia has already concluded major export and coproduction deals with China for technologically advanced weapons. Even if Sino-American political relations remain positive there is good reason for the United States to restrict export of its most militarily sensitive goods and technologies

to China. But there is little likelihood that broader trade restrictions, and the constriction of normal commerce on principles of comparative advantage, will significantly reduce Chinese power from what it would be with more trade. The evidence that unilateral limitations on trade in general can substantially reduce an adversary's power potential over the long run is very weak—much weaker, indeed, than the evidence that mutual trade can reduce the risk of violent conflict.5 Certainly the burden of proof is on those who would restrict such trade. Given the presence of two of the general inducements to conflict in the U.S.–China relationship (relative equality of power and major power status), the importance of powerful suppressors is evident. Yet despite substantial privatization of the economy and some relaxation of political restrictions, China is no democracy and is unlikely to become one in the next decade or two. Common Page 170 → international-organization memberships probably provide more suppressive capability, but the United States and China share few common IGO links outside of the quasi-universal ones of the expanded United Nations system. Global economically oriented IGOs are, however, an important exception, especially with China's accession to the World Trade Organization. It is hard to imagine continuing integration of China into this institutional network in the face of a significant U.S. effort to restrict China's access to the global economy. A systematic constriction of U.S.–China commercial exchange, and a retreat from the general economic and institutional openness of China's economy to the world, would mean weakening the most promising suppressors to violent conflict. In their chapter Christopher Gelpi and Joseph M. Grieco suggest that the interactive effect of trade and democracy is more powerful than simply an additive relationship would suggest. Our analyses have not supported that hypothesis. Should that interaction be convincingly demonstrated the combination would be better than either alone for reducing militarized conflict, but even in the absence of democracy trade may still be important. Foreign trade already represents a greater proportion of China's total economy (GDP) than it does of Japan's. Since many powerful elements of China's economic and political leadership benefit from that trade, it is worth strengthening their numbers and their incentives to maintain and expand it. The trade-to-GDP ratio for the two states is asymmetrical, larger for China than for the United States. Yet as noted earlier, there is little evidence that asymmetry in a trade relation bears any systematic relation to the likelihood that violent conflict will erupt between the trading partners. It is also possible that China's further integration into the world economy may strengthen the forces favoring its democratization, though this result is less certain than that greater economic ties will directly reduce the likelihood of violent conflict. In this volume Michael Mastanduno reminds us that economic engagement is a strategy and that the conditions for success of that strategy will vary. In a politically mixed pair like the United States and China, American democracy provides some advantages. Among other things, Lisa Martin (2000) shows that democratic politics, with open debate and decision, make it easier (not harder) for a government to make credible long-term economic commitments. Other evidence emerges that in mixed pairs democracies are less ready to initiate disputes than are their autocratic partners (Huth and Allee 2003). Since the United States is the status quo power, its concern should be to dissuade China from initiating or escalating a dispute. So the underlying problem is to devise incentives for the Chinese leadership not to use or threaten to use military force Page 171 → even as the balance of power shifts in their favor. Enmeshment in global networks of economic dependence changes the trade-offs leaders have to think about when they are tempted to pursue adventurous foreign policies. States that fight with others can be brutally punished by the global market, and their base for economic growth can vanish. It is worth remembering that the modern international system has experienced at least one peaceful power transition between a leading state and its closest rival. That was when, at the close of the nineteenth century and the beginning of the twentieth, the United States overtook the United Kingdom as the world's strongest economic power and potentially strongest military power. At that time they shared one very important suppressor to conflict (their shared democratic institutions and cultures), even though international organization at that time was in its infancy. Also, their commercial ties were strong. At that time—when their suppressor effect was most needed—U.S. and British trade and investment ties were very important to the national income of each country, far more than they would be for most of the next century (Russett 1963, chap. 3). Commerce provided not only a thick base of private interest in peaceful relations but also a conduit of information, communication, and the

development of shared perceptions and values. The policy implications of expanding or contracting China's economic linkages with the United States and other democratic and capitalist countries are therefore evident. A strategy of limiting trade would do little or nothing to diminish the incidence of the diplomatic and political inducements to such conflict. It might well exacerbate them. Any effect it might have to slow the growth of China's power would likely be minimal. And it would come at the cost of abandoning an alternative strategy with a powerful suppressor of violent conflict under circumstances when powerful suppressors would be badly needed. It could be a cosmic mistake.

NOTES This chapter reflects my long-term collaboration with John R. Oneal as well as his comments on an earlier draft. I am most grateful for both. I also thank several organizations for financial support, most recently the Ford Foundation and the Weatherhead Initiative on Military Conflict as a Public Health Problem. Of course neither John nor funding agencies are responsible for what I say. 1. See the correction on page 2 of Journal of Peace Research 20 (1999) of table 2 as erroneously typeset in Oneal and Russett 1999a. A corrected version of this article appears in the same volume as our commentary on it (Oneal and Russett 2003). Page 172 → 2. One such dispute is over controlling for time dependence by a spline correction or the generalized estimating equation (GEE). Both require assumptions that may be problematic (Alt, King, and Signorino 2001). Beck, Katz, and Tucker (1998) are sometimes cited as refuting Oneal and Russett's (1997) claim of a statistically significant relationship between trade and conflict in the period 1950–85. Beck, Katz, and Tucker's conclusion holds, however, for only a few analyses. GEE and the spline correction give substantially the same results for 1950–92 for politically relevant dyads, for all dyads when zero trade values are imputed for IMF members, and whether one uses the onset of disputes or their incidence (including continued disputes) as the dependent variable (Oneal and Russett 1999a). Similarly, for the long century beginning in 1885, Russett and Oneal (2001) and Oneal (this volume) find few differences with the two methods. Bennett and Stam (2000) draw similar conclusions. Debate about the desirability of using fixed-effects models in estimations of a rare-event binary variable (e.g., MIDs) has been decisively settled in the negative (Green, Kim, and Yoon 2001; vs. Beck and Katz 2001; King 2001; Oneal and Russett 2001; Bennett and Stam 2000; Heagerty, Ward, and Gleditsch 2001). 3. See Wagner 2000 for a rationalist explanation of why war is more probable in closely balanced dyads than in unbalanced ones, contra Waltz 1979. 4. It is also possible that a very unequal power ratio between the two states should be considered a suppressor of violent conflict, on the grounds that the weaker state will be deterred from any act that might provoke the stronger one to violent action. Yet this reasoning can be pushed back to the inducement stage, in that relatively equal power relations are likely to induce states to raise serious diplomatic issues in a realist world precisely because of the uncertainties of the impact of a power balance. Moreover, the power ratio shares with the other inducements the characteristic of being relatively stable and fixed, only slowly if at all subject to change by deliberate policy choice. The suppressor variables, by contrast, all are more clearly subject to political decision and change. For the sake of this discussion I leave the power ratio with the inducements, subject to further theorizing and empirical analysis. 5. See Morrow 1997 on the rarity of conditions when relative gains impede trade.

REFERENCES Achen, Christopher. 2000. Why Lagged Dependent Variables Can Suppress the Explanatory Power of Other Independent Variables. Paper presented at the annual meeting of the Political Methodology Section of the American Political Science Association, July, Los Angeles. Alt, James E., Gary King, and Curtis S. Signorino. 2001. Aggregation among Binary, Count, and Duration

Models: Estimating the Same Quantities from Different Levels of Data. Political Analysis 9:21–45. Baldwin, David. 1985. Economic Statecraft. Princeton: Princeton University Press. Page 173 → Beck, Nathaniel, and Jonathan N. Katz. 2001. Throwing the Baby Out with the Bath-water: A Comment on Green, Kim, and Yoon. International Organization 55:487–96. Beck, Nathaniel, Jonathan N. Katz, and Richard Tucker. 1998. Beyond Ordinary Logic: Taking Time Seriously in Binary-Time-Series-Cross-Section Models. American Journal of Political Science 42:1260–88. Bennett, D. Scott, and Allan Stam III. 2000. Research Design and Estimator Choices for Analyzing Interstate Dyads: When Decisions Matter. Journal of Conflict Resolution 44:653–85. Funigiello, Philip. 1988. American-Soviet Trade in the Cold War. Chapel Hill: University of North Carolina Press. Gartzke, Erik, Quan Li, and Charles Boehmer. 2001. Investing in the Peace: Economic Interdependence and International Conflict. International Organization 55:391– 438. Gleditsch, Kristian. 2002. Expanded Trade and GDP Data. Journal of Conflict Resolution 46:712–24. Green, Donald, Soo Yeon Kim, and David Yoon. 2001. Dirty Pool. International Organization 55:441–68. Heagerty, Patrick, Michael Don Ward, and Kristian Gleditsch. 2001. An Easy Way to Clean a Dirty Pool: Widowing Subseries Empirical Variance Estimators, with Application to Social Science. Paper presented at the 42nd annual meeting of the International Studies Association, February, Chicago. Hegre, Håvard. 2000. Development and the Liberal Peace: What Does It Take to Be a Trading State? Journal of Peace Research 37:5–30. Hirschman, Albert O. [1945] 1980. National Power and the Structure of Foreign Trade. Reprint, Berkeley and Los Angeles: University of California Press. Hufbauer, Gary, Jeffrey Schott, and Kimberly Elliott. 1990. Economic Sanctions Reconsidered: History and Current Policy. 2d ed. Washington, DC: Institute for International Economics. Huth, Paul, and Todd Allee. 2003. The Democratic Peace and Territorial Conflict in the Twentieth Century. New York: Cambridge University Press. King, Gary. 2001. Proper Nouns and Methodological Propriety: Pooling Dyads in International Relations Data. International Organization 55:497–507. King, Gary, James Honaker, Anne Joseph, and Kenneth Scheve. 2001. Analyzing Incomplete Political Science Data: An Alternative Algorithm. American Political Science Review 95:49–70. Kinsella, David, and Bruce Russett. 2002. Conflict Emergence and Escalation in Interactive International Dyads. Journal of Politics 64:1045–68. Kugler, Jacek, and Douglas Lemke. 2000. The Power Transition. In Handbook of War Studies, 2d ed., ed. Manus I. Midlarsky, 129–63. Ann Arbor: University of Michigan Press. Mansfield, Edward D., and Jon C. Pevehouse. 2000. Trade Blocs, Trade Flows, and International Conflict. International Organization 54:775–808. Page 174 →

Martin, Lisa. 1992. Coercive Cooperation: Explaining Multilateral Economic Sanctions. Princeton: Princeton University Press. _____. 2000. Democratic Commitments: Legislatures and International Cooperation. Princeton: Princeton University Press. Morrow, James D. 1997. When Do “Relative Gains” Impede Trade? Journal of Conflict Resolution 41:12–37. Mousseau, Michael. 2000. Market Prosperity, Democratic Consolidation, and Democratic Peace. Journal of Conflict Resolution 42:472–507. Oneal, John R., and Bruce Russett. 1997. The Classical Liberals Were Right: Democracy, Interdependence, and Conflict, 1950–1985. International Studies Quarterly 41:267–93. _____. 1999a. Assessing the Liberal Peace with Alternative Specifications: Trade Still Reduces Conflict. Journal of Peace Research 36:423–32. _____. 1999b. The Kantian Peace: The Pacific Benefits of Democracy, Interdependence, and International Organizations, 1885–1992. World Politics 52:1–37. _____. 2001. Clear and Clean: The Fixed Effects of Democracy and Economic Interdependence. International Organization 55:461–86. _____. 2003. Modelling Conflict while Studying Dynamics. In Globalization and Armed Conflict, ed. Gerald Schneider, Katherine Barbieri, and Nils Petter Gleditsch. Lanham, MD: Rowman and Littlefield. Oneal, John R., Bruce Russett, and Michael Berbaum. 2003. Causes of Peace: Democracy, Interdependence, and International Organizations, 1885–1992. International Studies Quarterly 47:72–94. Organski, A. F. K., and Jacek Kugler. 1980. The War Ledger. Chicago: University of Chicago Press. Pape, Robert. 1997. Why Economic Sanctions Do Not Work. International Security 22:90–136. Russett, Bruce. 1963. Community and Contention: Britain and America in the Twentieth Century. Cambridge, MA: MIT Press. Russett, Bruce, and John R. Oneal. 2001. Triangulating Peace: Democracy, Interdependence, and International Organizations. New York: W. W. Norton. Russett, Bruce, John R. Oneal, and Michaelene Cox. 2000. Clash of Civilizations or Realism and Liberalism Déjà Vu? Some Evidence . Journal of Peace Research 36:583–608. Spero, Joan, and Jeffrey Hart. 1997. The Politics of International Economic Relations. 5th ed. New York: St. Martin's. Thagard, Paul. 1999. How Scientists Explain Disease. Princeton: Princeton University Press. Wagner, R. Harrison. 2000. Bargaining and War. American Journal of Political Science 44:469–84. Waltz, Kenneth. 1979. Theory of International Politics. Reading, MA: Addison-Wesley.

Page 175 →

The Strategy of Economic Engagement: Theory and Practice Michael Mastanduno Much of the attention in political science to the question of interdependence and conflict focuses at the systemic level, on arguments and evidence linking the expansion of economic exchange among states on the one hand to the exacerbation of international conflict or the facilitation of international cooperation on the other. The approach taken in this chapter focuses instead at the state level, on the expansion of economic interdependence as a tool of state-craft. Under what circumstances does the cultivation of economic ties, that is, the fostering of economic interdependence as a conscious state strategy, lead to important and predicable changes in the foreign policy behavior of a target state? Students of economic statecraft refer to this strategy variously as economic engagement, economic inducement, economic diplomacy, positive sanctions, positive economic linkage, or the use of economic “carrots” instead of sticks. Critics of the strategy call it economic appeasement. Whatever one calls it, economic engagement is the subject of renewed interest among political scientists. It has profound policy significance as well. During the first decade following the Cold War, economic engagement proved to be the centerpiece of U.S. foreign policy toward its two most important potential competitors, Russia and China. In both cases, U.S. officials relied heavily on economic instruments in an effort to integrate would-be challengers into a U.S.–centered international order. The United States has also relied on economic incentives, since 1994, in an attempt to dissuade North Korea from breaking out of the nuclear nonproliferation regime. For at least a decade the Washington policy community has been debating seriously whether economic Page 176 → engagement toward Cuba would serve U.S. interests more effectively than the economic-isolation strategy that has been carried out by nine presidents across more than forty years. Similar arguments over the utility of economic incentives as opposed to economic sanctions inform contemporary U.S. policy debates about Iran. The policy appeal of economic engagement is not limited to the United States. For quite some time “civilian” great powers such as Germany and Japan have placed heavy emphasis on economic instruments in their foreign policy strategies. For both states, historical experience and the long memories of neighboring states have circumscribed the utility of traditional diplomatic and especially military instruments. Japanese officials have turned to economic incentives as a substitute means to promote foreign policy interests in relations with the states of Southeast Asia. Germany has long relied on economic diplomacy to serve its objectives in relations with Eastern Europe and the former Soviet Union. Policymakers seem increasingly eager to resort to strategies of economic engagement. Is the appeal of these strategies justified by scholarly research, or is this an example of practice preceding theory? Put another way, does there exist a sound social-scientific basis for the pursuit of economic engagement to serve foreign policy objectives? Our knowledge of the workings of economic engagement is still at a fairly preliminary stage. What we do know thus far leads, at best, to an assessment of cautious optimism. A recent series of case studies suggests that economic engagement can be effective as an instrument of statecraft. States have managed in certain situations to use economic relations to influence the foreign policies even of potential adversaries. Economic engagement is not simply synonymous with economic appeasement. Yet we must also appreciate the difficult conditions that must be met for economic-engagement strategies to succeed. Success requires the precise manipulation of domestic political forces in the target state. It requires some ability to control the effects of interdependence. It requires that domestic politics and the foreign policy of a target state be linked in predictable and desirable ways. And the success of this strategy requires the effective management of domestic political constraints in the sanctioning state. These conditions, outlined subsequently, are difficult to meet individually and all the more so cumulatively.

In the next section I assess the current state of social-science research on positive economic statecraft. The following section explores the conditions that must be met for economic engagement to work effectively. A concluding section Page 177 → briefly examines the policy implications in contemporary U.S. relations with China.

The Current State of Knowledge An examination of the scholarly literature on economic engagement as an instrument of statecraft reveals a striking pattern. Albert Hirschman's 1945 study, National Power and the Structure of Foreign Trade, is widely acknowledged today as a starting point for analysis (Hirschman [1945] 1980). Hirschman argued that the conscious cultivation of asymmetrical interdependence, if conducted strategically by the government of a powerful state, would lead weaker states to reorient not only their economies but also their foreign policies to the preferences of the stronger state. He developed a systematic framework for analysis and applied it to the trading and political relationships between Nazi Germany and its central and southeast European neighbors during the interwar period. Hirschman's study did not generate much of a research program in the decades following World War II. The reason, perhaps, had to do with the general lack of interest among political scientists in the substance of international political economy at the height of the Cold War. Political scientists concentrated on the “high politics” issues of warfare and diplomacy and left the “low politics” issues of economic interaction to the discipline of economics (Mastanduno 1998). Scholars who did focus attention on economic instruments focused more on negative sanctions than on positive inducements. David Baldwin observed in an important article in 1971 that “it is not that political scientists have said the wrong things about the role of positive sanctions in power relations; it is just that they have said little” (Baldwin 1971, 19). Baldwin's remark remained fairly accurate until the end of the Cold War. Over the last decade, however, there has been a sharp increase in the scholarly attention devoted to economic engagement. In a recent issue of Security Studies on the theme of economics and national security, for example, half of the articles focused on the analysis of economic engagement (Blanchard, Mansfield, and Ripsman 2000). The end of the Cold War has much to do with this revival of scholarly interest. The superpowers of the Cold War were, for the most part, economically independent of each other. Relations among major powers after the Cold War promise to be characterized by economic interdependence rather than independence. Once again, economic interactions have become, and will continue to be, an important part of great power politics. In this sense the Cold War is Page 178 → likely to prove to be an anomaly, a break in the “normal” pattern of strategic interaction among major states across politics, economics, and security. How the Cold War ended is also relevant. Economic engagement proved to be a key factor in Gorbachev's calculation that the Soviet Union should accept the risks and consequences inherent in the significant reform of its economy. Although Gorbachev eagerly anticipated the expansion of economic ties with the United States, the Cold War endgame was shaped even more profoundly by German economic statecraft. The West German government spent twenty years using economic incentives to build political confidence and trust in its relations with the Soviet Union. This enabled Germany, at the crucial moment, essentially to buy Soviet acquiescence to German unification on favorable terms with the current provision and future promise of expanded trade and generous credits and loans. The German government agreed to pay, among other things, to move Soviet troops from East Germany and house them back in the Soviet Union (Newnham 2002). Now that political scientists have rediscovered economic-inducement strategies, what is the state of our knowledge? The first point to make is that we are still in the very early stages of scholarship—we don't yet know all that much. Other contributions to this volume demonstrate advances that have taken place in the overall scholarship on interdependence and conflict. Over several decades, the work has become more nuanced theoretically and has made extensive use of both case-based and quantitative techniques in linking cause and effect. The literature on negative economic sanctions has progressed in a similar fashion. A large collection of case studies is now complemented by analysis of a large-N database (Hufbauer, Schott, and Elliott 1991). The fact

that the utility of that database is the subject of scholarly debate (Pape 1997; Elliott 1998) should not deflect us from appreciating its importance in providing a foundation for analysis of whether economic sanctions work. In the area of positive economic statecraft, no comparable database exists. Scholars don't yet agree on even the basic question of how common or widespread the practice of economic engagement actually is (cf. Baldwin 1985 and Drezner 1999–2000). Careful case-based research is only beginning to accumulate. It is fair to say at this point that sticks have been studied far more extensively than carrots. This is problematic in the sense that even a cursory examination suggests that positive economic measures have the potential to be as effective as negative ones, if not more so. Threats and coercion usually inspire resentment and resistance in a target state; rewards and inducements are more likely to prompt a willingness to bargain. Negative sanctions tend to produce the “rally around Page 179 → the flag” effect: as Fidel Castro's Cuba and Saddam Hussein's Iraq demonstrate, leaders can often mobilize internal political support for their regimes by pointing to the existence of an external threat. Economic-engagement strategies do not inspire this type of patriotic coalescence in the target country. Negative sanctions typically require multilateral support in order to be effective; economic engagement can benefit from multilateral support but can also work unilaterally. Finally, negative sanctions, unlike positive measures, carry the risk of escalation to more costly measures. If sanctions fail, leaders face the choice of accepting failure or escalating to military means of statecraft. None of this is intended to suggest that economic engagement works easily or without risk. Economic-inducement strategies clearly raise a moral-hazard problem, particularly when dealing with potential adversaries. No government wishes for the reputation of rewarding bad behavior. A state (or a parent) that “pays” for good behavior in one instance may find itself facing payment demands in subsequent instances. Economic engagement also carries the potential to raise political expectations unrealistically at home. As the experience of U.S.–Soviet détente suggested, the transition from economic warfare to economic cooperation can lead the public to expect a similarly rapid transition from political enmity to friendship, only to be disillusioned when conflicts of interest persist. The point here is not that economic engagement has a high probability of success or is risk or problem free, but that as a tool of statecraft it is worthy of the level of attention paid to negative economic sanctions. This point has been emphasized consistently by David A. Baldwin (1971, 1985), but only recently have scholars responded seriously to his challenge. The case-study work conducted over the last decade suggests that economic engagement can work and even can allow states to satisfy significant foreign policy objectives. Lars Skålnes demonstrates that economic engagement has been a critical component of the grand strategies of great powers over the past two centuries. His case studies include nineteenth-century France and Germany, Great Britain during the interwar years, and the United States after World War II (Skålnes 2000). Paul Papayoanou's examination of nineteenth-century international politics suggests the important role economic-inducement efforts played in bolstering the credibility of alliance commitments (Papayoanou 1999). Randall E. Newnham provides a careful examination of the contribution of German economic statecraft to the Cold War settlement with the Soviet Union (Newnham 2002). Patricia Davis turns attention to post-war West German–Polish relations to show that economic inducements were crucial to Germany's efforts to reassure its weaker and suspicious neighbor (Davis 1999). Jonathan Kirshner and Rawi Abdelal apply Hirschman's framework Page 180 → for economic engagement to an array of cases including U.S. diplomacy toward the Kingdom of Hawaii during the nineteenth and early twentieth centuries (Abdelal and Kirshner 1999–2000). Dale C. Copeland has taken an effective first cut at why U.S. economic engagement toward the Soviet Union seemed to succeed after 1985 but not during the 1970s (Copeland 1999–2000). William J. Long analyzes the effectiveness of economic inducements in cases involving U.S. relations with China (Long 1996). A primary achievement of this new literature is to showcase the past utility and future potential of economic engagement. Thus far the literature has tended to “select on the dependent variable,” that is, to choose cases in which economic inducement was an effective foreign policy instrument. Scholars are attempting to correct the research bias in favor of negative sanctions and to respond to the tendency among international relations practitioners as well as students to denigrate the utility of economic instruments of statecraft. They have demonstrated that economic engagement can work; the useful next step will be to develop systematic comparisons

of cases of success and failure to illuminate the conditions under which economic engagement is likely to be more or less effective.

Domestic Politics and the Logic of Economic Engagement Along those lines, the recent literature has served a second important purpose—to clarify the underlying logic of economic-engagement strategy and to point to some of the likely determinants of success or failure. In a striking convergence, virtually all of the recent studies highlight the linkages between domestic politics and foreign policy strategy as the key factors driving the potential effectiveness of economic engagement. The basic causal logic of economic engagement, and the emphasis on domestic politics, can be traced to Hirschman. He viewed economic engagement as a long-term, transformative strategy. As one state gradually expands economic interaction with its target, the resulting (asymmetrical) interdependence creates vested interests within the target society and government. The beneficiaries of interdependence become addicted to it, and they protect their interests by pressuring the government to accommodate the source of interdependence. Economic engagement is a form of structural linkage; it is a means to get other states to want what you want, rather than to do what you want. The causal chain runs from economic interdependence through domestic political change to foreign policy accommodation. Page 181 → In an intriguing way, the logic of negative sanctions similarly suggests a central role for domestic politics. In the standard logic, external economic pressure deprives the target population of economic welfare benefits. The population, in turn, pressures the government to relieve the pain. The governing coalition, faced with a discontented population on one hand and its foreign policy preferences on the other, becomes internally divided and, if the external pressure is sufficient, eventually capitulates. The problem, of course, is that target societies and governments do not always or even often react in the anticipated fashion. Over thirty years ago Johan Galtung, reflecting on Rhodesian defiance of international sanctions and isolation, termed this account the “naive” theory of economic sanctions (Galtung 1967). Even strategic bombing did not easily lead to popular or governmental capitulation; it was all the more naive to think economic sanctions would have that effect. If economic pressure does not easily lead to capitulation, should we expect economic engagement to lead easily to foreign policy accommodation? It obviously would be imprudent to make that assumption, that is, to expect a simple or predictable relationship among economic expansion, domestic transformation, and foreign policy change. Before assuming that economic engagement will have the predicted (for scholars) or desired (for policymakers) effect, we need to pay particular attention to the following four analytical (and, for policymakers, political) challenges. The first is the inherent unpredictability of target domestic politics and the difficulty of using economic exchange as an instrument to manipulate it. Scholars usefully distinguish nationalist from internationalist coalitions in the domestic politics of a target state (Solingen 1998; Papayoanou and Kastner 1999). These coalitions cut across the state and society. Nationalists are more conservative and inward-looking defenders of the status quo. Internationalists, from the perspective of the outside world, are more progressive; they embrace political and economic reform and the integration of the national economy into the global economy. Economic engagement is intended to strengthen the internationalists at the expense of the nationalists and eventually to tip the balance of domestic political power in favor of the former. It may in fact turn out that way. But it may not—there is no reason to assume the internationalists will always win. Economic engagement threatens the interests of the nationalist coalition. If that coalition prevails, it is likely to lash out against what it perceives as intrusive external influences. Depending on the configuration of target domestic politics, engagement could lead to a more confrontational, rather than a more accommodating, foreign policy. In Page 182 → international relations–theory terms, the potential exists for a security dilemma—the unintended and counterproductive consequences of a strategy designed to increase a state's security.

The experience of the United States in relations with Russia during the 1990s is sobering in this regard. Economic engagement and the integration of Russia into the global capitalist economy were intended to reinforce the position of the reformist coalition led by Boris Yeltsin. By the middle of the decade, it became clear that the balance of political authority in Russia had tipped from the reformers to the nationalists. U.S.–Russian relations soured in the latter part of the decade for numerous reasons, but critical among them was the feeling in Russia that international economic integration had been painful and exploitative. Nationalists, rather than internationalists, seem to have used the experience to consolidate their political position; from the perspective of the United States, Russian foreign policy over time became less rather than more accommodating. The second challenge involves the difficulty of controlling the “flow” of economic engagement. Hirschman's theory presumes a steady, gradual injection of interdependence into the target economy, leading to a gradual, almost imperceptible transformation in domestic politics. The medical analogy is apt; economic engagement is akin to the measured doses of medicine administered by a physician to a patient. The problem is that it is difficult to control the flow and effects of economic interdependence once a target economy begins to open to international economic exchange. The effects of interdependence are more likely to be shocking and disruptive than incremental and stabilizing. The experience of the “Asian tigers” during the 1990s illustrates the point dramatically. Countries opened their financial markets to find portfolio investment first flood in and then abruptly flood out, leaving collapsing currencies, plummeting growth rates, and beleaguered populations in its wake. China, whose financial markets were relatively unexposed to international movements of capital, managed to evade the detrimental effects of the 1997–98 crisis. If it had succumbed, it is difficult to imagine that the foreign policy consequences would have been salutary. The third challenge has to do with the malleability of a target state's foreign policy. The theory of economic engagement assumes that foreign policy is susceptible in some meaningful way to external influence. From the perspective of the initiating state, the strategy implies a trade-off between capabilities and behavior. Expanded economic interaction will strengthen the capabilities of a potential adversary—an outcome state leaders generally wish to avoid. But they are willing to accept that outcome on the expectation that interdependence Page 183 → eventually will lead to a positive change in target-state behavior, that is, to a more accommodating foreign policy. For the government of one state to know, with any degree of confidence, the foreign policy intentions of another is a difficult task. Hans Morgenthau (1978) observed that the ability to distinguish a status quo state from a revisionist state is the classic problem of diplomacy and that the fate of nations hangs in the balance. The logic of economic engagement presupposes that the target state is not unalterably revisionist and that if it is not currently a status quo state, it can be transformed into one. That assumption must obviously be tested on a case-by-case basis. If it is incorrect, and the target state is in fact embarked on a revisionist path, then economic engagement leads to the worst of both worlds—an uncooperative foreign policy and an adversary whose capabilities have been strengthened through international economic exchange. Skeptics of the U.S. engagement strategy toward China fear that Chinese foreign policy is less malleable than U.S. foreign policy officials have generally assumed. The fourth challenge directs our attention from domestic politics in the target state to domestic politics in the initiating state. Even if we assume that the target's foreign policy is susceptible to influence, that the flow of interdependence can be effectively managed, and that the domestic politics of the target can be effectively manipulated, there is still the question of whether the initiating state can actually carry out the engagement strategy effectively. Engagement is a long-term strategy that requires consistency, patience, and perseverance on the part of the initiating state. Some states are better equipped than others to employ it. West Germany, for example, possessed a national consensus, across political party lines and across state and society, on the desirability of engaging the Soviet Union and Eastern Europe during the twenty years preceding the end of the Cold War. Business and government worked closely together to assure a coordinated approach to foreign economic policy. The United States, on the other hand, appears less well equipped. The U.S. political system is fragmented and decentralized. The executive branch, when it manages to reach a consensus within its own ranks, has to share foreign policy authority with an often combative and disagreeable Congress. Business is a powerful force in foreign economic policy, but it shares the societal stage with an array of other interest groups. Interest groups and

their representatives in Congress have the potential to “capture” foreign policy to serve domestic or partisan purposes. This is especially true after the Cold War. Domestic actors perceive less at stake in security terms and are less willing to defer to the executive branch. The U.S. effort to engage the Soviet Union during the 1970s is illustrative of Page 184 → these problems. Richard Nixon and Henry Kissinger began, in their terms, to weave the Soviet Union into a web of interdependence with the hope of modifying its foreign policy. Congress had ideas of its own and proceeded through the Jackson-Vanik Amendment, which tied trade and financial benefits to Soviet emigration practices, to undermine the executive's engagement strategy. Soviet leaders were insulted by what they considered a very public violation of their sovereign autonomy. They abrogated their trade agreements with the United States and subsequently ceased foreign policy cooperation. It does not require great imagination to anticipate a similar dynamic in contemporary U.S.–China relations. Throughout the 1990s, the U.S. executive, with the support of business interests, pursued economic engagement. But critics of engagement—human-rights activists, proponents of religious freedom, and those who fear the growth of Chinese economic and military power and are suspicious of Chinese intentions—have increasingly gained strength. These forces gained the upper hand in the transition from the Clinton to Bush administrations. The simple point is that if economic engagement is a long-term strategy requiring consistent application over time, it is fair to question the extent to which the United States has the capacity and will to carry it out effectively.

Conclusion: What about China? A rich opportunity awaits those scholars who are rediscovering the use of economic inducements as instruments of foreign policy. The expertise of both international relations generalists and area-studies specialists is required to operationalize variables, conduct comparative case studies across time and space, and eventually build large-N databases. This very opportunity for scholarship, however, should be sobering to policymakers eager to employ economic engagement. The strategy can be effective—yet under conditions that we are only beginning to comprehend. U.S. officials pursued economic engagement more or less consistently in relations with China during the 1990s. Should they continue to do so? The answer for the immediate future is yes—but not because economic engagement has a high probability of success. The earlier discussion suggests the difficulty of manipulating Chinese domestic politics and foreign policy through economic relations. Engagement is no panacea, and U.S. officials should not oversell it. The study of economic statecraft teaches that the effectiveness of any strategy can only be evaluated in light of its plausible alternatives (Baldwin 1985). Page 185 → Compared to the alternatives, an economic-engagement strategy toward China fares better—even though its potential for success is modest in absolute terms. Economic containment is the most plausible alternative. That strategy would require the cooperation and support of other advanced industrial states— cooperation and support that is unlikely to be forthcoming so long as China remains an attractive market and refrains from military action against Taiwan or its neighbors. Moreover, any U.S.–inspired economic-containment effort is likely to be viewed by China as provocative. It would create a selffulfilling prophecy by giving China incentives to pursue a more confrontational foreign policy. The obvious downside of economic engagement is that, in seeking to moderate Chinese foreign policy behavior, the strategy simultaneously strengthens China's economic and potentially military capabilities. As Hirschman and so many others have noted, this trade-off between capabilities and behavior is a necessary complication of the economic-engagement strategy. The smaller the gap in relative power between the sanctioning state and its target challenger— that is, the more proximate the challenge—the more reluctant the sanctioning state should be to “take the chance” and pursue policies that strengthen the relative capabilities of a potential adversary through trade. The United States, however, finds itself in an enviable position. The relative power gap between the United States and China remains quite large. China will close that gap only slowly, and only if the more optimistic projections

regarding its economic growth, technological development, and political and social stability prove valid. If China does muster the combination of capabilities, intentions, and behavior to mount a revisionist challenge, the United States can react accordingly and mobilize for containment. In the meantime it can afford to experiment with economic engagement both for pragmatic reasons and with the hope of any gambler that long shots occasionally pay off.

REFERENCES Abdelal, Rawi, and Jonathan Kirshner. 1999–2000. Strategy, Economic Relations, and the Definition of National Interests. Security Studies 9:119–56. Baldwin, David A. 1971. The Power of Positive Sanctions. World Politics 24:19–38. ———. 1985. Economic Statecraft. Princeton: Princeton University Press. Blanchard, Jean-Marc F., Edward D. Mansfield, and Norrin M. Ripsman, eds. 2000. Power and the Purse: Economic Statecraft, Interdependence, and National Security. London: Frank Cass. Page 186 → Copeland, Dale C. 1999–2000. Trade Expectations and the Outbreak of Peace: Détente, 1970–1974, and the End of the Cold War, 1985–1991. Security Studies 9:15–58. Davis, Patricia. 1999. The Art of Economic Persuasion: Positive Incentives and German Economic Diplomacy. Ann Arbor: University of Michigan Press. Drezner, Daniel. 1999–2000. The Trouble with Carrots: Transaction Costs, Conflict Expectations, and Economic Inducements. Security Studies 9:188–218. Elliott, Kimberly Ann. 1998. The Sanctions Glass: Half-Full or Completely Empty? International Security 23:50–65. Galtung, Johan. 1967. On the Effects of International Economic Sanctions, with Examples from the Case of Rhodesia. World Politics 19:378–416. Hirschman, Albert O. [1945] 1980. National Power and the Structure of Foreign Trade. Reprint, Berkeley: University of California Press. Hufbauer, Gart Clyde, and Jeffrey Schott, with Kimberly Ann Elliott. 1991. Economic Sanctions Reconsidered: History and Current Policy. 2d ed. Washington, DC: Institute for International Economics. Long, William J. 1996. Trade and Technological Incentives and Bilateral Cooperation. International Studies Quarterly 40:77–106. Mastanduno, Michael. 1998. Economics and Security in Scholarship and Statecraft. International Organization 52:825–54. Morgenthau, Hans J. 1978. Politics among Nations: The Struggle for Power and Peace. 5th ed. New York: Knopf. Newnham, Randall E. 2002. Deutsche Mark Diplomacy: Economic Linkage in German-Russian Relations. State College: Pennsylvania State University Press. Papayoanou, Paul. 1999. Power Ties: Economic Interdependence and War. Ann Arbor: University of Michigan Press.

Papayoanou, Paul, and Scott Kastner. 1999–2000. Sleeping with the (Potential) Enemy: Assessing the U.S. Policy of Engagement with China. Security Studies 9:157–87. Pape, Robert. 1997. Why Economic Sanctions Do Not Work. International Security 22:90–136. Skålnes, Lars. 2000. Politics, Markets, and Grand Strategy: Foreign Economic Policies as Strategic Instruments. Ann Arbor: University of Michigan Press. Solingen, Etel. 1998. Regional Orders at Century's Dawn: Global and Domestic Influences on Grand Strategy. Princeton: Princeton University Press.

Page 187 →

Part 2. Conceptualization and Measurement of Interdependence and Conflict Interdependence Page 188 →

Page 189 →

Empirical Support for the Liberal Peace John R. Oneal Over the past twenty years, scientific research on the causes of war has progressed rapidly by focusing on the interstate relations of pairs of states observed through time. The analysis of dyadic time series is an important advance on the previous strategies of analyzing international relations at either the global or the national level of analysis. Attention to the behavior of pairs of states directly addresses the questions of greatest concern to political scientists and policy-makers alike: which states are likely to fight one another, and which will remain at peace? Research at the dyadic level has resulted in an appreciation of the separate peace that has existed among democratic states for over a hundred years, but the benefits of democracy do not exhaust the theoretical insights of the classical liberals. Immanuel Kant ([1795] 1970) suggested that international peace could be established on a foundation of republican constitutions, “cosmopolitan law” embodied in economic interdependence, and international law and organizations. It has only recently become possible to evaluate scientifically the “philosophical proposal” Kant put forward two hundred years ago: the triangle of institutions he emphasized has now come into existence in substantial parts of the world, the relevant data have been collected, appropriate statistical techniques have been developed, and the computing capacity necessary for the analysis is available. As Bruce M. Russett and I have reported elsewhere (Oneal and Russett 1999c; Russett and Oneal 2001), there is strong support for the pacific benefits of democracy, interdependence, and international organizations over the years 1886–1992. In this chapter, I focus on the second element of the Kantian peace: the pacifying influence of economic interdependence. I present new empirical tests Page 190 → using data that have only recently become available. These analyses are conducted using a variety of statistical estimators that allow for temporal dependence in the time series. They provide unqualified support for liberal theory. I also offer evidence that trade concentration does not make dyadic conflict more likely, contrary to an important statistical challenge to the liberal thesis (Barbieri 1996). I suggest that Barbieri's analysis is misspecified. Finally, I offer some thoughts on future directions for research on the liberal peace.

Confirming the Liberal Peace with New Data and Various Statistical Estimators In this section, I report new tests of the pacific benefits of economic interdependence for the period 1886–1992. The opportunity for these analyses is presented by Kristian S. Gleditsch's (2002) recent compilation of new data regarding states' bilateral trade and their gross domestic products (GDPs), 1950–92. Gleditsch has supplemented the International Monetary Fund's (IMF's) (1997) Direction of Trade statistics and the GDP data in the Penn World Table (Summers et al. 1995) with information from additional sources. There are two particularly important advantages to using Gleditsch's data. First, it includes dyads from within the Soviet bloc during the Cold War—observations absent from the IMF data. Most of the COMECON (Council for Mutual Economic Cooperation) countries were never members of the IMF during the Cold War, so trade within this group was unreported. Second, by drawing on additional sources, Gleditsch has eliminated the debate regarding missing values of trade in the Direction of Trade (Barbieri 1996; Oneal and Russett 1999a). For the pre–World War II period, I use the economic data that Russett and I assembled. These have been discussed in detail elsewhere (Oneal and Russett 1999c; Russett and Oneal 2001). The primary sources of our early trade data are The Statesman's Yearbook (Epstein 1913), for the years before World War I, and the League of Nations (various years), for the interwar years. These were carefully compared to Katherine Barbieri's (1998) data. Maddison (1995) provides estimates of per capita GDPs for a large number of countries, which permitted us to estimate the GDPs for many other countries as well. With this information, I calculated the bilateral trade-toGDP ratio and the total trade-to-GDP ratio, because trade is expected to influence international relations only if it is economically (and hence politically) important. The total trade-to-GDP ratio for many countries, 1950–92, is

found in Summers et al. 1995. With the publication of Gleditsch's economic data for the post–World War II period and the assembly of the best available data for the pre–World War I and interwar periods, Page 191 → the stateof-the-art data set on interstate commercial relations, 1885–1992, seems at hand. I also use Zeev Maoz's (1999) data regarding militarized interstate disputes (MIDs). Past efforts to produce dyadic data from the Correlates of War (COW) project's state-level MIDs file (Jones, Bremer, and Singer 1996) are problematic to a degree. States on one side of a multilateral contest, for example, may never have threatened, displayed, or used force against particular states on the opposing side. Maoz has corrected this problem and others to produce more accurate data for dyadic analyses. In the tests that follow, I report the results of two sets of analyses: the first for the onset of any militarized dispute and the second for particularly serious conflicts—those involving fatalities among the combatants. I rely on the basic specification used in recent publications with Russett: the likelihood of the onset of a dispute is a function of the lower democracy score and the lower bilateral trade-to-GDP ratio in each dyad. The less constrained state will be the weak link in the chain of peace. In this view, the state with the lower bilateral trade-to-GDP ratio in each dyad determines the danger of interstate violence, ceteris paribus: the more dependent this state is on its commercial relations with the other member of the dyad, the more averse it will be to engaging in a dispute, and the more peaceful the dyad.1 This specification has evolved from that of Bremer (1992) and Maoz and Russett (1993). Bennett and Stam (2000b); Hegre (2000); Mousseau (2000); Gartzke, Li, and Boehmer (2001); King and Zeng (2001); McDonald (2001); Heagerty, Ward, and Gleditsch (2002); and Kinsella and Russett (2002) have used similar models. To gauge the character of regimes, I use the Polity III data (Jaggers and Gurr 1996). The balance of power is measured using the logarithm of the ratio of the stronger state's military capabilities to those of the weaker in each dyad (Singer and Small 1995). There is an indicator for alliances; Singer's (1995) data were updated using Rengger with Campbell (1995). Because of the importance of geographical proximity, two different terms are included: the log of the great circle distance in miles between the two states' capitals (or major ports for the largest countries) and a measure of contiguity.2 The effect of distance in constraining conflict, however, is less for the major powers, so a binary variable indicating whether a dyad includes a major power (as defined by the Correlates of War project) is included. To evaluate the effect of interdependence on the likelihood of a fatal militarized dispute, 1886–1992, I use logistic analyses of pooled time series for the politically relevant dyads: contiguous states and pairs that include at least one major power. Robust standard errors are reported, which allows for Page 192 → unspecified correlation within the dyadic data. To account explicitly for autocorrelation, I use the method of generalized estimating equations (GEE) and correct for an AR1 process. Zorn (2001) has drawn attention to the flexibility of this statistical technique. I also estimate key analyses using Beck, Katz, and Tucker's (1998) correction for duration dependence and the Window Subseries Empirical Variance (WSEV) estimator of Heagerty, Ward, and Gleditsch (2002). Because all the theoretical relationships are directional, one-tailed tests of statistical significance are reported throughout. Limiting the analysis to the politically relevant dyads is not uncontroversial. Focusing on this set of conflict-prone states is, in effect, selecting cases based on the value of the dependent variable. Accordingly, the sample is not representative of the population of all possible pairs of states. The counterargument, however, is simple: the risk of conflict in the “nonrelevant” population is so small—about .04 of 1 percent (Oneal and Russett 1999a)—that the nonrepresentativeness of the sample is less important than the enhanced precision gained in assessing the probability of conflict for those states for which the risk is important as a practical matter (Gleditsch and Hegre 1997; Oneal and Russett 1999a; Gleditsch and Ward 1999). Neither does the danger of bias seem as great as once thought (Lemke and Reed 2001), and the results do not depend upon a particular definition of “political relevance” (Gates and McLaughlin 1996).3 Analyzing the contiguous and major power dyads has other advantages. Most important, it substantially increases the proportion of the cases from the pre-1950 years. It is important to insure that the liberal peace is not limited to

the Cold War era (Farber and Gowa 1997) and its unusual bipolar character. By analyzing the politically relevant pairs, 30 percent of the cases are drawn from the pre–World War II years. If all possible pairs are used, this drops to 13 percent. With such a small number of cases from the pre–World War II period, it would be difficult to generalize the results beyond the Cold War period. It is important to analyze as many years as possible, as Edward D. Mansfield and Brian M. Pollins emphasize in their contribution to this volume. In any event, the analyses following were redone using all dyads for which data are available. The results do not differ substantially from those reported. Table 1 gives the results of estimating four models of the onset of a militarized dispute using GEE (AR1). Model 1 is the basic model Russett and I have used recently (Oneal and Russett 1999c; Russett and Oneal 2001). As seen in column 1, the lower bilateral trade-to-GDP ratio is very significantly related to a reduction in the likelihood of conflict (p < .001). As expected, the lower democracy score, too, reduces the risk of interstate violence. Consistent with realist tenets, a preponderance of power—not a balance—and an alliance are Page 193 → associated with more peaceful interstate relations. Proximate states fight more than those more distant, and contiguous states are prone to violence. Contiguous pairs and those that are nearly neighbors have more to fight about and are better able to take military action should they do so. Finally, major powers are prone to violence. These results are consistent with the great majority of previous studies that have adopted the dyadic perspective.4 Page 194 → The analysis in column 2 approximates the tests reported by Barbieri (1996), who found that extensive economic interdependence increased the likelihood of dyadic disputes, 1870–1938. Barbieri used the concentration of trade—the ratio of two states' bilateral trade to their total trade with all commercial partners—to gauge interdependence, rather than the trade-to-GDP ratio most researchers prefer (e.g., Polachek 1980; Domke 1988; Mansfield 1994; Hegre 2000). The decision to use trade concentration, not the economic importance of trade, was dictated by a lack of data regarding nations' gross domestic products prior to 1950. As Barbieri surmised, trade concentration and bilateral-trade dependence are correlated, though imperfectly. The correlation is.68 for the cases analyzed here. The results in column 2 do not suggest that states are more likely to fight when trade is concentrated. The estimated coefficient is positive but is far from statistical significance (p < .31) for the long period analyzed here. If the analysis is limited to the pre–World War II period, as in Barbieri (1996), trade concentration is significantly associated with an increased incidence of conflict (p < .003). Rather than indicating that trade concentration—much less economic interdependence—increases the risk of a dispute, it is more likely that the model is misspecified. Recall that trade concentration equals trade ij /total trade i. The measure of the economic importance of state i's trade with j is trade ij /GDP i ; and economic openness equals total trade i /GDP i. Trade concentration is equal, therefore, to the bilateral trade-to-GDP ratio divided by openness. If the dependence of state i on its trade with j is held constant and open-ness i is decreased, trade concentration ij will increase. There is substantial evidence that states are constrained from fighting when they are tightly linked to the international economy (Domke 1988; Oneal and Russett 1997, 1999b; Russett and Oneal 2001). Barbieri's (1996) finding that trade concentration is associated with higher levels of conflict, 1870–1938, may indicate only that the total trade-to-GDP ratio, as well as the bilateral ratio, constrains states from using force. The third model in table 1 allows the independent pacific benefits of economic openness to be assessed. The lower trade-to-GDP ratio and the same state's total trade-to-GDP ratio are both included. As seen in column 3, both are closely associated with greater dyadic peace. Indeed, the estimated coefficient for economic openness is more statistically significant (p < .001)— though less important substantively (as will be seen)—than the coefficient of the bilateral measure (p < .003). Apparently, national leaders recognize that militarized disputes have consequences for economic relations not only with a Page 195 → potential adversary but also with third parties. Even limited bilateral trade and investment do not mean that decision makers are unconstrained by economic forces from taking military action. Rather, conflict is most likely when states are unhampered by important external economic relations, whether bilateral or multilateral.

The link between openness and peace is more important than has been generally recognized. It provides a strong indication that trade plays a causal role in limiting conflict. Deteriorating political relations and the rising prospects of military violence may lead a state to reduce its dependence on a potential adversary, but it would be virtually impossible for a state to reduce significantly its economic ties with all its commercial partners simultaneously. In their introduction to this volume, Mansfield and Pollins express concern over the use of a trade-to-GDP measure to indicate interdependence. They note that it will be correlated with economic size. Small states do tend to be more open. Trade is the means by which they achieve economies of scale in production and acquire the diverse array of resources people desire to consume. Mansfield and Pollins also note that large states tend to be politically powerful and disproportionately likely to become involved in military conflict. There are several reasons to doubt that the evidence presented for the liberal peace can be subsumed under a realist account of this sort. First, past studies do control for economic size by including in the regression models the capability ratio and an indicator of whether the dyad includes a major power. States with large GDPs are, of course, usually populous, industrialized, and in possession of substantial military forces—the constituents of the COW military capabilities index. Second, the weight of the evidence is clear: an imbalance of power deters conflict, as Kenneth Organski and others (Organski and Kugler 1980; Bremer 1992; Lemke and Kugler 1996) have suggested. The analyses reported in table 1 corroborate this view. It is true, however, that those states identified as the “major powers” by historians have been more frequently involved in militarized disputes than other states, as the results in table 1 show. It is easy to confirm that trade-based measures of interdependence are not acting as additional proxies for relative power or great power status. In the final column of table 1, I have added the log of the ratio of the larger to the smaller GDP in each dyad to the model in column 3. The coefficient of the GDP ratio is positive and not statistically significant (p < .12). The coefficients of the other variables are virtually unchanged. Thus far, I have focused attention on the statistical significance of the estimated coefficients, but of course, it is the substantive impact of the variables that is more important. A small effect, even though very significant statistically, Page 196 → is inconsequential, but an effect that is large on average, even if uncertain to a degree, is important. In table 2, I report the reductions in the probability of the onset of a militarized dispute, expressed in percentages, from increasing the measures of interdependence by one standard deviation. These effects are calculated relative to a “typical” contiguous dyad, one that has the median level of the lower bilateral trade-toGDP ratio, has the mean values for the continuous variables in the relevant model, is not allied, and does not include a major power. The first and fourth rows in table 2 indicate the pacific benefits of interdependence associated with models 1 and 3 in table 1. Clearly, they are substantial. An increase in the bilateral trade-to-GDP ratio reduces the risk of conflict by 34 percent using model 1. This is similar to the results we have reported elsewhere (Oneal and Russett 1999c; Russett and Oneal 2001). Bilateral interdependence reduces the probability of a dispute by 29 percent in model 3, while openness has an independent benefit of 24 percent. These figures can be compared to the consequence of making the dyad allied, which is also reported in the table. Increasing both the bilateral and total trade-to-GDP ratios in model 3 reduces the probability of a dispute by 46 percent. We have explained our preference for the GEE estimator elsewhere (Oneal and Russett 1999a, forthcoming), but Beck, Katz, and Tucker (1998; Beck forthcoming) have proposed an alternative method for correcting for dependence in the dyadic time series. They suggest controlling for the years of peace since the last dispute. We have two concerns about this proposal. First, it does not give primacy to the theoretically derived model. Rather, it introduces the years of Page 197 → peace into the specification of the regression equation as coequals of the theoretical terms. Second, to the extent that the theoretical variables are correlated with the years since the last dispute, Beck's control apportions to itself some of the variance in the dependent variable that would otherwise have been explained by the theoretical terms. A similar problem arises in the analysis of continuous variables when a lagged dependent variable is included on the right-hand side of the regression equation (Achen 2000). Beck's correction is, however, a valuable method for addressing duration dependence. The support for the pacific

benefit of the liberal peace using this technique is important (Oneal and Russett 1999a, 1999b, 1999c; Bennett and Stam 2000b; Hegre 2000; Mousseau 2000; Gartzke, Li, and Boehmer 2001; King and Zeng 2001; Heagerty, Ward, and Gleditsch 2002; and Beck forthcoming).

Consequently, I reestimated models 1 and 3 using Beck, Katz, and Tucker's correction. The coefficient of the lower trade-to-GDP ratio is again significant at the .001 level. The substantive implications are reported in the second and fifth rows of table 2. The benefits of economically important bilateral trade, as indicated in model 1, are smaller (–24 percent in model 1, –20 percent in model 3) than the estimates produced by GEE, but they are still substantively important, as is the reduction in risk afforded by economic openness (–17 percent, model 3). Recently, Heagerty, Ward, and Gleditsch (2002) have proposed a correction for spatial and temporal dependence in pooled time series that is consistent with the view that it is the standard error, not the coefficient, that is in doubt when observations are not independent. After all, when data are autocorrelated, we should be primarily concerned about the reliability of an effect, as indicated by the size of the standard error, not the magnitude of the coefficient itself. Heagerty, Ward, and Gleditsch note that the experiences of dyads at a single point in time are not independent, in part due to the contagion effect of interstate conflict. Failing to take this into account confounds standard estimators of variance. As a remedy, they propose the Window Subseries Empirical Variance estimator, which averages empirical variance estimations drawn from different subsamples of the data through time. This produces consistent estimates of the variances of the regression coefficients. The essence of the method is straightforward. Instead of using dyads to define the clusters for producing robust standard errors, the WSEV estimator uses slices of time. When these slices include more than one year, this technique also adjusts for autocorrelation in the time series.5 The period from 1886 to 1992 can be broken into five periods of historical significance: 1886–1903, 1904–14, 1920–39, 1951–73, and 1974–92.6 The four break-points in time correspond to the formation of the Entente Cordiale Page 198 → between France and Great Britain in 1904, the outbreak of World War I, the start of the interwar years, the onset of the Cold War, and the beginning of détente between the United States and the USSR. Thus, I assume that the geographical pattern of disputes is similar within each period. By estimating robust standard errors using multiyear clusters, adjustment is made for temporal as well as spatial dependence. The results of reestimating models 1 and 3 using Heagerty, Ward, and Gleditsch's (2002) estimator are presented in the third and sixth rows of table 2. Interdependence reduces the probability of a militarized dispute in model 1 by 38 percent, and the estimated coefficient is significant at the .001 level. As seen in the sixth row, openness is also important, reducing the probability of the onset of a dispute by 25 percent, even when the effect of bilateral trade is also estimated (–31 percent). This much seems clear from table 2: the pacific benefit of economic interdependence is statistically significant and substantively important, whichever estimator is used.

Analyses of Fatal Disputes Underscore the Benefits of Interdependence The results reported in tables 1 and 2 understate the case for the liberal peace. Focusing attention on particularly serious disputes—those involving fatalities among the combatants—shows that the benefit of interdependence is even more important than indicated there. Table 3 reports the results of estimating the same models as before, again using GEE with a correction for an AR1 process, but in this table only the onset of fatal disputes—conflicts that resulted in the death of at least one combatant—are considered. Limiting the analysis to disputes with at least one military death reduces the geographical bias in the reporting of conflict at the low end of the spectrum and insures that we are addressing the interstate violence of greatest concern. The results reported in table 3 are very similar to those discussed earlier. The lower trade-to-GDP ratio is highly significant in both model 1 (p < .001) and model 3 (p < .003), and the total trade-to-GDP ratio is very significant (p < .001) in model 3. Model 2 shows that trade concentration is unrelated to the incidence of fatal militarized disputes, and model 4 dispels concern that this evidence for the liberal peace is an artifact of the size of states.

Table 4 indicates the dramatic effect that economically important trade has on the likelihood of a fatal dispute. Using the estimated coefficient in model 1, table 3 indicates that interdependent states are 68 percent less likely than the Page 199 → typical pair of states to be involved in a conflict with military fatalities. With Beck, Katz, and Tucker's (1998) correction, the reduction in risk is 58 percent. Interdependent states are 71 percent less likely to experience a fatal dispute if the WSEV estimator is used. The lower bilateral trade-to-GDP ratio is very significant in all three estimations, and its pacific benefit is greater than the reduction in the risk of conflict achieved by making the pair of states allies. Page 200 → Economic openness, too, even controlling for bilateral trade, is an important force for peace. It is always statistically significant (at least the .005 level) and substantively important, reducing the likelihood of a fatal dispute by 26 percent (with Beck, Katz, and Tucker's correction) to 32 percent (GEE and WSEV).

Conclusion and Suggestions for Future Research As weapons of mass destruction become accessible to more and more states, it is essential to find a safer, more reliable foundation than military might on which to build a peaceful international system. Nor does America's predominance insure world peace (Oneal and Russett 1999b; Russett and Oneal 2001). Democracy, interdependence, and international organizations—the three legs of the Kantian tripod—offer this promise. If we are to have confidence in this prognosis and encourage policies promoting liberal institutions and values, we must, however, be as sure as science will allow that the liberal influences actually cause a reduction in interstate conflict. We must do what we can to guard against the possibility that peace encourages trade (Anderton and Carter 2001) but not vice versa. Indeed it is very likely that there are reciprocal relations between interdependence and peace because, liberals argue, it is the fear that conflict will disrupt beneficial commerce that leads states to refrain from using military force. Careful consideration of the problem of endogeneity, in order to confirm the causal effect of economic interdependence, should be, therefore, a priority in future research. Page 201 → There have been efforts to address this issue in the past. Polachek (1980, 1997) used two-stage and three-stage least-squares regression analysis to take into account current political and military conflict; and Mansfield (1994) estimated a system of two simultaneous equations, one predicting the level of global trade and the other the number of wars in the international system. All these studies found that trade reduces the incidence of conflict, independent of past levels of violence. More recently, Kim (1998) used simultaneous equations to disentangle the reciprocal relations between trade and interstate conflict, 1950–85. Her work is particularly important because she uses a dyadic analysis, considers militarized disputes and war (not political tensions), and has a large number of cases. Furthermore, she uses sophisticated models to account for both bilateral trade and the likelihood of a dyadic dispute. Kim concludes that the effect of trade on conflict is stronger than the effect of conflict on trade. Beck, Katz, and Tucker's (1998) correction for duration dependence is an alternative method of dealing with the effect of past conflict on the current probability of a dispute. Most previous efforts to estimate simultaneously a system of equations for trade and conflict have included only contemporaneous terms, implying that conflict in a particular year affects trade only in the same year and vice versa. It seems more likely that the occurrence of a militarized dispute in one year will affect the decisions of investors and traders for some number of years into the future. This is Beck's (Beck, Katz, and Tucker 1998) essential insight. But it is possible that important commercial relations, too, have a long-term effect: the likelihood of conflict in the future may be lower for a dyad with a long history of close economic ties, even if it is currently involved in a dispute, than if the two states had never been interdependent. Simply controlling for the years of peace does not allow for this possibility. This assumes that the time elapsed since the last dispute is independent of the influences of the theoretical variables included in the regression analysis. It is much more likely that the years of peace a dyad has enjoyed are a function of the past character of the two states' political systems, their trade, and so on. Distributed-lags models, vector autoregression (VAR), and Granger-causality testing are related methods that provide additional means of addressing the endogeneity of conflict and interdependence (Gasiorowski and Polachek 1982; Reuveny and Kang 1996; Reuveny 2001). Using Granger's (1969) logic, economically important

trade, democracy, and so on, can plausibly be considered causes of peace if their past values can be used to predict the current likelihood of a dispute more accurately than using dyads' histories of disputes alone. This approach has several advantages. First, it does not assume that reciprocal effects are simultaneous but allows for disputes to affect the current likelihood Page 202 → of conflict over a period of years. Second, it controls for dependence in the time series in a manner that is substantially richer and more complete than a count of the time elapsed since the last dispute. In estimating the current likelihood of conflict, it would distinguish, for example, a dyad that enjoyed nine years of peace and then suffered a military dispute from a pair of states that was involved in a dispute every year for ten years. Third, it allows past values of the explanatory variables to influence the current likelihood of conflict. Thus, long-term benefits of interdependence, and so on, can be detected. Fourth, it provides protection against accepting a spurious correlation as evidence of a causal relation, because the indicators of past militarized disputes act as proxies for explanatory variables omitted from the regression equation (Burkhart and Lewis-Beck 1994). Finally, it is simple to implement. Application of distributed-lags models (Oneal and Russett 2001) reinforces the findings presented here. Military conflict does adversely affect states' commercial relations, as the liberals expected, but economically important trade is also a powerful force for peace, even when the history of dyadic conflict is taken into account. The pacific benefit of trade is particularly great for fatal disputes, as the results presented earlier also suggest. Certainly it is important to clarify the limits of the liberal peace and to determine if it is contingent upon institutional means for peacefully resolving conflicts of interest (Hess this volume; Mansfield this volume), the nature of states' political regimes (Gelpi and Grieco this volume), or their level of economic development (Hegre 2000). Recent work indicates, however, that the benefit of interdependence is not contingent on either regime type or development (Mousseau, Hegre, and Oneal 2003). In the meantime, there are now even better reasons to believe that the classical liberals were right: interdependence, as well as democracy, reliably and dramatically increases the prospects for peace.

NOTES I would like to thank Neal Beck, Margit Bussmann, Dan Reiter, Bruce Russett, Mike Ward, and Christopher J. W. Zorn for helpful comments. I am also very grateful to Edward D. Mansfield, Brian M. Pollins, Pat McDonald, and others who organized the conference at the Mershon Center of the Ohio State University that resulted in this book. I benefited greatly from the presentations and comments of the other participants. 1. The higher trade-to-GDP ratio does not influence the probability that force will be used (Oneal and Russett 1997, 1999a, 1999c; Russett and Oneal 2001). 2. Because of widespread colonial empires for much of the period we analyze, these two measures are not highly correlated (r = –0.67), especially prior to World War I. 3. Bennett and Stam's (2000a) tests of Bueno de Mesquita and Lalman's (1992)Page 203 →international interaction game also suggest that focusing on the politically relevant dyads is sensible. 4. The log-files of the regression analyses are posted at . 5. I am grateful to Mike Ward for drawing my attention to this point. 6. Independent variables measured in year t explain disputes in year t + 1, so democracy, interdependence, and so on, in 1885 explain conflict in 1886. There are no data for the years of World Wars I and II or for the immediate post–World War II period (1946–49).

REFERENCES Achen, Christopher H. 2000. Why Lagged Dependent Variables Can Suppress the Explanatory Power of Other Independent Variables. Paper presented at the annual meeting of the Political Methodology Section of the American Political Science Association, July 20–22, Los Angeles. Anderton, Charles H., and John R. Carter. 2001. The Impact of War on Trade: An Interrupted Times-Series Study. Journal of Peace Research 38:445–57.

Barbieri, Katherine. 1996. Economic Interdependence: A Path to Peace or a Source of Interstate Conflict? Journal of Peace Research 33:29–49. ______. 1998. Dyadic Trade Data Set. Available at . December. Beck, Nathaniel. Forthcoming. Modeling Dynamics in the Study of Conflict. In Globalization and Conflict, ed. Katherine Barbieri, Nils Petter Gleditsch, and Gerald Schneider. London: Sage Publications. Beck, Nathaniel, Jonathan N. Katz, and Richard Tucker. 1998. Taking Time Seriously: Time-Series-Cross-Section Analysis with a Binary Dependent Variable. American Journal of Political Science 42:1260–88. Bennett, D. Scott, and Allan C. Stam. 2000a. A Cross-Validation of Bueno de Mesquita and Lalman's International Interaction Game. British Journal of Political Science 30:541–60. ______. 2000b. Research Design and Estimator Choices in the Analysis of Interstate Dyads: When Decisions Matter. Journal of Conflict Resolution 44:653–85. Bremer, Stuart A. 1992. Dangerous Dyads: Conditions Affecting the Likelihood of Interstate War, 1816–1965. Journal of Conflict Resolution 36:309–41. Bueno de Mesquita, Bruce, and David Lalman. 1992. War and Reason. New Haven: Yale University Press. Burkhart, Ross E., and Michael S. Lewis-Beck. 1994. Comparative Democracy: The Economic Development Thesis. American Political Science Review 88:903–10. Domke, William K. 1988. War and the Changing Global System. New Haven: Yale University Press. Epstein, Martin, ed. 1913. The Statesman's Yearbook, 1913. London: Macmillan. Page 204 → Farber, Henry, and Joanne Gowa. 1997. Common Interests or Common Polities? Journal of Politics 57:393–417. Gartzke, Eric, Quan Li, and Charles Boehmer. 2001. Investing in the Peace: Economic Interdependence and International Conflict. International Organization 55:391– 438. Gasiorowski, Mark, and Solomon W. Polachek. 1982. Conflict and Interdependence: East-West Trade and Linkages in the Era of Detente. Journal of Conflict Resolution 26:709–29. Gates, Scott, and Sarah McLaughlin. 1996. Rare Events, Relevant Dyads, and the Democratic Peace. Paper presented at the 37th annual meeting of the International Studies Association, April 16–20, San Diego. Gleditsch, Kristian S. 2002. Expanded Trade and GDP Data Version 1. Journal of Conflict Resolution 46:712–24. Gleditsch, Kristian S., and Michael D. Ward. 1999. A Revised List of Independent States Since the Congress of Vienna. International Interactions 25:393–413. Gleditsch, Nils Petter, and Havard Hegre. 1997. Peace and Democracy: Three Levels of Analysis. Journal of Conflict Resolution 41:283–310. Granger, Clive W. J. 1969. Investigating Causal Relations by Econometric Models and Cross-Spectral Methods. Econometrica 37:424–38. Heagerty, Patrick, Michael D. Ward, and Kristian S. Gleditsch. 2002. Windows of Opportunity: Window Subseries Empirical Variance Estimator in International Relations. Political Analysis 10:304–17. Hegre, Håvard. 2000. Development and the Liberal Peace: What Does It Take to Be a Trading State? Journal of

Peace Research 37:5–30. International Monetary Fund. 1997. Direction of Trade (ICPSR 7628). Washington, DC: International Monetary Fund. Distributed by Inter-university Consortium for Political and Social Research, Ann Arbor, MI. Jaggers, Keith, and Ted Robert Gurr. 1996. Polity III Data. May. Available at . Jones, Daniel M., Stuart A. Bremer, and J. David Singer. 1996. Militarized Interstate Disputes, 1816–1992: Rationale, Coding Rules, and Empirical Patterns. Conflict Management and Peace Science 15:163–213. Kant, Immanuel. [1795] 1970. Perpetual Peace: A Philosophical Stretch. In Kant's Political Writings, ed. Hans Reiss. Cambridge: Cambridge University Press. Kim, Soo Yeon. 1998. Ties That Bind: The Role of Trade in International Conflict Processes, 1950–1992. Ph.D. diss., Yale University. King, Gary, and Langche Zeng. 2001. Explaining Rare Events in International Relations. International Organization 55:693–715. Kinsella, David, and Bruce Russett. 2002. Conflict Emergence and Escalation in Interactive International Dyads. Journal of Politics 64:1045–68. League of Nations. Various years. International Trade Statistics. Geneva: League of Nations. Page 205 → Lemke, Douglas, and Jacek Kugler, eds. 1996. Parity and War. Ann Arbor: University of Michigan Press. Lemke, Douglas, and William Reed. 2001. The Relevance of Politically Relevant Dyads. Journal of Conflict Resolution 45:126–44. Maddison, Angus. 1995. Monitoring the World Economy: 1820–1992. Paris: Organization for Economic Cooperation and Development. Mansfield, Edward D. 1994. Power, Trade, and War. Princeton: Princeton University Press. Maoz, Zeev. 1999. Dyadic Militarized Interstate Disputes Dataset—Version 1.1 (DYMID1.1). Available at . August. Maoz, Zeev, and Bruce M. Russett. 1993. Normative and Structural Causes of Democratic Peace, 1946–1986. American Political Science Review 87:624–38. McDonald, Patrick J. 2001. Private Property and the War Machine: Rethinking Liberal IR Theory. Paper presented at the 97th annual meeting of the American Political Science Association, August 30–September 2, San Francisco. Mousseau, Michael. 2000. Market Prosperity, Democratic Consolidation, and Democratic Peace. Journal of Conflict Resolution 42:472–507. Mousseau, Michael, Håvard Hegre, and John R. Oneal. 2003. How the Wealth of Nations Conditions the Liberal Peace. European Journal of International Relations 9. Oneal, John R., and Bruce M. Russett. 1997. The Classical Liberals Were Right: Democracy, Interdependence, and Conflict, 1950–85. International Studies Quarterly 41:267–94.

______. 1999a. Assessing the Liberal Peace with Alternative Specifications: Trade Still Reduces Conflict. Journal of Peace Research 36:423–42. ______. 1999b. Is the Liberal Peace Just an Artifact of Cold War Interests? Assessing Recent Critiques. International Interactions 25:213–41. ______. 1999c. The Kantian Peace: The Pacific Benefits of Democracy, Interdependence, and International Organization, 1885–1992. World Politics 52:1–37. ______. Forthcoming. Modeling Conflict while Studying Dynamics: A Reply to Nathaniel Beck. In Globalization and Conflict, ed. Katherine Barbieri, Nils Petter Gleditsch, and Gerald Schneider. London: Sage Publications. Oneal, John R., Bruce Russett, and Michael Berbaum. 2003. Causes of Peace: Democracy, Interdependence, and International Organizations, 1885–1992. International Studies Quarterly 47:72–94. Organski, A. F. K., and Jacek Kugler. 1980. The War Ledger. Chicago: University of Chicago Press. Polachek, Solomon W. 1980. Conflict and Trade. Journal of Conflict Resolution 24:57–78. ______. 1997. Why Do Democracies Cooperate More and Fight Less: The Relationship between Trade and International Cooperation. Review of International Economics 5:295–309. Rengger, N. J., with John Campbell. 1995. Treaties and Alliances of the World. 6th ed. New York: Stockton. Page 206 → Reuveny, Rafael. 2001. Bilateral Import, Export, and Conflict/Cooperation Simultaneity. International Studies Quarerly 45:131–58. Reuveny, Rafael, and Hejoon Kang. 1996. International Trade, Political Conflict/Cooperation and Granger Causality. American Journal of Political Science, 40:943–70. Russett, Bruce M., and John R. Oneal. 2001. Triangulating Peace: Democracy, Interdependence, and International Organizations. New York: W. W. Norton. Russett, Bruce M., John R. Oneal, and David R. Davis. 1998. The Third Leg of the Kantian Tripod for Peace: International Organizations and Militarized Disputes, 1950–1985. International Organization 52:441–67. Singer, J. David. 1995. Alliances, 1816–1984. Ann Arbor: University of Michigan, Correlates of War Project. Singer, J. David, and Melvin Small. 1995. National Military Capabilities Data. Ann Arbor: University of Michigan, Correlates of War Project. Summers, Robert, Alan Heston, Daniel Nuxoll, and Bettina Aten. 1995. The Penn World Table (Mark 5.6a). Cambridge, MA: National Bureau of Economic Research. Zorn, Christopher J. W. 2001. Generalized Estimating Equation Models for Correlated Data: A Review with Applications. American Journal of Political Science 45:470–90.

Page 207 →

Models and Measures in Trade-Conflict Research Katherine Barbieri Scholars and policymakers have long grappled with the question of whether trade promotes peace. The answer is not yet clear. While political commentators increasingly become engaged in debates over the consequences of trade and the implications of globalization, academics apply social-science methods to advance their understanding of the relationship between trade and conflict. Scholars have focused their energies on developing better measures of central concepts and employing more appropriate methodological techniques to answer the basic question of whether trade promotes peace. Even so, a scholar's theoretical perspective affects the way he or she conceptualizes trade, interdependence, and their consequences. This, in turn, affects a scholar's explanatory models, operational measures, and focus of inquiry. In this chapter, I explore the ways in which varying conceptions of the trade-conflict relationship may impact research findings and conclusions. In doing so, I argue that we in the community of trade-conflict researchers are best served by pursuing multiple means to capture varying dimensions of what we recognize to be complex relationships. If anything, differences in empirical findings in the literature might be telling us more about systematic variations in the trade-conflict relationship than we suspect. There may be some instances where the appropriate measurement choice is obvious. More generally, a multimethod strategy is the desirable solution. This chapter will be divided into two sections. First, I will discuss the ways in which different theoretical orientations impact varying conceptualizations of key variables in the trade-conflict relationship. Second, I will present examples of what I discuss in the first section by showing how alternative conceptualizations of dependence and interdependence might impact measurement Page 208 → decisions. This, combined with practical constraints, may pose hurdles to scholars as they work toward solving the trade-conflict puzzle. I argue that there should not be a monolithic approach to the study of trade and conflict but that we should consider which conceptualization best fits the questions being studied. In so doing, we may be able to tell a more nuanced tale about the conditions under which trade does and does not promote peace.

The Impact of Theoretical Orientation on Conceptualization The first problem that often confronts the trade-conflict research community arises from the ideological tone that can accompany any evaluation of commerce and interdependence. Debates about trade's perceived positive and negative consequences can be found throughout history (Selfridge 1918; Viner 1937). Surprisingly, little has changed over the ages in the issues of dispute or in the fervor with which people cling to their beliefs about commerce's impact on society and relations between societies. People tend to emphasize the good or the bad, the benefits or the costs of commerce, and the opportunities or potential vulnerabilities that come with dependence on foreigners and interactions with the “other.” We should not be surprised to see diversity among scholars interested in the trade-conflict relationship. Regardless of their degree of commitment to social-scientific principles, scholars always bring their personal biases to the table. This, in itself, should not be problematic. But it does mean that we must consider how theoretical leanings might affect the questions we ask and the answers we secure. The current line within the trade-conflict divide appears to be drawn between those who accept the liberal portrayal of trade as having a universally pacifying influence in international relations and those who question this view. The latter group lacks a unified voice, and its members are best described as “critics” or “skeptics” of liberalism. These scholars may be realists, Marxists, or some “neo” variant of these theoretical traditions. The critics vary in the extent of their skepticism about trade. Some believe that commerce causes conflict, while others simply believe that trade has a neutral effect on conflict. There seem to be few agnostics involved in discussions about trade and interdependence. Understanding a scholar's theoretical orientation is important, because it may shape subsequent stages of his or her research. This is true for even the most committed social scientist. Scholars with alternative conceptions of

interdependence may choose to focus their attention in different areas, employing different explanatory models, different samples, and different measures.1 This Page 209 → can lead to reinforcing tendencies, whereby scholars marshal evidence that supports their preconceived notions about trading relationships. For example, a scholar who immediately envisions the European Union when thinking about peaceful trading relationships may focus attention on that area and reach very different conclusions than someone who immediately conjures up images of colonialism and imperialism when thinking about trading relationships. In each case, scholars would offer alternative explanatory models, samples for analysis, and conclusions. Theoretical biases may seem less problematic among social scientists conducting large-N analyses than in other areas, but they can still be present in subtle ways. For example, the tendency to focus on interstate relations will limit our ability to capture the diversity of economic relationships that have existed over time and may exist in the future, such as relationships between colonial states and their colonizers. Most commonly, our theoretical leanings influence our analysis through our construction of explanatory models and operational measures. The link of trade to peace or trade to conflict requires some conduit. What is it about trade that leads to peace? The causal mechanisms that get us from point A to point B will vary, depending upon our initial assumptions about trading relationships. Moreover, the characteristics of trading relationships that we deem most important for our stories will also vary, depending upon theoretical orientation. This will lead to different operational measures of central constructs. In most instances, the theories germane to our research enterprise lack the necessary degree of specificity to identify definitively the most important theoretical constructs and causal mechanisms. This leaves many things open for interpretation. That should not be problematic, but it may hamper our ability to move forward as a research community and to achieve cumulativeness in our research endeavors. On the bright side, theoretical diversity might enable us to fill in many missing pieces in the trade-conflict puzzle. It is possible to construct and test competing models that capture alternative portrayals of the trade-conflict relationship. Yet we must be cognizant of the differences that exist in our research and the ways in which they might impact our findings. Differences in empirical findings in the literature may be telling us something about real variations in the tradeconflict relationship across cases. Within trade-conflict research we have a number of competing theories about the impact of commerce on conflict between states. Scholars tend to describe two types of relationships as emerging with the expansion of trade— interdependent versus dependent relations.2 Liberals argue that trade leads to interdependence between states, which in turn promotes peace. The liberal Page 210 → argument is based on both economic and sociological factors, with economic arguments enjoying greater popularity among contemporary scholars. According to the economic argument, trade generates wealth and other economic benefits; leaders are deterred from engaging in conflict with important trading partners for fear of losing the economic benefits associated with trade (Polachek 1980, 1992). The sociological strand of liberalism focuses on the bonds that emerge between actors that engage in extensive interactions, such as trade. Integration theorists argue that the expansion of interstate linkages in one area stimulates further cooperation in other areas (Deutsch et al. 1957; Haas 1958, 1964; Mitrany 1964). The assumption here is that increased contact is good for peace—contact reduces misperceptions, increases understanding, leads to a convergence of cultures, fosters formal and informal institutions to facilitate trade, and has spillover effects into other areas that lead to greater cooperation. Critics of liberalism offer alternative scenarios to the trade-promotes-peace hypothesis. Among them is the view that trade need not produce interdependent relationships but may instead produce relations of dependence. Such relationships may be highly conflictual. Moreover, the interdependence that emerges through extensive interactions, whether through trade or other ties, may simply increase the opportunities for conflict to emerge (Waltz 1979). Many scholars agree that highly unequal dependence produces a dynamic very different from the dependence existing in relations of mutual need. Liberals tend to neglect this analytical distinction when describing the benefits of trade, but neo-Marxist theorists have highlighted the contrasts among different types of economic relationships, drawing upon Marxist-Leninist thought (Marx [1887] 1906; Lenin [1939] 1990). Dependency theorists, in particular, explain how asymmetrical dependence affords the more independent state a position of power over the dependent state, whereby the latter becomes vulnerable to political and economic

manipulation (see Dos Santos 1970). In addition, for radical economists, the costs and benefits of trade are assumed to be highly unequal in relations of dependence. A reading of critical theories of trade leads one to infer that trade's impact on interstate relations is contingent upon the distribution of costs and benefits in that relationship, which is closely tied to the symmetrical nature of dependence in a given relationship. Thus, one might hypothesize that tensions are more likely to arise in asymmetrical relations due to the exercise of power derived through such relations, the perception of negative consequences of dependence, or concerns about relative gains. Liberals believe that the benefits of trade far outweigh any costly aspects of it. A clear link is therefore established between expanded trade and peace Page 211 → within the liberal tradition. This link of trade to peace is tenuous for those who maintain that trade might entail net costs, for those who view states' concerns about absolute gains as subordinate to concerns about relative gains (see Baldwin 1993), and for those who view increased contact as harmful to interstate relations (see Forbes 1997). It is possible to reconcile the variances in the propositions advanced by different theoretical traditions by considering the commonalties among the arguments. The basis of trade's pacifying effect is presumed to arise from the benefits derived from economic linkages. When such ties are believed to contribute to poverty or domestic disequilibria, the pacifying influence of trade may be neutralized. In fact, it may be reversed, whereby increased trade leads to increased conflict. If one state perceives poverty to arise from economic relations, trading relations might become hostile, whereas when states enjoy increased prosperity from trade, trading relations might become harmonious. In sum, beneficial trade may deter conflict, while detrimental effects might be associated with conflict. The question, of course, arises over which types of trading relationships are most likely to produce hostile and which ones peaceful relationships. In my own research, I have tried to distinguish between the concepts of interdependent and dependent relationships to characterize the types of trading relationships described by liberals and their critics, respectively. I have argued that interdependence emerges in relationships that have both salient and symmetrical ties (Barbieri 1995, 1996, 2002). Trade ties must be extensive, rather than weak, to distinguish interdependent from relatively independent relationships. Furthermore, they must be balanced, as opposed to asymmetrical, to denote interdependent as opposed to dependent relations. The analytical distinction between dependence and interdependence is important to shed light on the different ways in which trading relationships are characterized in the literature. Scholars on all sides of the trade-conflict debate frequently use the terms trade, dependence, and interdependence interchangeably, while describing very different dynamics in the trade-conflict relationship. These conceptual distinctions are important to make before developing measures of central constructs.

Measuring Central Concepts Given the differences in our theoretical conceptions of the trade-conflict relationship, researchers should not be surprised that we often construct different measures of the key variables central to our analysis. However, measurement choice may have unintended consequences for our analysis that have been inadequately explored. At a minimum, employing alternative measures of key Page 212 → constructs makes it difficult to compare empirical findings across studies. More important, measurement choice could influence the cases we analyze and the conclusions we reach. This suggests that a multimethod approach to analyzing the trade-conflict relationship may provide valuable information to scholars. Barbieri and Schneider (1999) provide a summary of quantitative studies that assess the impact of trade on conflict. Their review indicates the diversity of measures, samples, analytical techniques, and findings among trade-conflict studies. Scholars engaged in trade-conflict research might benefit from taking a closer look at the efforts of integration and dependency theorists who decades ago dealt with the same types of issues we confront today. Within each of these research programs, scholars employed a number of alternative operational measures of the phenomenon they wished to capture, whether integration or dependence. For example, Hirsch (1986, 117) identifies more than sixteen operationalizations of trade dependence used in two decades of dependency literature. Since the time of Hirsch's review, researchers have introduced new measures of dependence and interdependence and continue to do so. While not ideal, having numerous measures of interdependence may be far less problematic

than having one measure that does not fit our conceptualization of the phenomenon. But having so many measures makes it difficult to compare empirical findings across studies and to accumulate knowledge. Unfortunately, we have not thought enough about what different measures represent and what the findings produced with alternative measures might be telling us. Hughes's (1971) assessment of commonly used measures of integration seems timely for those contemplating which measures are most suitable for assessing the trade-conflict relationship. Hughes analyzes three measures of integration: (1) the Savage and Deutsch (1960) null model; (2) the export-percentage model (exports/total exports); and (3) the gross national product (GNP) model (exports/GNP). Hughes argues that these are all reasonable measures but adds that they may produce different findings. How then does one decide which measure is best to employ? Hughes believes that the way researchers opt to introduce the size control for trade relative to either total transactions or total income can be traced to their notion of whether integration is a constant-sum or variable-sum process. For the first two models listed earlier, an increase in trade with one state must lead to a concurrent decline in integration with other states, since total trade represents, figuratively speaking, a pie being divided among partners. When GNP is used as a control for dyadic trade flows, integration with all trade partners can grow over time if GNP is also growing. If researchers instead choose to use total trade in their denominator Page 213 → to control for the importance of dyadic trade flows, the growth in trade with one partner would signify a decline in dependence on all other states. Thus, for the total-trade-based model, increased dependence on one trade partner signifies a decreased dependence on all other partners. Clearly, we could think of interdependence as the relative importance of trade for the economy or the importance of a given trade partner relative to others. It is inappropriate to say that one is “wrong” and the other “right,” since each strategy captures important pieces of information. In my own research, I have developed measures that capture two types of dependence, which I term “partner-dependence” and “economy-dependence” measures. The first is a total-trade-based measure, and the second is based on gross domestic product (GDP) (Barbieri 1995, 2002). These measures are highly correlated, making it difficult to estimate their independent effects simultaneously in a single model. Yet these measures are conceptually distinct and at times empirically distinct. For example, in the pre–World War II era, the major powers, which have high estimated GDP values, were also the major traders. Therefore, total trade and GDP would be positively correlated. However, in the post–World War II period, there is more variation in the association between GDP and total trade. In general, there is an inverse relationship between a state's dependence on foreign trade, relative to GDP, and the state's level of development (Domke 1988). Thus, the states that rank high on measures of dependence might vary, depending upon the operational measure one employs. It is possible for states to have high partner dependence and low economy dependence. Such a situation would arise if a state's economy was not heavily dependent upon foreign trade, but the same state elected to conduct most of its trade with a few trade partners. If the state could not acquire certain products domestically and relied on one partner for those products, high partner dependence might reflect a form of dependence not captured with the economy-dependence measure, where the dyadic trade flow would appear minor relative to the national product. If, on the other hand, a state was very dependent on foreign trade, relative to national production, but chose to rely on a large number of trading partners, it might have a low partner-dependence score for a relationship but a high economy-dependence score. In my own research, I have found that the partner-dependence score for most states tends to be higher than the corresponding economy-dependence score for the state. The economy-dependence measure tends to be higher in cases where states have low GDP values. While most of the dyadic level measures of interdependence are sound, one Page 214 → increasingly employed measure of dependence warrants reconsideration. Oneal and Russett (1999) advise relying exclusively on one state's dependence score (i.e., the score of the state with the lower dependence level) as a means to measure dyadic interdependence. The rationale for this approach rests on Dixon's (1993) argument that it is the “weakest link” in a relationship that is less constrained to refrain from force and, therefore, it is this weakest link that determines the degree of conflict in the dyadic relationship. In the case of trading relationships, the less dependent state is viewed as the weakest link, since this state needs the relationship less. Yet, one state alone does not define a dyadic relationship. Even if one state is less constrained to use force, the state that is more constrained (more dependent

in this case) should work harder to resolve conflicts of interest before they escalate. Therefore, relationships that contain at least one heavily dependent state should be different from relationships where both states have little dependence upon each other. We would not observe these differences by looking at the least dependent state alone. For example, imagine that state A has two trading partners, state B and state C. State A is not very dependent on either partner, but state B is highly dependent on A, while state C is not. State A's degree of constraint may not vary in these two relationships, but its two partners should vary in their desire and effort to maintain peace in the relationship. Looking exclusively at state A's level of dependence would not explain the variation that would exist in these two separate dyadic relationships. In general, it is problematic to employ the characteristics of only one nation to describe the characteristics of a dyad. A dyadic relationship, by definition, reflects the interaction of two actors. This approach introduces an additional problem. Within trade-conflict studies, the less dependent state is generally the more developed, democratic state. For example, almost any dyadic relationship that contains the United States would be characterized by the dependence level for the United States. The same would be true for Germany or Japan. Does such an approach make sense? It artificially truncates the variance, as noted previously. Further, since the weakest link approach relies exclusively on data for the less dependent state (the more developed state), the resulting sample would contain a disproportionate number of dyads that contain at least one major power. Those dyads that contain only developing states, which, as mentioned subsequently, are more likely to have missing data, will be more likely to drop out of the sample than other dyads. While the issue of missing data for developing states is problematic in general, the weakest link approach exacerbates the problem of biased sampling. Page 215 → How does one decide which measure is most appropriate to employ in studies of the trade-conflict relationship? Does measurement choice affect research findings? There may be instances in which theory serves as an appropriate guide for measurement selection, but the theories employed in the trade-conflict research seldom provide definitive guidance. This means that in most instances it may be more useful to employ multiple measures for testing hypotheses. We want to be certain that our measurement choice does not affect our empirical findings, particularly if we view our measures as substitutable. Different measures may capture different dimensions of interdependence. By employing alternative measures, we obtain more information than we could acquire with only one measure. In some cases, one measure will conform more closely than another commonly employed measure with what a given scholar has in mind when positing a relationship between trade and peace or conflict. Take, for example, the case of analyzing whether trading states are more peaceful than other states. Here, the commonly used measure of trade dependence, total trade/GDP, would produce a measure where developing states rank highest. That would be appropriate if we wished to measure the extent to which a given state is dependent upon the global economy. Yet that may not be what we have in mind when we think about trading states. Instead, we may envision the most active traders in the system or those states that dominate the global trading system. Such states tend to be the most developed states in the system and tend to rank low on measures of economy dependence (total trade/GDP). If we wished to capture the extent to which states were dominant trading states, we would be better served by employing a measure that incorporated the absolute value of trade for a state or that state's share of total global trade, rather than the commonly used ratio measure of total trade/GDP. In this case, our choice of measures would lead to vastly different stories. In most instances, our conceptualizations of a given phenomenon will be the best guide to measurement selection. However, even when we can envision our ideal measures of dependence, we face practical constraints in constructing operational measures. These practical constraints may have a dramatic influence on the samples we analyze, since missing data will lead us to exclude cases. This could produce samples for our analyses, as mentioned previously, that are not as representative as we hope and that make it difficult for us to generalize about our conclusions. It may also mean that studies employing different measures of central constructs and relying on different data will lead scholars to analyze different cases of trading states. Differences in results may mean differences in the experiences of the states or relationships contained in our Page 216 → sample. We must, inevitably, select operational measures that balance our notion of the ideal measure of dependence with those for which data are readily available.

For example, GDP-based measures of dependence might be preferred in some instances, but data for GNP or GDP are available for only a small number of states in the pre–World War II period. In addition, historical data on total trade tend to be more accurate than data on GNP or GDP, since national trade statistics were recorded for most countries in the nineteenth and earlier twentieth centuries, while the data used to calculate GDP were not systematically compiled by the majority of countries until after 1950. Even then, the accuracy of these statistics for developing countries has been questioned by scholars, which leads to one of many sources of bias in accurately capturing the Third World experience. One characteristic of developing states is that a large percentage of economic activities take place outside the formal sector; this is true today but was much more the case prior to World War II. In general, scholars wishing to rely on GDP-based measures restrict the focus of their analysis and the accuracy of those measures for different classes of states. Scholars have made efforts to estimate GDP figures for the pre–World War II era and to adjust post–World War II estimates to account for some of the criticisms of these measures. Still, these estimates are far less reliable than those of recorded economic activities. Data limitations pose many barriers to the cases that we might analyze. For example, data availability is highly correlated with the level of development of a nation, particularly in the nineteenth century. This results in an inherent bias toward the disproportionate inclusion of developed states in our research samples. While most researchers seek to overcome this bias, it continues to pose a major problem for those wishing to capture a broad range of trading relationships. In addition, trade data contain their own set of problems. One problem is that these data are inherently biased in favor of important trading relationships, because data for a given state's top trading partners are more frequently reported in distinct categories than in cases of minor trading partners. Generally, the values of trade for minor trading partners are reported in an aggregated category labeled “all other countries.” Data are unavailable about which states fall within this category. Thus, it is impossible to determine from most data sources whether two states refrained from trade or simply carried on a limited amount of trade. While minor trade flows may be similar to zero trade from a statistical perspective, there is an important conceptual difference between states that have minor ties and those that may have no ties. Since contact is a necessary, but not sufficient, condition for conflict, it would be useful Page 217 → to distinguish between minor and zero trade. Unfortunately, this is not always possible. There are other ways in which data constraints might shape the trading relationships we portray, the cases we analyze, and the conclusions we reach. Trade-conflict studies that rely exclusively on International Monetary Fund trade statistics exclude states that are not members of that organization. This means that trade within the Soviet Bloc during the Cold War is excluded from many trade-conflict studies. The end of the Cold War has led to the release of data that should facilitate greater understanding about the dynamics of relationships between states in the Eastern Bloc and between those individual states within and outside the region. This will help us examine some of the issues that distinguish trade-related arguments from theories of capitalism. Although GDP- or total-trade-based measures are the most commonly used means of assessing trade dependence, researchers have proposed a number of alternative measures. For example, Polachek (1980) measures the absolute value of dyadic trade flows in his early studies of the trade-conflict relationship, while also including GDP as a separate variable among his controls for national attributes. Anderton and Carter (2001) argue that it is the value of trade, not the volume, that is germane in assessing the liberal hypothesis. Barbieri and Levy (2001) argue that it is both. In fact, the value of trade may be more consistent with economic explanations of the liberal peace, while the volume traded might be more reflective of sociological arguments that focus on increased interaction leading to peace. Polachek and McDonald (1992) emphasize the importance of using elasticity of supply and demand to measure the importance of trade flows and to capture a more realistic measure of gains from trade. Similarly, Blanchard and Ripsman (1994) develop a measure of dependence based on the strategic importance of commodities traded with a particular partner. Researchers make a valid point in stressing the need to consider the importance of commodities traded in evaluating trade dependence. However, here too practical constraints make it difficult to implement their recommendation on a general basis. Unfortunately, the elasticity measures and the measures of the strategic importance of commodities traded require information that is not available for a large number of countries or for a significant period of time. For example, a measure of strategic importance requires information about a state's

strategic needs, as well as potential sources of supply and the availability of the commodity. The strategic importance of commodities varies by nation and over time, and data reflecting those variations accurately are difficult to obtain. In addition, for national-level studies, it would be useful to assess the Page 218 → extent to which domestic and international prices converge to capture the degree to which a state has been integrated into the global economy. We could imagine a number of ways in which more complete information would advance our understanding of the trade-conflict relationship. Unfortunately, when the goal of a particular project includes generalizing findings to a broad group of states, the options available for selecting measures of interdependence become limited. There are many other weaknesses in our ability to capture information central to our conception of dependence. One important element that we would like to understand is the extent to which states have the freedom to divert trade to third-party states. This is not apparent from a trade figure alone but could be important for explaining the trade-conflict relationship. Hirschman ([1945] 1980) developed an operational measure of a state's freedom to divert trade to alternative suppliers and buyers, but employing his measure proves difficult as a result of data constraints. This is still an area that requires much more exploration. The competitiveness of a market and the availability of suppliers and buyers would be relevant to trade-conflict researchers, since limited choice may lead to greater variation in political relations among trade partners. For example, if states could trade with whomever they wished, they might opt against trading with a state whose policies they opposed; with limited choice, they might not have the luxury to be as selective. Trade may not aid states in overcoming bad political relationships between states that engage in trade only because they have little choice. While the trade-promotes-peace hypothesis suggests that commerce could help overcome deep-rooted differences, I would argue that the likelihood of conflict in this type of relationship would be much higher than between states that trade with each other out of preference and where trade merely reinforces other types of preexisting bonds. Clearly, there are many areas in which we could improve our understanding about trading relationships. Researchers attempt to capture the characteristics that they believe make trading relationships important, significant, peaceful, or conflictual. Scholars disagree about these points even before measurement construction begins. Therefore, we should not expect to have one measure of dependence. More important, we should not view having one measure of dependence or interdependence as the ideal solution. Alternative measures capture different dimensions of the phenomenon we seek to illuminate. In many respects, we still require more information, rather than less information. We have yet to capture fully the complexities that come to mind when we envision dependent and interdependent relationships. If anything, more attention needs to be paid to acquiring the information that would enable us to construct Page 219 → Models and Measures in Trade-Conflict Research 219 more realistic and complete measures of dependence and interdependence. This may further complicate our ability to reconcile inconsistent empirical findings across research studies, but it will further our knowledge about the trade-conflict relationship.

Conclusion The purpose of this chapter has been to identify areas in which theoretical differences might impact our empirical research. The images that initially come to mind when we think about trading relationships will impact our expectations about the trade-conflict relationship. This may affect our choice of conceptual measures and our focus of inquiry. Admittedly, information constraints pose real problems for trade-conflict researchers, as they do for any empirical analysis. We must be aware of the problems data constraints pose for trade-conflict researchers. This is particularly important when our choice of measures influences the samples we analyze and the conclusions we reach. Yet, equally important, we must consider the significance of the information that we have attained with existing measures. Different measures might be telling us important things about the world. We have yet to uncover the extent to which variations in empirical findings across studies reflect real differences in the tradeconflict relationship. This is an area that requires further research, and it is one that might hold many answers to the trade-conflict puzzle.

NOTES I would like to thank John Geer and Jack S. Levy for their valuable comments on this chapter. 1. Scholars also differ in their conception of what constitutes a peaceful relationship. This topic, however, is beyond the scope of this chapter. 2. Obviously, I am excluding “independent” or autarkic states that do not engage in trade and relationships that have no trade, although one could argue that this is a type of trading relationship.

REFERENCES Anderton, Charles H., and John R. Carter. 2001. The Impact of War on Trade: An Interrupted Times-Series Study. Journal of Peace Research 38:445–57. Baldwin, David A, ed. 1993. Neorealism and Neoliberalism: The Contemporary Debate. New York: Columbia University Press. Page 220 → Barbieri, Katherine. 1995. Economic Interdependence and Militarized Interstate Conflict, 1870–1985. Ph.D. diss., Binghamton University. ______. 1996. Economic Interdependence: A Path to Peace or a Source of Interstate Conflict? Journal of Peace Research 33:29–49. ______. 2002. The Liberal Illusion: Does Trade Promote Peace? Ann Arbor: University of Michigan Press. Barbieri, Katherine, and Jack S. Levy. 2001. Does War Impede Trade? A Response to Anderton and Carter. Journal of Peace Research 38:619–24. Barbieri, Katherine, and Gerald Schneider. 1999. Globalization and Peace: Assessing New Directions in the Study of Trade and Conflict. Journal of Peace Research 36:387–404. Blanchard, Jean-Marc F., and Norrin M. Ripsman. 1994. Peace through Economic Interdependence? Appeasement in 1936. Paper presented at the annual meeting of the American Political Science Association, New York. Deutsch, Karl W, Sidney Burrel, Robert Kann, Maurice Lee, Martin Lichterman, Raymond Lindgren, Francis Loewenheim, and Richard van Wagenen. 1957. Political Community and the North Atlantic Area. Princeton: Princeton University Press. Dixon, William. 1993. Democracy and the Management of International Conflict. Journal of Conflict Resolution 37:42–68. Domke, William K. 1988. War and the Changing Global System. New Haven: Yale University Press. Dos Santos, Theotonio. 1970. The Structure of Dependence. American Economic Review 60:231–36. Forbes, Hugh D. 1997. Ethnic Conflict: Commerce, Culture, and the Contact Hypothesis. New Haven: Yale University Press. Haas, Ernst B. 1958. The Uniting of Europe: Political, Social, and Economic Forces,1950–1957. Stanford: Stanford University Press. ______. 1964. Beyond the Nation-State: Functionalism and International Organization.

Stanford: Stanford University Press. Hirsch, Leonard Paul. 1986. Incorporation into the World Economy: Empirical Tests of Dependency Theory. In Dependency Theory and the Return of High Politics, ed. Charles Abel and Mary Ann Tetreault, 101–24. New York: Greenwood Press. Hirschman, Albert O. [1945] 1980. National Power and the Structure of Foreign Trade. Reprint, Berkeley: University of California Press. Hower, Gretchen. 1990. The Dynamics of Hostile and Cooperative Behavior. Ph.D. diss., University of Illinois, Urbana-Champaign. Hughes, Barry B. 1971. Transaction Analysis: The Impact of Operationalization. International Organization 25:132–45. Lenin, Vladimir Ilich. [1939] 1990. Imperialism: The Highest Stage of Capitalism. Reprint, New York: International Publishers. Marx, Karl. [1887] 1906. Capital: A Critique of Political Economy, vol. 1. Ed. Friedrich Engels. Trans. Samuel Moore and Edward Bibbins Aveling. Reprint, New York: Modern Library. Page 221 → Mitrany, David. 1964. A Working Peace System. Chicago: Quadrangle. Oneal, John R., and Bruce M. Russett. 1999. Assessing the Liberal Peace with Alternative Specifications: Trade Still Reduces Conflict. Journal of Peace Research 36:423–42. Polachek, Solomon W. 1980. Conflict and Trade. Journal of Conflict Resolution 24:57–78. ______. 1992. Conflict and Trade: An Economics Approach to Political International Interactions. In Economics of Arms Reduction and the Peace Process, ed. Walter Isard and C. H. Anderton, 89–120. Amsterdam: NorthHolland. Polachek, Solomon W., and Judy McDonald. 1992. Strategic Trade and the Incentive for Cooperation. In Disarmament, Economic Conversion and Peace Management, ed. Manas Chatterji and Linda Forcey, 273–84. New York: Praeger. Rosecrance, Richard. 1986. The Rise of the Trading State: Commerce and Conquest in the Modern World. New York: Basic Books. Savage, J. Richard, and Karl W. Deutsch. 1960. A Statistical Model of the Gross Analysis of Transaction Flows. Econometrica 28:551–72. Selfridge, H. Gordon. 1918. The Romance of Commerce. London: John Lane, the Bodley Heard. Spiegel, Henry W. 1991. The Growth of Economic Thought. 3d ed. Durham and London: Duke University Press. Viner, Jacob. 1937. Studies in the Theory of International Trade. New York: Harper and Brothers Publishers. Waltz, Kenneth. 1979. Theory of International Politics. New York: Random House.

Page 222 →

Preferential Peace: Why Preferential Trading Arrangements Inhibit Interstate Conflict Edward D. Mansfield The relationship between foreign trade and political conflict is a longstanding source of controversy among scholars of international relations, and the rapid expansion of international economic activity in recent decades has stimulated renewed interest in this topic. Echoing arguments first advanced centuries ago, many contemporary observers have expressed hope that rising trade flows will increase economic interdependence and thereby inhibit hostilities. This view, however, has prompted substantial criticism, and the strength and nature of trade's effect on interstate conflict remain the subject of fierce debate. Lately, various empirical attempts have been made to resolve this debate, but these studies have focused primarily on the links between trade flows and military disputes. In doing so, they have largely ignored how the institutional context in which foreign commerce is conducted bears on the relationship between trade and conflict. Preferential trading arrangements (PTAs) have been instrumental in shaping this context during the period since World War II. In this chapter, I argue that parties to the same PTA are less prone to interstate conflict than other states. Moreover, I maintain that PTA membership affects the relationship between the flow of overseas trade and military confrontation. Although the amount of commerce conducted by states that are not members of the same PTA has relatively little influence on their prospects of becoming embroiled in hostilities, the likelihood of antagonism between parties Page 223 → to the same preferential grouping dips precipitously as trade increases between them.

The Growth of Commercial Regionalism Preferential trading arrangements include free trade areas, customs unions, and common markets, as well as other economic groupings (Anderson and Blackhurst 1993; de Melo and Panagariya 1993). These arrangements vary along a number of institutional dimensions. For example, both customs unions and free trade areas eliminate trade barriers among members, but the former establish a common external tariff, while the latter do not. Common markets not only liberalize trade among participants but also facilitate the movement of factors of production among them. Despite these differences, however, all such arrangements grant each participant preferential access to every other member's market. Preferential groupings have long been important features of the international political economy, but they have become increasingly pervasive of late. Over fifty of them have been formed since World War II, and almost every member of the World Trade Organization (WTO) is currently party to at least one (Mansfield and Milner 1999). These developments have prompted heated debates about whether the proliferation of PTAs will generate economic conflicts among competing trade blocs that will fragment the multilateral economic system or whether this process will serve as a building block to greater economic cooperation and a strengthened multilateral system (Bhagwati 1993; Bhagwati and Panagariya 1996; Krugman 1993; Lawrence 1996; Pomfret 1997). Equally heated are debates about why PTAs have proliferated (Frankel 1997; Mansfield 1998). But whereas many empirical studies have analyzed the effects of preferential commercial arrangements on international economic relations, very little empirical research has addressed their effects on political-military relations. It is on the latter topic that I will focus in this chapter. The notion that PTAs might influence the security of member states is hardly new. Near the end of the nineteenth century, for example, Wilfred Pareto argued that “customs unions and other systems of closer commercial relations [could serve] as means to the improvement of political relations and the maintenance of peace” (quoted

in Machlup 1977, 143). At the conclusion of World War I, John Maynard Keynes (1919, 249) made a similar point, noting that “a Free Trade Union, comprising the whole of Central, Eastern, and South-Eastern Europe, Siberia, Turkey, and (I should hope) the United Kingdom, Egypt and India, might do as much for the peace and prosperity of the Page 224 → world as the League of Nations itself.” Since World War II, various scholars have again argued that PTAs dampen hostilities between participants, and policymakers have established various preferential groupings in an effort to stimulate political cooperation as well as economic gains (Fernández and Portes 1998; Nye 1971; Organization for Economic Cooperation and Development 1993, 25–26). The best-known attempt to foster peace through economic integration was, of course, the European Community (EC), whose founders sought through its creation to render war between France and Germany not only “unthinkable, but also materially impossible” (Swann 1984, 17). But the desire to promote political cooperation also contributed heavily to the formation of the Association of Southeast Asian Nations (ASEAN), the Economic Community of West African States (ECOWAS), and Mercado Común del Sur (MERCOSUR), among other preferential trading arrangements. Although various observers have argued that PTAs inhibit conflict and achieving this end has been a key objective of certain groupings, whether parties to the same PTA actually tend to be more pacific than other states is far from clear. While some preferential arrangements (like the EC) have experienced little political tension among members, others have been marked by periods of considerable discord (for example, the Central American Common Market and the East African Common Market). Moreover, the scarcity of systematic empirical research on this topic is striking. As noted in the introduction to this volume, various statistical studies have addressed the effects of trade flows on the incidence of hostilities, but very few of them directly assess the influence of international institutions designed to shape economic exchange. This gap in the literature is surprising not only because a number of prominent observers have speculated that PTA membership reduces the prospect of interstate antagonism. Equally important is the fact that many of the classical liberal arguments that these studies aim to test center squarely on the effects of free trade (Cain 1979; Doyle 1997). Directly testing these arguments requires a measure of trade policy among states, not just an analysis of trade flows (McDonald 2002). Whether or not states participate in the same PTA is one of the very few measures of this sort (though it is obviously a crude one) since such arrangements liberalize trade among members.

Preferential Trading Arrangements and Conflict In a number of recent studies, Jon C. Pevehouse, David H. Bearce, and I (Mansfield and Pevehouse 2000, 2003; Mansfield, Pevehouse, and Bearce Page 225 → 1999–2000) attempted to fill this gap in the literature by conducting some preliminary analyses of the relationship between PTA membership and the outbreak of military disputes. In these studies, we argued that researchers who ignore PTAs risk arriving at misleading conclusions about the influence of commerce on conflict. PTAs inhibit political-military disputes between members, and the pacifying effects of these arrangements grow more pronounced as trade flows rise between the participants. In the same vein, the high levels of trade that liberals view as contributing to peace are much more likely to reduce hostilities between parties to the same PTA than between other states. One reason why PTAs help to dampen conflict is that hostilities threaten the economic benefits that states expect to realize from membership. Central to these expected benefits is that PTAs cut trade barriers among members and restrict their ability to raise such barriers in the future. Consequently, acceding to a PTA helps a country to insure against the possibility that other members of the arrangement will become increasingly protectionist in the coming days (Fernández and Portes 1998; Mansfield 1998; Whalley 1998; Yarbrough and Yarbrough 1992). Moreover, the desirability of obtaining such insurance rises if the other members include this country's key trade partners, since heightened protectionism on their part would be especially harmful. The expected benefits of joining a PTA also grow if states fear being isolated from lucrative foreign markets unless they belong to one, a concern that has contributed heavily to the recent proliferation of states entering preferential groupings (de Melo and Panagariya 1993, 5–6; Yarbrough and Yarbrough 1992, 105–6). In addition to insuring against the breakdown of trade relations with key commercial partners, PTAs hold out the promise of increased investment for members. Firms locating investments in a country that belongs to a PTA gain

preferential access to every other participant's market as well. Further, since PTAs reduce the ability of governments to behave opportunistically, firms investing in a state that belongs to such an arrangement limit the prospect that their assets will be threatened by state actions (Fernández and Portes 1998; Yarbrough and Yarbrough 1992). Indeed, even if there are few economic complementarities among a group of states—and therefore few efficiency gains to be derived by lowering their trade barriers—the group may expect that forming a PTA will prompt a surge in foreign investment.1 Equally, countries forming a preferential arrangement often expect to bolster their position in international economic negotiations, since the grouping is likely to have more leverage with respect to third parties than any individual member. Hence, there are a wide variety of benefits that states entering a PTA have Page 226 → reason to anticipate. Interstate conflict can scuttle these expected gains by undermining commitments to sustain commercial liberalization, inhibiting investment on the part of firms that are reluctant to operate in unstable regions, and damaging the bargaining power of members in negotiations with third parties. The adverse economic consequences of hostilities within PTAs are generally well understood by the members of such arrangements. For example, it is widely recognized that the participants in MERCOSUR anticipated deriving substantial economic gains from this grouping (Guedes da Costa 1998; Peña 1993). Since its members also realized that MERCOSUR's economic goals would be very difficult to meet in the face of political antagonism, they implemented a series of measures intended to facilitate cooperation in the Southern Cone (Smith 1993). These measures have contributed to the diminution of tensions between Brazil and Argentina. More generally, Monica Hirst (1998, 113) points out that “regional security cooperation has become a spill-around effect of the expansion of economic ties among Southern Cone countries.” Even if little trade is conducted between PTA members and they obtain few economic gains from membership—a situation that has marked various preferential arrangements—these groupings can inhibit conflict if participants expect to derive substantial economic benefits in the future. That countries often attempt to revive moribund PTAs and seldom abandon them without establishing a replacement suggests that states in unsuccessful groupings frequently expect to realize future economic gains from membership. When countries trade extensively, however, the future benefits of membership are likely to appear especially large. Preferential arrangements deepen integration and avert the erosion of economic relations, thereby holding out the promise that trade between key partners will continue to flourish. Since hostilities threaten to scuttle existing trade relations as well as future economic exchange, states that trade heavily and participate in the same PTA have a particularly strong incentive to avoid conflict. While political disputes may not immediately affect either trade flows or a PTA, conflicts can drag on and escalate, gradually damaging the grouping and economic relations among its members. The upshot is that PTA members marked by extensive commercial ties have good reason to avoid the use of force. Preferential arrangements help to dampen conflicts through other means too. Centrally important is that such arrangements establish a forum for bargaining and negotiation among members that assists them in resolving disagreements before open hostilities break out (Nye 1971, 109). Many preferential arrangements, for example, have created dispute-settlement mechanisms to mediate economic conflicts. These mechanisms may be particularly useful for Page 227 → states that trade heavily: extensive commercial ties increase the costs of military conflict but can also stimulate economic disputes that, unless contained, may promote political tensions (Stein 1993, this volume). Preferential arrangements help to limit and resolve economic conflicts before they degrade political relations. Many of these arrangements also mediate political tensions that arise between participants, thereby limiting the resort to force. For example, ASEAN is widely credited for managing discord in Southeast Asia (Huxley 1996; Snitwongse 1998). MERCOSUR has performed a similar function throughout the Southern Cone (Manzetti 1993–94; Smith 1993). Finally, it is widely argued that concerns over the distribution of the gains from economic exchange are a substantial impediment to international cooperation (Grieco 1990; Mastanduno 1991; Mearsheimer 1995). PTAs help to address such concerns by fostering reciprocity among members (Anderson and Blackhurst 1993, 5). These arrangements attempt to ensure that economic concessions made by one participant are repaid rather than exploited by its counterparts (Fernández and Portes 1998, 213). Preferential groupings also address such concerns by facilitating the flow of information about the gains and losses accruing to members, thereby reducing their

uncertainty about the distribution of benefits stemming from economic activity. More generally, PTAs can establish focal points that forestall breakdowns in cooperation by shaping states' expectations about what constitutes acceptable behavior and facilitating the identification of deviations from such behavior (Garrett and Weingast 1993).

The Evidence To test these arguments, Pevehouse, Bearce, and I have conducted a series of analyses. Initially, we built on a study by John R. Oneal and Bruce M. Russett (1997), who examined the effects of democracy and commercial interdependence on militarized interstate disputes (MIDs) during the period from 1950 to 1985. They argued that, for the purposes of explaining bilateral disputes, primary attention should be directed at “politically relevant” dyads—that is, pairs of contiguous states and pairs including a major power—since other dyads rarely have either a reason or the capacity to fight.2 Based on an analysis of these pairs, Oneal and Russett found that both democracy and interdependence, as well as other factors, strongly influence the likelihood of MIDs. They did not, however, account for the effects of PTAs. To gain a very general sense of whether participation in a PTA affects the incidence of hostilities, we identified whether each pair of states included in Page 228 → Oneal and Russett's sample participated in the same PTA and whether a MID broke out between them for every year from 1950 to 1985. We then compared the number of MIDs occurring between PTA members to the number occurring between other states. The results shown in table 1 indicate that, on average, PTA members have become involved in MIDs only about three-quarters as often as other countries, thus providing some preliminary support for our argument. To more fully analyze the relationship between participation in a preferential grouping and conflict, we estimated a model of military disputes that included democracy, a measure of economic interdependence, interstate power relations, political-military alliances, and economic growth, as well as a dummy variable indicating whether, for each year from 1950 to 1985, each given pair of countries participated in the same PTA (Mansfield, Pevehouse, and Bearce 1999–2000). We found strong evidence that PTA membership reduces the likelihood of a military dispute. Based on our estimates, states that are not members of the same PTA are between one-third and one-half more likely to engage in military disputes than PTA members. Interestingly, our results also indicated that Oneal and Russett's measure of economic interdependence had little bearing on disputes. In a subsequent set of studies, we addressed whether PTAs influence the relationship between trade flows and disputes. In the first study (Mansfield and Pevehouse 2000), we again analyzed Oneal and Russett's data on politically relevant dyads. As noted in the introduction to this volume, many researchers use the ratio of bilateral trade to gross domestic product (GDP) as a measure of interdependence. Oneal and Russett took that tack too.3 However, we argued that it is more appropriate for dyadic studies of conflict to break down this measure into its constituent parts: the GDP of each trade partner and the volume of commerce between them. One reason is that studies relying solely on the ratio of trade to GDP implicitly assume that commercial flows and national income have an interactive effect on hostilities and ignore the independent effects of these factors. Yet an assessment of the independent effects of trade and national income is needed to ensure that any observed relationship between the ratio of bilateral trade to GDP and military disputes does not stem from the influence of national income alone. This issue merits careful consideration since economically larger states tend to be politically powerful and strong countries are disproportionately prone to belligerence (e.g., Organski and Kugler 1980). Page 229 → To address whether PTA membership influences the relationship between trade flows and hostilities, we analyzed a model of military disputes that included a dummy variable indicating whether each pair of states in our sample participated in a PTA, the volume of trade between them, and the interaction between these factors. Our results provided strong evidence that trade flows have little bearing on conflict between states that do not belong to the same PTA. For PTA members, however, heightened commerce does indeed inhibit hostilities. In addition, the pacific effects of PTA membership grow increasingly large as the volume of trade rises.

Figure 1 illustrates these results. It shows the predicted probability of a dispute for PTA members and for other states, respectively, varying the annual level of trade from $0 to $10 billion, and holding constant the remaining variables in our model.4 Clearly, increasing the value of trade does more to reduce the prospect of antagonism between PTA members than between other countries. Increasing the annual amount of commercial activity from $1 billion to $5 billion, for example, yields roughly a two-thirds reduction in the likelihood of conflict between parties to the same PTA, but only about a 10 percent reduction in the likelihood of a dispute between states that do not participate in such an institution. Equally, the probability of belligerence between states conducting $1 billion of commerce annually is more than one-third lower if the trading partners belong to the same PTA than if they do not. Non–PTA members that engage in $5 billion of trade are about five times more likely to engage in conflict than members of a preferential grouping, and non–PTA members conducting $10 billion of commerce are about twenty times more likely to do so. Our final study included a broader test of the arguments presented earlier (Mansfield and Pevehouse 2003). In it, we analyzed all country-pairs rather than only politically relevant dyads, we examined a somewhat longer period of time (1950–92) than before, and we compared our results across different data on trade flows.5 Our results continued to indicate that PTAs inhibit disputes between participants, that preferential groupings are increasingly likely to Page 230 → dampen conflict as trade flows rise, and that high levels of trade are more likely to reduce hostilities between PTA members than between other states. Finally, in interpreting the results of these studies, it is important to recognize that the effects of preferential trading arrangements do not seem to reflect any more general tendency for intergovernmental organizations (IGOs) to dampen conflict. Russett, Oneal, and David R. Davis (1998) found considerable evidence that the more IGOs in which a pair of states hold common membership, the lower the likelihood of a military dispute between them. However, we included their measure of IGO membership in our model and found no evidence of this sort (Mansfield and Pevehouse 2000). Equally, our findings do not stem from the influence of the General Agreement on Tariffs and Trade (GATT). We included a variable indicating whether both states in each country-pair were parties to GATT and found that it also had little bearing on hostilities (Mansfield and Pevehouse 2000, 2003). Moreover, accounting for the influence of IGO membership and participation in GATT had no impact on the observed relationships between PTAs and trade flows, on the one hand, and political-military conflict, on the other. That PTAs have a more pronounced impact on military disputes than GATT may reflect the tendency for preferential groupings to do a better job Page 231 → than global institutions of providing certain political and economic benefits to member states, thereby rendering participants especially hesitant to jeopardize these groupings by engaging in conflict. For example, recent research indicates that PTAs are often better able than GATT to (1) solve the time-inconsistency problems faced by governments that want to make a credible commitment to trade liberalization or political reforms and (2) serve as a focal point for actors with a preference for free trade (Bagwell and Staiger 1997; Fernández and Portes 1998).6 Equally, participation in a PTA is likely to provide a state with insurance against any substantial weakening of GATT—especially if the arrangement includes its key trade partners—since the state would continue to have open access to the markets of other PTA members. In addition, such participation enhances a state's bargaining power in GATT negotiations (Mansfield 1998; Whalley 1998).7 Consequently, it is not surprising that PTA membership has a stronger influence on hostilities than GATT membership.

Conclusions During the past decade, a burgeoning empirical literature has emerged on the relationship between international trade and political conflict. Most of this literature addresses the influence of bilateral trade flows on military disputes. By itself, however, this tack is inadequate, since it takes no account of the institutions set up to guide trade relations. Preferential trading arrangements are centrally important in this regard. In this chapter, I have argued that they dampen military disputes between members and have a strong bearing on the relationship between trade flows and conflict.

Preferential groupings help to mute military tensions by generating the expectation of future economic gains on the part of members. Since the outbreak of hostilities threatens to scuttle these gains, participants in the same PTA have reason to avoid involvement in military conflicts. In addition, many preferential groupings create a forum for bargaining and negotiation that reduces tensions among participants, helps to resolve conflicts that do occur, and promotes the establishment of focal points that shape states' expectations and facilitate the identification of deviations from accepted norms. Moreover, the tendency for preferential arrangements to inhibit disputes is likely to become more pronounced as trade flows rise. For PTA members that trade extensively, the future stream of gains from membership is likely to seem particularly large—thereby creating an especially potent deterrent to military disputes—and institutional mechanisms exist for resolving tensions that do emerge before open hostilities break out. The findings of a number of recent studies accord with these arguments. For Page 232 → states that do not belong to the same PTA, the flow of trade has only a weak impact on hostilities. For PTA members, however, rising commerce strongly reduces the likelihood of military conflict. Furthermore, parties to the same PTA are less prone to engage in military disputes than other states, an influence that grows increasingly large as the flow of trade expands. These results indicate that the relationship between commerce and conflict depends on the institutional context in which trade is conducted. This relationship is therefore more complex than liberals suggest. At the same time, however, there is little support for the claim made by realists and many others that trade has no systematic bearing on belligerence. While the available evidence points to the importance of PTAs in dampening political tensions, it is clear that some arrangements have been more effective in this regard than others. Further work is needed to address why this is so. Such research might focus on whether institutional variations across PTAs, the political and economic characteristics of members, and the stability of the international political economy bear on the effects of PTAs and trade flows on conflict. Equally important is the need for empirical work on this topic covering the period prior to World War II. Finally, the studies described in this chapter analyze how PTAs influence political-military relations among members. It would be useful to examine how these arrangements affect political-military relations between members and third parties as well. Research on these issues is likely to be of more than just academic interest. The past half-century has been marked by a rapid increase in the number of PTAs and the number of countries participating in them. Both the Clinton and the Bush administrations have made spreading regional economic arrangements a foreign policy priority. While the economic effects of establishing PTAs have been debated at great length, the political effects of doing so have received far shorter shrift. However, there is ample reason to expect that the growth of these arrangements will promote peace and cooperation among member states, an important political benefit that has not been fully appreciated in debates on the consequences of regionalism.

NOTES I am grateful to David H. Bearce and Jon C. Pevehouse for allowing me to draw on our collaborative research in this chapter and to Patrick McDonald and Brian M. Pollins for helpful comments. This chapter is dedicated to the memory of Jaak Holemans. 1. For example, such was the case for members of ASEAN. See Saxonhouse 1993, 410–11. Page 233 → 2. During the period from 1950 to 1985, the major powers were China, France, Great Britain, the Soviet Union, and the United States. 3. See also Oneal (this volume) and Russett (this volume). 4. More specifically, the regime type of each state, their rates of economic growth, the distribution of capabilities between them, each state's GDP, a measure of the global distribution of power, and the length of time since the states in question last engaged in a MID are evaluated at their respective medians. Further, we assume that the states are contiguous and are not allied. 5. Making this sort of comparison is useful since recent research by Oneal and Russett (1999) and Katherine Barbieri (1996, 1998) suggests that differences across data sets on foreign trade strongly influence the observed relationship between commerce and conflict.

6. Fernández and Portes (1998, 205) speculate that PTAs have an easier time resolving time-inconsistency problems because, within GATT, the countries injured by another party's actions are responsible for identifying that party and punishing it if GATT rules in their favor. In GATT—a large organization with a diffuse structure—the incentive to take these steps is “likely to be much smaller for any single member, and the process likely to be slower and the outcome more uncertain, than within a regional arrangement. In a [PTA], it is much clearer who has the responsibility to punish, and the reputational loss from not doing so should accordingly be greater.” 7. Our tests focused on GATT and did not address the WTO because of the time period we analyzed. However, the arguments made in this paragraph apply to both GATT and the WTO.

REFERENCES Anderson, Kym, and Richard Blackhurst. 1993. Introduction and Summary. In Regional Integration and the Global Trading System, ed. Kym Anderson and Richard Black-hurst, 1–15. New York: Harvester Wheatsheaf. Bagwell, Kyle, and Robert Staiger. 1997. Multilateral Tariff Cooperation during the Formation of Free Trade Areas. International Economic Review 38:291–319. Barbieri, Katherine. 1996. Explaining Discrepant Findings in the Trade-Conflict Literature. Paper presented at the 37th annual convention of the International Studies Association, April 16–20, San Diego. ______. 1998. International Trade and Conflict: The Debatable Relationship. Paper presented at the 39th annual convention of the International Studies Association, March 18–21, Minneapolis. Bhagwati, Jagdish. 1993. Regionalism and Multilateralism: An Overview. In New Dimensions in Regional Integration, ed. Jaime de Melo and Arvind Panagariya, 22–51. New York: Cambridge University Press. Bhagwati, Jagdish, and Arvind Panagariya. 1996. Preferential Trading Areas and Multilateralism: Page 234 → Strangers, Friends, or Foes? In The Economics of Preferential Trade Agreements, ed. Jagdish Bhagwati and Arvind Panagariya, 1–78. Washington, DC: AEI Press. Cain, Peter. 1979. Capitalism, War and Internationalism in the Thought of Richard Cobden. British Journal of International Studies 5:229–47. de Melo, Jaime, and Arvind Panagariya. 1993. Introduction. In New Dimensions in Regional Integration, ed. Jaime de Melo and Arvind Panagariya, 3–21. New York: Cambridge University Press. Doyle, Michael W. 1997. Ways of War and Peace: Realism, Liberalism, and Socialism. New York: W. W. Norton. Fernández, Raquel, and Jonathan Portes. 1998. Returns to Regionalism: An Evaluation of Nontraditional Gains from Regional Trade Agreements. World Bank Economic Review 12:197–220. Frankel, Jeffrey. 1997. Regional Trading Blocs. Washington, DC: Institute for International Economics. Garrett, Geoffrey, and Barry R. Weingast. 1993. Ideas, Interests, and Institutions: Constructing the European Union's Internal Market. In Ideas and Foreign Policy, ed. Judith Goldstein and Robert O. Keohane, 173–206. Ithaca: Cornell University Press. Grieco, Joseph M. 1990. Anarchy and the Limits of Cooperation: A Realist Critique of the Newest Liberal Institutionalism. International Organization 42:485–507. Guedes de Costa, Thomaz. 1998. The Role of the Armed Forces in Brazil's Democratization. In Civil-Military Relations: Building Democracy and Regional Security in Latin America, ed. David R. Mares, 223–37. Boulder:

Westview. Hirst, Monica. 1998. Security Policy, Democratization, and Regional Integration in the Southern Cone. In International Security and Democracy, ed. Jorge I. Dominguez, 102–18. Pittsburgh: University of Pittsburgh Press. Huxley, Tim. 1996. ASEAN's Role in the Emerging East Asian Regional Security Architecture. In Fragmented Asia: Regional Integration and National Disintegration in Pacific Asia, ed. Ian G. Cook, Marcus A. Doel, and Rex Li, 29–52. Aldershot: Avebury. Keynes, John Maynard. 1919. The Economic Consequences of the Peace. London: Macmillan. Krugman, Paul. 1993. Regionalism versus Multilateralism: Analytical Notes. In New Dimensions in Regional Integration, ed. Jaime de Melo and Arvind Panagariya, 58–79. New York: Cambridge University Press. Lawrence, Robert Z. 1996. Regionalism, Multilateralism, and Deeper Integration. Washington, DC: Brookings Institution. Machlup, Fritz. 1977. A History of Thought on Economic Integration. New York: Columbia University Press. Mansfield, Edward D. 1998. The Proliferation of Preferential Trading Arrangements. Journal of Conflict Resolution 42:523–43. Mansfield, Edward D., and Helen V. Milner. 1999. The New Wave of Regionalism. International Organization 53:589–627. Page 235 → Mansfield, Edward D., and Jon C. Pevehouse. 2000. Trade Blocs, Trade Flows, and International Conflict. International Organization 54:775–808. ______. 2003. Institutions, Interdependence, and International Conflict. In Globalization and Armed Conflict, ed. Gerald Schneider, Katherine Barbieri, and Nils Petter Gleditsch, 233–50. Lanham, MD: Rowman and Littlefield. Mansfield, Edward D., Jon C. Pevehouse, and David H. Bearce. 1999–2000. Preferential Trading Arrangements and Military Disputes. Security Studies 9:96–118. Manzetti, Luigi. 1993–94. The Political Economy of MERCOSUR. Journal of Interamerican Studies and World Affairs 35:101–41. Mastanduno, Michael. 1991. Do Relative Gains Matter? America's Response to Japanese Industrial Policy. International Security 16:73–113. McDonald, Patrick J. 2002. The Invisible Hand of Peace: Capitalism, the War Machine, and Liberal IR Theory. Ph.D. diss., Ohio State University. Mearsheimer, John J. 1995. The False Promise of International Institutions. International Security 19:5–49. Nye, Joseph S. 1971. Peace in Parts: Integration and Conflict in Regional Organization. Boston: Little, Brown. Oneal, John R., and Bruce M. Russett. 1997. The Classical Liberals Were Right: Democracy, Interdependence, and Conflict, 1950–1985. International Studies Quarterly 41:267–94. ______. 1999. Assessing the Liberal Peace with Alternative Specifications: Trade Still Reduces Conflict. Journal of Peace Research 36:423–42.

Organization for Economic Cooperation and Development. 1993. Regional Cooperation and Developing Countries. Paris: Organization for Economic Cooperation and Development. Organski, A. F. K., and Jacek Kugler. 1980. The War Ledger. Chicago: University of Chicago Press. Peña, Felix. 1993. Strategies for Macroeconomic Coordination: Reflections on the Case of MERCOSUR. In The Challenge of Integration: Europe and the Americas, ed. Peter H. Smith, 183–200. New Brunswick, NJ: Transaction Publishers. Pomfret, Richard. 1997. The Economics of Regional Trading Arrangements. New York: Oxford University Press. Russett, Bruce M., John R. Oneal, and David R. Davis. 1998. The Third Leg of the Kantian Tripod for Peace: International Organizations and Militarized Disputes, 1950–1985. International Organization 52:441–67. Saxonhouse, Gary R. 1993. Trading Blocs and East Asia. In New Dimensions in Regional Integration, ed. Jaime de Melo and Arvind Panagariya, 388–415. New York: Cambridge University Press. Smith, Peter H. 1993. The Politics of Integration: Concepts and Themes. In The Challenges of Integration: Europe and the Americas, ed. Peter H. Smith, 1–14. New Brunswick, NJ: Transaction Publishers. Page 236 → Snitwongse, Kusuma. 1998. Thirty Years of ASEAN: Achievements through Political Cooperation. Pacific Review 11:183–94. Stein, Arthur A. 1993. Governments, Economic Interdependence, and International Cooperation. In Behavior, Society, and Nuclear War, ed. Philip E. Tetlock, Jo L. Husbands, Robert Jervis, Paul C. Stern, and Charles Tilly, 3:241–324. New York: Oxford University Press. Swann, Dennis. 1984. The Economics of the Common Market. 5th ed. Harmondsworth: Penguin. Whalley, John. 1998. Why Do Countries Seek Regional Trade Arrangements? In The Regionalization of the World Economy, ed. Jeffrey A. Frankel, 63–83. Chicago: University of Chicago Press. World Trade Organization. 1995. Regionalism and the World Trading System. Geneva: World Trade Organization. Yarbrough, Beth V., and Robert M. Yarbrough. 1992. Cooperation and Governance in International Trade: The Strategic Organizational Approach. Princeton: Princeton University Press.

Page 237 →

Conflict Page 238 →

Page 239 →

Trade and Conflict: Does Measurement Make a Difference? Jon C. Pevehouse The debate over trade interdependence and its influence on international conflict has intensified over the past decade. As new data concerning the levels of trade between nation-states has become more accessible, as estimation methods have advanced to analyze increasingly complex models, and as computational power has made it easy to deal with very large data sets, scholarly research concerning the impact of trade flows on interstate relations has grown dramatically. Much of this literature focuses on the impact of trade flows on the propensity of states to become engaged in military disputes involving the threat or the use of force. Although this explosion in the interdependence literature has advanced the empirical study of interdependence, there is still far from universal agreement on the relationship between trade and conflict. One difference that has surfaced among various sides in this debate concerns the issue of measurement (Oneal and Russett 1997, 1999; Barbieri 1996). Largely, however, this discussion (as well as similar differences in previous quantitative interdependence work) centers on the measurement of interdependence. What seems to have been taken for granted in this recent wave of quantitative interdependence work is the operationalization of the dependent variable—international conflict. Indeed, in their introduction to this volume, Edward D. Mansfield and Brian M. Pollins suggest that “there has been remarkably little discussion of either why this focus is theoretically appropriate or the implications of shifting among these different data sets.” This chapter will examine issues regarding the measurement and operationalization Page 240 → of international conflict. Specifically, I will argue that events data provide a potentially better measurement of conflict and cooperation to use in testing theories of interdependence. My contention is important not only for purposes of knowledge cumulation but for theory testing as well. After examining some of the empirical literature concerning trade interdependence, including how various authors have operationalized conflict, I review some problems relating to measuring international conflict. I then discuss the use of events data versus the now popular Militarized Interstate Dispute (MID) data set. Finally, I compare two of the major events data sets to the MID data set to show that varying the measurement of conflict may influence the results of our empirical investigations of commercial liberalism.

The Interdependence Debate The claim that trade inhibits conflict has existed for nearly two millennia (Irwin 1996). The nineteenth-century Manchester liberals (see Blainey 1988) brought this claim to the forefront of policy debates, and since then the argument has occupied a crucial space in international relations theory. While I will not review the causal claims and counterclaims of the commercial liberals and their opponents, it is important to note that the theoretical claims of this literature do not merely predict the absence of major war between states that are interdependent. That is, commercial liberals hold that free trade fosters a sense of international community, as well as the development of mutual respect and harmonious relations that reduce interstate tensions (Viner 1951, 261). Opponents of the commercial liberals hold that as commerce rises, so does the range of economic issues over which disputes can emerge. Kenneth Waltz (1970, 205, 222), for example, argues that since close interdependence means closeness of contact and raises the prospect of at least “occasional conflict. . . the [liberal] myth of interdependence. . . asserts a false belief about the conditions that may promote peace.” For Waltz, heightened trade could actually stimulate belligerence—especially low-level “occasional” conflict. I emphasize these particular points to illustrate that interdependence theorists expect trade and other commercial ties to influence the propensity for cooperation and for all levels of conflict. As the commercial liberal view makes clear, trade should do more than prevent violent conflict: commerce should engender cooperation among trading states. While early quantitative tests of commercial liberalism incorporated these ideas, recent studies have

largely failed to do so, focusing only on violent interstate disputes. Page 241 →

Empirical Investigations of Trade Interdependence Despite the long theoretical tradition of interdependence-related work, little systematic investigation existed until the past thirty years. A first wave of quantitative interdependence literature blossomed in the late 1970s and early 1980s. Most, but not all, of these early studies supported the liberal hypothesis that trade was correlated with peace.1 For example, based on an analysis of thirty pairs of states during the period from 1958 to 1967, Solomon W. Polachek (1980) concluded that higher levels of trade dampen conflict. Mark Gasiorowski and Polachek (1982) found that heightened commerce between the United States and the Soviet Union from 1967 to 1975 contributed to a reduction in tensions and the onset of détente. Later work by Gasiorowki (1986), however, found that trade flows were inversely related to the onset of conflict during the Cold War but that conflict became more pervasive as the costs of severing commercial relations rose, casting doubt on the classic liberal argument. What all of these studies have in common is their use of events data (specifically Edward E. Azar's Conflict and Peace Data Bank—COPDAB; Azar 1993) as the measure of conflict and cooperation. Although there was some criticism of the actual methodology employed to use the data (see Gasiorowski 1986, 27–8), COPDAB and Charles A. McClelland's (1978) World Event Interaction Survey (WEIS) were the best options to measure conflict and cooperation at the time. The second wave of quantitative interdependence research, however, would move away from this dependent variable.2 Work by John R. Oneal et al. (1996) argued that higher levels of trade inhibited military disputes during the era from 1950 to 1985, especially between contiguous states. These results have been confirmed in subsequent studies by Oneal and Bruce M. Russett (1997); Russett, Oneal, and David R. Davis (1998); and D. Scott Bennett and Alan Stam (2000). Other studies, however, have cast doubt on these findings. Katherine Barbieri (1996), for instance, found that higher levels of commerce increased the prospects of hostilities between 1870 and 1938 but inhibited conflict throughout the post–World War II era. Mansfield and Jon C. Pevehouse (2000) find that the effect of trade is conditioned by the presence of regional trading institutions and that outside these institutions, trade has little influence on political-military hostilities. Erik Gartzke (2000) finds that controlling for state preferences (measured by the similarity in United Nations voting patterns) minimizes the impact of trade on disputes. Thus, the liberal interdependence argument is by no means universally supported. Page 242 → Although each of these studies finds differing support for the theoretical argument, all of these studies utilize MID data to measure the level of conflict and cooperation. There is one set of the recent interdependence literature that does utilize events data to measure interdependence. Rafael Reuveny and Heejoon Kang (1996) use events data and Granger Causality to show that the relationship between conflict and trade is both reciprocal and dyaddependent. They utilize COPDAB data to measure conflict and cooperation between a subset of dyads. Although this study does use events data, because the sample of states and model specification vary from other empirical investigations, it is difficult to compare this work to the other quantitative work in the field. While almost every one of these studies addresses the issue of measuring the concept of interdependence, none of them problematizes the nature of the dependent variable. This is not surprising given early problems in gathering data on trade dependence and a general debate over what operationalization best captured the concept itself. Various authors have differed in their definitions of dependence, sensitivity, and vulnerability (for an overview see Keohane and Nye 2001). While it is certainly important to operationalize these concepts in a careful manner, these debates have pulled attention away from measurement issues of the dependent variable.

Measuring International Conflict and Cooperation Most studies of the trade and conflict nexus rely on conflict data drawn from the MID project (see Gochman and

Maoz 1984). The MID data measure “united historical cases” in which one state directs at least the threat of military force against another state, its territory, or representatives (Jones, Bremer, and Singer 1996, 168). As most scholars engaged in quantitative research in international relations are familiar with the MID data, I will not provide an in-depth discussion of the data set here.3 Another major source of data on international conflict is events data sets, two of which were coded throughout the 1970s and were more widely used in the first wave of interdependence research. WEIS and COPDAB are data sets made up of discrete interactions between states. Both sets utilized human coders to code interstate interactions out of widely available news sources. Both of these data sets were meant to code all interstate interactions, as opposed to more episodic events data sets such as the Behavioral Correlates of War (BCOW; see Leng 1993).4 In theory, MID data should represent aggregated episodes that should be found in either events data set: because MIDs were Page 243 → coded in their distinct crisis phases, then combined into larger episodes, constituent components should exist in WEIS or COPDAB. One reason for the wide use of the MID data set over events data sources such as WEIS and COPDAB is the availability of the data. Larger sample sizes carry obvious advantages in statistical investigation, and this is an attractive characteristic of the MID data set. MID data cover dyadic interactions among nearly all states from 1816 to 1992, but both major events data sets have temporal limitations. COPDAB's coverage ranges from 1948 to 1978, while WEIS covers 1960 to 1991. Although most recent studies of trade data begin in 1950 and progress through 1992, Oneal and Russett's (1999) most recent work extends the data backward temporally until the late nineteenth century. The MID data set is able to provide more data on conflict interactions over the relevant period. There are spatial-overlap issues as well. Although in theory events data sets cover all possible interactions, many dyads within the period of observation have few, if any, events in either WEIS or COPDAB. For example, the Oneal and Russett (1997) data on trade and conflict contains 20,990 observations on “politically relevant” dyads between the years 1951 and 1985. If one merges the relevant WEIS data into the data set, there are only 4,788 overlapping observations (77 percent of the 20,990 are missing). For COPDAB, 9,541 cases match, leaving slightly over 50 percent of the data (54 percent) missing.5 Thus, MIDs clearly provide more complete data. However, more data does not necessarily translate into more appropriate data. Two possible issues arise from the use of the MID data. First, some disputes that do not involve violence or the explicit threat of violence may be missed by the data. The five-tier intensity scale for the level of violence associated with the MID data set allows a range of hostile interactions to be coded, yet few nonviolent episodes exist in the data. The vast majority of MIDs fall into the show-of-force or use-of-force category, suggesting that most disputes in the international system turn violent. While this is certainly a theoretical possibility, this seems a bizarre conclusion, given a cursory glance at everyday world politics. One possibility is that there are selection issues that arose during the coding of the MID data. A brief comparison of the three data sets suggests that this is a distinct possibility. MID equivalents for the events data sets can be computed by charting the coding rules of the MID data into their WEIS and COPDAB equivalents. Table 1 shows which events match between the data sets.6 If these events occur between dyad members in either events data set, that dyad-year is coded as experiencing a MID-like event. Table 2 shows the number of MID-like events Page 244 → that are present in the Oneal and Russett sample. In their data there are 947 MIDs in total, which constitutes 4.5 percent of all observations.7 As table 2 illustrates, COPDAB and WEIS show much more frequent conflict events in those dyads that have nonmissing data. For the WEIS data, 880 dispute-equivalent events exist, which (given the smaller sample size) shows that roughly 18 percent of the observations display MID-like behavior. For COPDAB, 2,596 dyad-years show dispute-equivalent events, which is over 27 percent of the observations for the overlapping data. Thus, both events data sets show a much higher propensity for conflict between dyads. Another method for comparing these data sets is to break down the frequency of various levels of hostility. For example, does either data set find threats of force or displays of force more frequently than the MID data or vice versa? To check this possibility, I utilize the same event-to-dispute mappings shown in table 1. I then compute the

number of each dispute type (threat of force, display of force, use of force) across the three data sets on a yearly basis. Two conservative coding decisions were made in order to enhance the prospects that the data sets would show similar trends of conflict. First, I counted only one incident per dyad-year. Thus, if two states were coded as engaging in the use of force by the COPDAB data ten times during the year, this was counted as merely one use of force for the entire year. This was done since the MID data could have coded multiple disputes for that year, yet the Oneal and Russett data included only one dispute. Another possibility is that since the MID data set aggregates events into crisis episodes, all ten uses of force captured by the COPDAB data could be aggregated into one MID. Page 245 → Second, I chose to include threats or displays of force in the events data sets only when there was no event involving the use of force in the same month. Because of the episodic nature of the MID data, the only variable reported in the MID records is the highest level of hostility reached for a given dispute. Thus, one explanation for the appearance of the MID data as missing lower-level disputes is that these threats and displays of force are aggregated into violent MIDs once they move to the use-of-force category. As mentioned earlier, this would imply that most interstate threats and displays of force end in militarized conflict, given that the vast majority of MIDs involve the use of force. Nonetheless, this disaggregation would still explain why more low-level disputes exist in the events data sets. While excluding threats and displays of force in the same month when the use of force occurs should help to ameliorate this potential problem, it is still possible that the events data will show low-level disputes that were included in a MID. For example, if a dispute continues for five months and ends in a use of force, up to five threats of force could be coded by COPDAB or WEIS yet be missing in the MID data since the threats would be part of a larger, highly aggregated violent dispute. This is not a critique of the MID coding scheme, but rather a caveat as one compares events across these data sets.8 Figures 1 to 3 show comparisons of the three levels of MIDs: figure 1 compares threats of force, figure 2 compares displays of force, and figure 3 compares uses of force. Figure 1 shows the large number of threats measured in the COPDAB data compared to the MID and even the WEIS data. While these later data sets show quite low levels of threat behavior during the period of observation, COPDAB records a significantly larger number of dyads engaging in threat-based behavior. Figure 2 shows a similar trend, where COPDAB consistently shows a much higher number of displays of force than WEIS or MID data. Although the differences are generally much smaller in this category of conflict behavior, they remain quite noticeable. Figure 3 shows a much higher level of consistency across each of the three data sets. With respect to the use of force, each data set tracks quite closely with the others. This is especially reassuring for those scholars who have utilized Page 246 → only category 4 and 5 MIDs in their research (e.g., Gowa 1999). This consistency between the MID data and the other data sets could be taken as evidence that each set is consistently measuring the same conflict phenomenon, but because the MID data set aggregates disputes to their highest level of hostility, it appears that far fewer low-level disputes occur when using the MID data. If this is the case, there may be little advantage gained in moving to events data for the purposes of capturing more nuanced conflict behavior. Alternatively, this high correlation among all data sets could be the result of selection: violent disputes are more likely to be reported by the mainstream media and therefore picked up by researchers.9 Thus, if one only wishes to code episodes of violence between states, one may be likely to miss the “dogs that don't bark” (or, in this case, only growl). If one is coding an entire range of diplomatic behavior, it is plausible that one will more readily find this type of behavior occurring among everyday interstate interactions, which may not make it onto the radar screen of the media and, therefore, researchers.10 Unfortunately, we cannot tell whether the low-level disputes are missing or are aggregated into temporally aggregated episodes. Thus, the MID data may be missing some disputes, even in the smaller subsample in the Oneal and Russett data set. This could cast doubt on the findings of the interdependence and peace literature. Of course, some of the extra disputes could be miscoded in either events data set, or, if the dispute-equivalent events were highly isolated, it may not be important to include them. Yet, given the high disparity between the frequencies of “disputes” in these three data sets, one must at least question whether some disputes (especially low-level disputes) are missed in the

MID data. Page 247 → Page 248 → From a theoretical level, it is essential to include these low-level conflicts, as this could help to flesh out realist ideas concerning the promotion of conflict through trade. Waltz's (1970) logic states that because states interact on more levels with higher trade, there are more issues over which states may “fight.” Yet, the average observer of the postwar system would usually not expect these differences to spill over into armed military conflict or even threats of military conflict. That is, no matter how nasty the “banana war” gets, no one expects a resort to violence. Hostility, nonetheless, may exist between states over trade issues. By looking at “lesser” dispute-related events, both the realist and the liberal ideas could be correct. States may engage in diplomatic bickering more, yet the overall levels of hostility may not rise above a certain level. By using only MID data this dynamic would be missed. Tracking Cooperation and Conflict The second problem with the use of the MID data lies in the fact that it is only conflict data. Traditional theories of interdependence hold that trade interdependence may lead not only to less conflict but to higher levels of cooperation. The use of the MID data would capture only half of the picture. While this is certainly not a problem with the data itself (the data set was not intended to track cooperative behavior), it could be an issue for testing the theory at hand. It is possible, for example, that high levels of trade quicken the de-escalation process or make disputes between states quite brief and that cooperation will ensue after these “spikes” of conflict. In only measuring conflict, however, this type of dynamic would go undetected. Again, an examination of the data indicates that a very dynamic process may be occurring under situations of trade interdependence. If one converts the COPDAB and WEIS data to interval-level data using their intensity weights, then aggregates the data into net cooperation scores, one can get a feel for the overall level of conflict and cooperation in the dyad-year.11 Using this method, it would appear that there is very little value added in using events data. When Page 249 → a MID is present in the Oneal and Russett data, the mean COPDAB net cooperation score for that dyad-year is –80. For the WEIS data, the net cooperation score is –151. Thus, adding the cooperative side of political interactions may yield similar results to studies that use MIDs only. Yet, upon closer inspection, the dynamic is more complex. For the WEIS data, there are 443 cases of MIDs. In 121 of those 443 dyad-years (27 percent) there is positive net cooperation. For the COPDAB data, there are 575 dyad-years of MIDs, 252 of which show positive net cooperation (43 percent). Thus, there are clearly significant variations in behavior in these dyads. This indicates the importance of moving to the analysis of sequences of conflictual and cooperative events to get a deeper understanding of how trade may influence the process of conflict escalation and de-escalation. Although scholars have begun to address these questions using existing trade and conflict data sets (see Reed 2000), these descriptive statistics indicate that the use of the MID data may not be the best way to approach this particular question. Conflicts may escalate and de-escalate within a year. Thus, incorporating cooperative data into measures of interstate behavior may more appropriately capture the overall tenor of relations between states. Of course, depending on the question under investigation, one may or may not be interested in this dynamic. If one wishes to test the theory that trade inhibits any conflict between states, capturing only conflict between states may be appropriate. If, however, one wishes to test the broader observable implications of commercial liberalism or the influence of trade on the de-escalation of disputes, then one would certainly want to include a measure of cooperation.

Conclusions Although the descriptive statistics presented here should not be taken as the final answer to the question of how to operationalize international conflict, the evidence does suggest that looking at disputes not captured by MID coding as well as examining the impact of trade on cooperation would be fruitful. On one hand, this is not a new

argument—the initial quantitative studies of interdependence took this very track in analyzing the influence of trade on conflict. Most recent statistical research, though, exclusively uses the MID data. I suggest that it would be fruitful to revisit the events data realm in order to expand the scope of the questions investigated in the trade and conflict debate. As pointed out by Mansfield and Pollins, in their introduction to this volume, by more explicitly linking theories of interdependence to various stages of international Page 250 → conflict, new analyses may add depth to our understandings of the dynamics of trade and conflict. As I have shown, because MIDs tend to be highly aggregated, their appropriateness for these tasks may be limited. Of course, events data are not without their own drawbacks. Apart from the important issue of data availability and the potential bias this would introduce in a sample, some have criticized the overall quality of the events data. The data are often noisy, and this is problematic when engaging the first type of comparison presented here—one miscoded dispute event could be counted as a full dispute using the WEIS or COPDAB data. Although there are potential solutions to this problem, it is one that researchers should be cognizant of when utilizing this data. In addition, part of the attraction of events data is that they are a rich source of political interactions—here, aggregating them to the yearly level washes out most of the interesting interactions and dynamics within the data. Unfortunately, other economic data, such as trade and gross domestic product (GDP), are computed at yearly intervals, which makes it difficult to disaggregate to “lower” temporal density. Nonetheless, if one could somehow take advantage of the sequenced nature of the events, one could test hypotheses about the influence of trade on escalation processes or strategic interaction games. Ideally, one would expect different types of sequences under different levels of interdependence. Although one would have to move away from the pooled time-series setup to accomplish this, this would promise to provide a more microlevel understanding concerning the dynamics between trade and conflict. Another prospect for future research would be to utilize the richness of the COPDAB or WEIS data in count models. Rather than simply coding the presence of a dispute (of any type) as a 0/1 dummy variable, one could count the number of various types of disputes to gain a sense of the intensity of conflict or cooperation between states. A final suggestion for future research concerns disaggregating the bilateral nature of existing trade and conflict data sets. In most recent empirical studies, researchers measure trade in total bilateral flows (country A → B + country B → A). The International Monetary Fund's Direction of Trade statistics, however, report trade in directed form, just as WEIS and COPDAB report conflict and cooperation in such a directed form. Thus, some purchase could be made if one were to examine directional trade and conflict in a pooled time-series setup. In conclusion, this chapter has suggested a number of research avenues that can be opened by reconsidering how we measure international conflict. Many assumptions and coding decisions lie behind the MID data, and some of these Page 251 → may influence the appropriateness of their use in answering certain questions. I have tried to show that one such question is that of trade and conflict. Although they clearly possess problems of their own and are by no means a panacea, events data can offer a effective source of data to test theories relating to interdependence, conflict, and cooperation.

NOTES The author would like to thank Edward D. Mansfield, Brian M. Pollins, and the participants in the Conference on Economic Interdependence and International Security held at the Mershon Center, Columbus, Ohio, for helpful comments. 1. For a more comprehensive review of this literature, see McMillan 1997 and Edward D. Mansfield and Brian M. Pollins's introduction to this volume. 2. COPDAB data has also been used in the trade and conflict debate as an independent variable. For example, see the seminal studies of Pollins (1989a, 1989b), which investigated conflict's role in shaping trade flows. 3. For such an overview see Jones, Bremer, and Singer 1996. 4. For a description of global-event data sets (e.g., WEIS or COPDAB) versus episodic-event data sets, and

the use of events data in general, see Schrodt 1994. Rafael Reuveny (this volume) also describes some of the more recent data sets and programs in the events data field. 5. In theory, one could treat the missing observations within the temporal overlap as zeros since presumably nothing happened between the states in question. At this point, I will not make that assumption, preferring to leave these observations as missing. 6. The decision concerning which events to map to which level of militarized dispute was made by consulting the coding criteria and description of each event/hostility level. 7. This figure assumes the use of the dispute variable rather than the onset variable used in later Oneal and Russett studies. 8. In a focused comparison between data concerning U.S. foreign policy, Fordham and Sarver (2001) find that the MID data undercounts U.S. uses of force in the 1870–1990 period. Again, this suggests that the MID data set may miss a number of important conflict events. 9. On the propensity for different media services to pick up various types of events, see Huxtable and Pevehouse 1996. 10. This would also explain why COPDAB picks up far more low-level dispute events than WEIS. WEIS relies only on New York Times reports, whereas COPDAB combines a wide variety of international media sources to code events. 11. This is equivalent to yearly aggregation of the data using their respective intensity-weighting scales.Page 252 → For COPDAB, Azar's (1993) original scale is used, and for WEIS, Goldstein's (1992) scale is used.

REFERENCES Azar, Edward E. 1993. Conflict and Peace Data Bank (COPDAB), 1948–1978. Study no. 7767. Ann Arbor, MI: Inter-university Consortium for Political and Social Research. Barbieri, Katherine. 1996. Economic Interdependence: A Path to Peace or a Source of Interstate Conflict? Journal of Peace Research 33:29–49. Bennett, D. Scott, and Allan C. Stam. 2000. Research Design and Estimator Choices in the Analysis of Interstate Dyads. Journal of Conflict Resolution 44:653–86. Blainey, Geoffrey. 1988. The Causes of War. 3d ed. New York: Free Press. Fordham, Benjamin O., and Christopher Sarver. 2001. Militarized Interstate Disputes and United States Use of Force. International Studies Quarterly 45:455–66. Gartzke, Erik. 1998. Kant We All Just Get Along? Opportunity, Willingness, and the Origins of the Democratic Peace. American Journal of Political Science 42:1–27. ______. 2000. Preferences and the Democratic Peace. International Studies Quarterly 44:191–212. Gasiorowski, Mark. 1986. Economic Interdependence and International Conflict: Some Cross-National Evidence. International Studies Quarterly 30:23–38. Gasiorowski, Mark, and Solomon W. Polachek. 1982. Conflict and Interdependence: East-West Trade and Linkages in the Era of Detente. Journal of Conflict Resolution 26:709–29. Gochman, Charles S., and Zeev Maoz. 1984. Militarized Interstate Disputes, 1816–1976: Procedures, Patterns, and Insights. Journal of Conflict Resolution 28:585–616. Goldstein, Joshua S. 1992. A Conflict-Cooperation Scale for WEIS International Events Data. Journal of Conflict Resolution 36:369–85.

Gowa, Joanne. 1999. Ballots and Bullets. Princeton: Princeton University Press. Huxtable, Phillip A., and Jon C. Pevehouse. 1996. Validity Problems in Events Data Collection: News Media Sources and Machine Coding Protocols. International Studies Notes 21:8–19. Irwin, Douglas A. 1996. Against the Tide: An Intellectual History of Free Trade. Princeton: Princeton University Press. Jones, Daniel M., Stuart A. Bremer, and J. David Singer. 1996. Militarized Interstate Disputes, 1816–1992: Rationale, Coding Rules, and Empirical Patterns. Conflict Management and Peace Science 15:163–213. Keohane, Robert O., and Joseph S. Nye. 2001. Power and Interdependence. 3d ed. New York: Addison-Wesley Longman. Leng, Russell. 1993. Interstate Crisis Behavior, 1816–1980: Realism versus Reciprocity. Cambridge: Cambridge University Press. Page 253 → Mansfield, Edward D., and Jon C. Pevehouse. 2000. Trade Blocs, Trade Flows, and International Conflict. International Organization 54:775–808. McClelland, Charles A. 1978. Warnings in the International Events Flow: EFI and Roz as Threat Indicators. International Interactions 5:135–204. McMillan, Susan M. 1997. Interdependence and Conflict. Mershon International Studies Review 41:33–58. Oneal, John R., Frances H. Oneal, Zeev Maoz, and Bruce M. Russett. 1996. The Liberal Peace: Interdependence, Democracy, and International Conflict, 1950–85. Journal of Peace Research 33:11–28. Oneal, John R., and Bruce M. Russett. 1997. The Classical Liberals Were Right: Democracy, Interdependence, and Conflict, 1950–85. International Studies Quarterly 41:267–94. ______. 1999. The Kantian Peace: The Pacific Benefits of Democracy, Interdependence, and International Organization, 1885–1992. World Politics 52:1–37. Organski, A. F. K., and Jacek Kugler. 1980. The War Ledger. Chicago: University of Chicago Press. Polachek, Solomon W. 1980. Conflict and Trade. Journal of Conflict Resolution 24:57–78. Pollins, Brian M. 1989a. Conflict, Cooperation, and Commerce: The Effect of International Political Interactions on Bilateral Trade Flows. American Journal of Political Science 33:737–61. ______. 1989b. Does Trade Still Follow the Flag? American Political Science Review 83:465–80. Reed, William. 2000. A Unified Statistical Model of Conflict Onset and Escalation. American Journal of Political Science 44:84–93. Reuveny, Rafael, and Hejoon Kang. 1996. International Trade, Political Conflict/Cooperation and Granger Causality. American Journal of Political Science 40:943–70. Russett, Bruce M., John R. Oneal, and David R. Davis. 1998. The Third Leg of the Kantian Tripod for Peace: International Organizations and Militarized Disputes, 1950– 1985. International Organization 52:441–67. Schrodt, Philip A. 1994. Event Data in Foreign Policy Analysis. In Foreign Policy Analysis: Continuity and Change in Its Second Generation, ed. L. Neack, J. Hey, and P. Haney, 145–66. Englewood Cliffs, NJ: Prentice Hall.

Viner, Jacob. 1951. Peace as an Economic Problem. In International Economics, ed. Jacob Viner, 247–67. Glencoe, IL.: Free Press. Waltz, Kenneth. 1970. The Myth of National Interdependence. In The International Corporation, ed. C. P. Kindleberger, 205–23. Cambridge, MA: MIT Press.

Page 254 →

Measuring Conflict and Cooperation: An Assessment Rafael Reuveny Political conflict and cooperation are frequently observed in international relations. In broad terms, conflict and cooperation cover a spectrum of activities. On the conflict side of the spectrum there are actions such as wars, blockades, sanctions, and threats. On the cooperation side there are peace accords, cease-fires, economic links, and verbal support. These actions interact with many social, political, and economic forces, including international trade. The nature of the relationship between international trade and political conflict/cooperation is controversial. The contemporary debate has been largely empirical, with many studies using statistical models. Statistical models require quantitative measures of conflict and cooperation. These measures are typically constructed by compiling reports taken from various publicly available news sources and historical texts. The empirical measurement of conflict and cooperation comes in two types. One focuses on the presence or absence of militarized interstate disputes (MIDs). In this approach, the cooperation side of the international relations spectrum is basically equated with the total absence of militarized conflict.1 A second approach considers the full spectrum of international relations, generating continuous measures of conflict and cooperation. The dichotomous measure of conflict is coded based on the information provided in the MID data set, a record of the occurrences of MIDs. The continuous measures of conflict and cooperation are constructed based on events data sets, or recorded daily interactions. Several such data sets have been collected. The evolution of the events data research program is relevant to this chapter. Briefly, the research program was begun in the early 1960s.2 By the early 1970s, the Conflict and Peace Data Bank (COPDAB) and the World Event Page 255 → Interaction Survey (WEIS) had been assembled under the leadership of Edward Azar and Charles McClelland, respectively. From the 1970s to the early 1990s, events data sets were used extensively in the literature on conflict and cooperation interactions (the action-reaction literature) and in trade and conflict studies.3 In the 1970s and early 1980s, events data also were used by the U.S. government for foreign policy analysis and crisis management (Laurance 1990). Despite their growing popularity during the 1970s and early 1980s, events data sets have faced criticisms. Consequently, their collection has declined (Laurance 1990; Merritt 1994). From the early 1980s to the early 1990s, the collection of large-scale events data focused on updating the WEIS data set. Since the mid-1990s, the overwhelming majority of statistical trade and conflict studies (and conflict studies in general) have used the MID data set.4 These studies have enhanced our understanding of conflict processes. However, the MID data set also has certain limitations that make it unsuitable for the empirical study of many conflict/cooperation–related topics, including the relationship between trade and conflict. I argue that it would be beneficial for the field of international relations to go back and routinely use events data sets. The argument is not meant to urge the discontinuation of the use of the MID data set but to illustrate a number of insights to be gained from using events data sets that cannot be investigated by using only the MID data set. In other words, the two approaches to the measurement of conflict/cooperation are complements rather than substitutes. The next section describes the basics of events data sets and the MID data set. The following two sections discuss the limitations and benefits of both data sets and evaluate their uses in the study of trade and conflict, respectively. The last section summarizes and suggests further research.

The Basics of Events Data and MID data

An event is an action involving an actor and a target. The actor is the unit generating the action, and the target is the unit toward which the action is sent. The description of an event involves an issue, the topic about which the activity revolves, and an activity—the action directed by the actor toward the target. The coding of events requires one to register the actions actors direct toward each other. Since the spectrum of international actions is broad, some categorization of actions according to predefined types is required. Two other questions to be answered are: (1) Which types of actions are to be observed? and (2) Who are the relevant actors whose actions need to be observed? One possibility Page 256 → is to observe all the actions taking place among all international actors, including governments, international organizations, and multinational corporations. The effort entailed by this approach is monumental. Another possibility is to simplify the inquiry by focusing on subsets of actions and actors. Events data sets deal with these issues in approximately the same way. The actions observed cover the full spectrum of international relations, and the actors whose actions are to be observed are nations, conceptualized as unitary actors. The information sources of events data sets are publicly available news sources and historical chronologies. The reports are classified according to predefined types of activities involving one pair of nations at a time: for example, events data sets may report that on date t 1, nation X warned nation Y, and on date t 2, X and Y signed some agreement. To illustrate, consider the popular COPDAB and WEIS events data sets, which are collected by trained coders. COPDAB covers events taking place from 1948 to 1978, whereas WEIS covers events from 1966 to 1992. The COPDAB data set is based on about seventy sources. The raw reports are separated into fifteen categories that differ in their degree of friendliness and hostility, ranging from war to a voluntary political unification. COPDAB is based on many sources in order to reduce the bias in data collection that could arise when using only a few sources (Azar 1984; Reuveny and Kang 1996a). The WEIS data set is based only on the New York Times. According to McClelland et al. (1971), the use of one source reduces the noise introduced when the same event is interpreted differently by several news sources. WEIS events are categorized according to sixty-one actions such as “force,” “threaten,” and “agree.”5 Several other smaller events data sets also are available. For example, the SHERFACS and Behavioral Correlates of War (BCOW) data sets focus on specific conflict episodes in greater detail than COPDAB or WEIS.6 WEIS and COPDAB have been extended by the Kansas Event Data System (KEDS) and the Global Event Data System (GEDS), respectively. The KEDS and the GEDS are data sets collected with the aid of a computer program designed to read news reports and code events. The scope of these two data sets currently is relatively small.7 While events data sets record events on a daily basis, many economic and political variables are available only annually or quarterly. To be used with annual or quarterly data in empirical analysis, the daily events data need to be aggregated over time. Typically, the events are first weighted to reflect the relative strength of actions. The weighted events are then aggregated over years or quarters, respectively. The weights are generated based on surveys that ask respondents to rank actions according to their relative level of hostility or Page 257 → friendliness. The weights most often used for the COPDAB data set have been developed by Azar and Havener (1976). Scholars have developed several weighting schemes for the WEIS data set. Goldstein (1992) summarizes the literature on weighting events data and provides another set of weights for the WEIS data set.8 Moving to the second approach to the measurement of conflict discussed in this chapter, the MID data set focuses on intense conflict episodes, referred to as militarized interstate disputes. Like the COPDAB data set, the MID data set also is coded based on many sources. However, unlike events data, MID data focus on militarized interstate disputes as the unit of analysis, and the cooperation side of the conflict-cooperation spectrum only can be observed (indirectly) as the total absence of militarized conflict.9 The first version of the MID data set included 965 MIDs taking place from 1816 to 1976, focusing on major powers (Gochman and Maoz 1984). The second version of the MID data set has extended the scope of the data set to include more than 2,000 MIDs taking place from 1816 to 1992 and has added several new variables describing each MID (Jones, Bremer, and Singer 1996).

The information provided on each MID has several attributes, including the nations involved in the dispute, the beginning and ending dates of the dispute, an indication of whether the dispute was reciprocated, the fatality level, and the outcome of the dispute. The actions taken in the dispute are reported using the dichotomous variables “hostility level of dispute” and “highest action in dispute.”10 The “hostility level of dispute” covers the action types “threat to use force,” “display of force,” “use of force,” and “war.” The “highest action in dispute” covers twenty-two action types, ranging from “threat to use force” to “joins interstate war.” The MID data set released in 1996 has been converted into a dyadic format by Zeev Maoz.11 The dyadic form of the MID data set is used extensively in contemporary trade and conflict studies as well as in other conflict studies. The information provided in the dyadic form of the MID data set is basically similar to the information provided in the dispute form of the data set, but it is presented along dyad-years. Currently a new version of the MID data set (MID3)—extending the coverage of the existing data set to include MIDs that took place between 1993 and 2001—is being coded.12

The Limitations of the MID Data Set and of Events Data In statistical analysis, it is important to understand the limitations of the data used and the constraints they impose on the modeling enterprise. This section Page 258 → first will discuss the limitations of the MID data set and then turn to a discussion of events data sets. First, the MID data set focuses on the conflict side of the conflictcooperation spectrum. However, dyads often engage in actions of conflict and cooperation in the same time (Pollins 1989a, 1989b).13 To deal with these cases, many studies aggregate weighted conflict and cooperation reports from events data sets to generate a single measure, referred to as net conflict.14 The MID data set does not allow the investigation of such situations. Researchers typically convert MID data into a yearly series of dyadic disputes due to the yearly nature of other indicators in their models. However, since MID episodes are rare events, following this transformation almost all the yearly entries for “hostility level of dispute” and “highest action in dispute” register “no militarized action” (i.e., no MID occurred that year). As noted by Oneal and Russett (2000), this makes the job of statistical analysis very hard because the input data seldom change. Partly because of this “problem,” MID studies typically set their yearly conflict measure to 1 whenever a MID of any type is reported and to 0 otherwise. Oneal and Russett also note that the rank ordering of the “hostility level of dispute” and “highest action in dispute” variables is considered questionable, providing yet another reason to ignore it.15 But Jones, Bremer, and Singer (1996) note that no effort has been made by the coders of the MID data set to rank the intensity of the militarized incidents that fall within the categories of the “highest action in dispute.” Regardless of the reasons for the use of a 1–0 indicator of MID, warnings to use military force, for example, are quite different from wars. In other words, the use of a 1–0 indicator of MID introduces aggregation bias (James, Solberg, and Wolfson 2000).16 Third, the MID data set is not well positioned to inform researchers about conflict intensity. However, intensities are important aspects of scientific inquiries. For example, one question that cannot be fully addressed using the MID data set is how the relationship between conflict and democracy holds up with respect to cooperation—or less severe conflict than a MID (Polachek and Robst 1998). Moreover, empirical measures that exhibit both low and high levels of intensity increase the ability of statistical analysis to detect patterns in the data, which provides yet another reason to use events data sets. Fourth, the use of a dichotomous MID measure complicates statistical analysis. We generally know more about estimating models from continuous dependent variables than from dichotomous dependent variables.17 There also is the issue of simultaneity. In our context, many scholars argue that trade and conflict affect each other. However, the estimation of simultaneous equation models involving dichotomous dependent variables is not well understood. Another example involves the debate on using MID-onset versus MID- Page 259 → involvement as the dependent variable.18 This issue is more complicated when using a measure that “suddenly” jumps from 0 to 1, because in reality political relations typically do not “jump,” but change gradually. With a continuous measure of conflict, the question of how trade affects the onset of MIDs is naturally dealt with by checking the extent to which trade can maintain the level of conflict below some threshold of hostility.

In general, the MID data set is not well suited to investigating issues pertaining to conflict and cooperation dynamics. For example, many studies have demonstrated the importance of action-reaction conflict dynamics, where conflict depends on reciprocity and inertia.19 The MID data set is not appropriate for the investigation of the action-reaction model because it only provides “a highly condensed summary of the evolution of the dispute and not an accurate description of the complete interaction process” (Jones, Bremer, and Singer 1996, 173). Finally, statistical models estimated based on the MID data set generate statements on the likelihood of conflict, as opposed to the level of conflict. For example, some MID models predict that when the level of bilateral trade rises by X units, the probability of MID falls by Y percent. Probabilistic statements of this type are informative, but policymakers may well want some idea of what the level of conflict will be when trade rises by X units. While events data are not subject to these problems, they also have limitations.20 First, they are noisy. Some scholars attribute the noise to coder error, arguing that the coding depends on news reporters whose sources and understanding of the issues are not known. The interpretations of the coder and the reporter also are factors. In principle, when a news report is not clear, the coder could interview the reporter. However, this typically is not done. Scholars also argue that events data could be biased because some of the actions between nations are exchanged in secret and not reported in open sources, and news reports may reflect their paper's worldview. The problems of noise and bias also apply to the MID data set because it is also coded manually based on news reports. However, MID data involve events on the high end of the conflict scale that are easier to discern than, say, the intricacies of cooperation. Second, scholars argue that some international interactions are episode related. However, once events are coded, the context in which they took place is lost. The user of events data is provided with a “black box” whose accuracy cannot be checked against the sources upon which data have been coded. This point does not seem to apply to the MID data set because it focuses on episodes. However, most studies that use the MID data set also do not dwell on the details of each of the MIDs included in their large-N samples. It also is possible Page 260 → to argue that the MID data set provides fewer details than events data sets, because it does not specify the sequence of events leading to each MID episode. Another argument concerns the number of sources used. When events data sets are based on many sources, there could be differences across sources, resulting in measurement noise. However, relying on one source could bias the data, reflecting the worldview of that source, and could result in longer periods without reports because some sources tend to focus on particular geographical regions. MID data-set coding also relies on many sources and thus is subject to a similar problem of measurement noise. Fourth, events data sets assume that states are unitary actors. However, international interactions may involve nonstate actors (e.g., the Organization of Petroleum Exporting Countries [OPEC], the Palestinians) and domestic groups (e.g., the army, businesses), neither of which are reported in events data sets. Obviously, this shortcoming also applies to the MID data set, as it also assumes that states are unitary actors. Finally, events data sets use different sources and classifications of events. Whether these data sets provide similar information pertaining to the same event is discussed in the literature. In particular, some scholars argue that the COPDAB and WEIS data sets are not compatible, which suggests that the decisions of coders are to some extent subjective. Other studies argue that the two data sets are generally compatible, a point to which I will return. This argument obviously does not apply to the MID data set, since there is only one such data set in the field.

Evaluating the Limitations and Discussing the Benefits Empirical measures typically have limitations, particularly when the measured concepts are complicated, as is the case here. Scholars using the MID data set, as well as those using events data sets, have made important contributions. That said, it is important to understand the implications of the limitations as well as the benefits of the two approaches. The MID data set has several advantages. First, the likelihood of errors in the data is relatively low because MID

episodes are salient events. The saliency streamlines the process of making sure that these episodes are reported correctly. The MID data set also is superbly collected and closely managed by leading scholars in the field. If coding errors are found in the data, they are quickly corrected. A second benefit of using the MID data set has to do with its long time span, beginning in 1816. Events data sets are only available from 1948. In practice, the Page 261 → long time span of the MID data set may not be important for the study of trade and conflict, because the time span of most studies is 1950 to 1992 (due to the availability of bilateral trade data and other economic variables). Considering the centrality of militarized disputes in the field of international relations, the focus on MIDs is another benefit offered by the MID data set. It also is possible to argue that while cooperation and relatively lowlevel conflict may occur at the same time, when countries fight the level of cooperation is likely to be low. Hence, the omission of cooperation from a story focusing on intense conflict may not introduce a significant error. The MID data set is relatively simple. The amount of MID data to sift through is relatively small, as MID episodes are rare. Almost all studies in the MID tradition simplify the picture even further by coding an annual on-off measure of dyadic militarized conflict. In contrast, using events data sets requires decisions on such issues as the temporal unit of aggregation (as these are daily data), the treatment of days without reports, the treatment of days with more than one report, the weighting scheme used to aggregate events, and the construction of measures for empirical analysis (e.g., net conflict, separate conflict and cooperation measures). The simplicity of the MID data set comes with a price. In particular, conflict dynamics and cooperation activities are not well represented in the MID data set. Events data sets are more suitable for studying detailed dynamic conflict/cooperation mechanisms, because their data are reported daily, record the direction of actions, and report both conflictual and cooperative actions. As noted, events data sets also have limitations. In summary, scholars argue that events data are noisy, that they omit covert actions, that their unitary-state-actor assumption is restrictive, that they do not provide episode-related details, and that different data sets may not be compatible when observing the same event. However, noise does not bias the data (Goldstein and Freeman 1990). To the extent that studies find statistically significant results, noise is not a problem. Noisy data is one of the reasons statistical analysis is used. The possible omission of covert actions from public sources is an issue about which little can be done. Still, to the extent that countries' public moves eventually agree with their secret moves, this omission generates only noise, not bias. In general, the issues of noise and bias seem less problematic in the MID data set because MID episodes are relatively salient. The unitary-actor assumption employed by events data sets is a simplification. However, this assumption is common in international relations and also is shared by the MID data set. Next, the level of detail provided by events data may not be high enough for certain analyses, but it is adequate for Page 262 → many models. If one requires a higher level of detail on some episode than that provided in events data sets, one might as well employ a case study. Some scholars also argue that dyads are affected by other dyads, which is not well reflected in events data sets. While this argument is valid, studies using events data could include the effects of other dyads in the modeling stage (e.g., Kang and Reuveny 2001). In any case, these arguments also apply to the MID data set. Finally, there is the issue of compatibility of different events data sets. Reuveny and Kang (1996a) find that COPDAB and WEIS reports of the same event are mostly compatible and conclude that the two data sets could be spliced, but this issue is subject to debate. Taking a broader view, having more than one events data set could be considered an asset. If the results obtained based on different events data sets match, they are more likely to be correct and robust. In any case, when events data sets do not agree with each other, the sources of the incompatibility can be investigated, a topic discussed in the last section.

Trade and Conflict Research: Events Data or MIDs? Both the MID data set and events data sets have been used in trade and conflict studies, although currently the MID data set is much more widely used than events data sets. This section evaluates the use of these data sets in

the study of trade and conflict. To begin with, events data sets and the MID data set imply a different conceptualization of the trade and conflict nexus. Studies that employ events data assume that the political conflict-cooperation spectrum is a continuum of actions and that all these actions interact with trade. In the MID data set, cooperation can only be observed as the total absence of conflict. Hence, studies that use the MID data set assume (implicitly or explicitly) that cooperation and no conflict are the same concept and that all actions of cooperation interact with trade in the same manner regardless of their type or intensity. The MID data set divides conflict into categories, as reflected by the entries of the variables “hostility level of dispute” and “highest action in dispute.” However, as noted, the overwhelming majority of studies focus on the presence or absence of an MID of any type. This implies that joining wars and threats, for example, are assumed (implicitly or explicitly) to interact with trade in the same manner. This assumption may be correct, but it needs to be tested. Of course, testing this assumption based on the MID data set is not easy due to the paucity of the different MID subcategories. It could be done by using events Page 263 → data, since the different types of MIDs basically imply different intensities of conflict, which can be measured using events data. Another problem with the on-off approach to conflict is noted by Penubarti and Ward (2000). MID studies typically code the presence of a MID as 1 for the entire year, regardless of its timing. If a MID occurs on December 31, it is still coded as 1 for that year. It follows that this MID may have no detectable effect on trade, as the trade volume reported for this year has occurred before the MID. This problem is not unique to the MID data set, but it is exacerbated by it. With events data, one could observe a deterioration in political relations leading to a dispute, as countries do not simply “jump” into a MID. This deterioration could be reflected in traders' calculations. Because annual aggregates of events data reflect the intensity of relations throughout the year, they are better positioned to measure the effect of a MID on annual trade. Moreover, with events data, one could choose to generate quarterly measures of conflict/cooperation, which further helps to alleviate this problem. Taking a broader view, the choice of using MID data or events data in trade and conflict studies boils down to the question of what type of conflict is to be investigated. The MID data set offers a restricted depiction of cooperation and a more detailed portrayal of conflict. Consequently, the MID data set does not allow full evaluation of the intricacies involved in the basic international relations ideologies that drive the trade and conflict debate, that is, liberalism, realism, and neo-Marxism. Realists expect that trade will not affect intense international conflict or cooperation, linking political relations to considerations of national security. Liberals expect that trade will suppress all levels of conflict and promote all levels of cooperation, regardless of their intensity. Neo-Marxists expect that trade will intensify all levels of conflict and reduce all levels of cooperation. Proponents of the position that conflict reduces trade and cooperation raises trade also do not restrict their arguments to certain levels of conflict or cooperation. From any perspective, the intensity of conflict/cooperation is an important issue. Therefore, it seems that events data are better suited to the task of fully evaluating contemporary trade and conflict theories, as these data encompass all levels of conflict and cooperation.

Moving Forward: Some Possibilities Using the MID data set, the field of international relations has generated important insights. However, there also are issues that the MID data set has limited ability to address, including trade and conflict dynamics, Page 264 → conflict/cooperation interactions, the role of cooperation, and the relationship between trade and the levels of conflict and cooperation (as opposed to the probability of a MID). Such issues can be more readily investigated from events data. The two approaches need to be viewed as complements, rather than substitutes. That said, a number of research extensions could prove beneficial. There is a need to construct events data sets to cover periods before 1948. While this is a monumental task, many of the news sources used by current events data sets are available for periods before 1948. Thus, it is a matter of finding funding for a project of this magnitude. The National Science Foundation has recently funded the extension of the MID data set to cover the 1993–2001 period; perhaps the next step would be to extend events data sets to include periods before 1948.

Some of the problems associated with the collection of events data can be alleviated by using computer-based data collection. For example, one could conceivably adapt the KEDS and GEDS computer programs to extend the scope of current events data sets. As a first step, KEDS and GEDS could be used to generate data sets encompassing all countries from 1948 to the present, enabling the replication of analyses currently performed based on the MID data set. Once these data sets are in place, several cross-check analyses are required. First, KEDS and WEIS, and GEDS and COPDAB, could be compared during their overlapping years.21 Second, KEDS and GEDS could be compared to one another. Third, KEDS and GEDS could be compared to the MID data set. The more these data sets match, the greater their reliability. Since MID episodes are easier to discern than some of the actions in events data sets, where there are discrepancies between the GEDS/KEDS and the MID data sets, the GEDS/KEDS data sets may need to be changed to match the MID data set. MIDs would probably be reflected at the higher conflict end of events data sets. However, this question has not been evaluated yet (Mansfield and Pollins 2001). An extension of this analysis could be to examine whether countries that engage in MIDs also cooperate over some issues. It also is important to evaluate whether the trade and conflict nexus is affected by the overall political relations as opposed to MIDs. Another possibility would be to use events data sets to better understand the circumstances of each MID. For example, events data sets may better allow the investigation of the escalation into a MID or the resolution of a MID. Finally, events data sets were used by policymakers in the 1970s and early 1980s. The use of events data sets by policymakers has declined since then, partly due to the controversies surrounding the data and partly because user friendly Page 265 → software tools were not well developed at that time. The large effort required to collect events data also has been an issue. I believe that the utility of the MID data set for real-time policy-making is not considerable. Of course, the insights academics gain from using the MID data set are potentially important for policymakers, but the MID data set cannot be used in real time for early warning or analysis of recent events. In contrast, daily events data collected in, or close to, real time have proven to be useful for policymakers, and these data sets could be further modified to increase their utility. In conclusion, I believe that it is in the best interest of the field of international relations to return to the use of events data sets on a large-scale basis—not necessarily as a replacement of the MID data set but as an addition to it. Despite the criticisms that have been levied against using events data, some of which I feel are exaggerated, there are many useful aspects of these data that make them a source of beneficial insights for both academics and policymakers.

NOTES 1. As noted by Jones et al. (1996), there is no intention to code cooperative acts within the MID tradition; rather, the intention is to focus on hostile interactions between states. 2. For details, see Laurance 1990 and Merritt 1994. 3. For action-reaction studies, see Ward 1981; Dixon 1986; McGinnis and Williams 1988; Goldstein and Freeman 1990; and Ward and Rajmaira 1992. For trade and conflict studies, see Polachek 1980; Gasiorowski and Polachek 1982; Gasiorowski 1986; Sayrs 1989; Pollins 1989a, 1989b; and de Vries 1990. 4. Contemporary trade and conflict studies using events data sets include Reuveny and Kang 1996b, 1998; Polachek 1997; Polachek and Robst 1998; and Kang and Reuveny 2001. 5. On COPDAB, see Azar 1984 and Azar and Havener 1976. On WEIS, see World Event Interaction Survey 1993 and McClelland et al. 1971. 6. On SHERFACS, see Sherman 1994; on BCOW, see Leng 1993. SHERFACS stands for the names Frank Sherman, Lee Farris, Hayward R. Alker Jr, and Kathleen Carley. 7. GEDS uses reports of news wires such as Reuters and TASS (Telegrafnoe Agentstvo Sovetskogo Soiuza [Telegraph Agency of the Soviet Union]), and KEDS employs only Reuters reports. On KEDS, see and Schrodt, Davis, and Weddle 1994. GEDS is described at and in Davis 1994. 8. Gasiorowski (1986) emphasizes the importance of using weights, but not all scholars endorse aggregation of events using weights. For example, see McClelland 1983. Page 266 →

9. On the MID data set, see and Jones, Bremer, and Singer 1996. 10. The names of these variables are those listed at . 11. For details, see Maoz 1999. 12. The new data set also includes a few changes of action types for “the highest action in dispute.” For details, see . 13. For example, often during the Cold War the United States and the USSR cooperated over some issues while having intense conflict over others. We might also consider U.S.–China relations today. 14. For example, see Richardson 1960; Polachek 1980; Gasiorowski 1986; Pollins 1989a, 1989b; Goldstein and Freeman 1990; Reuveny and Kang 1996b, 1998; and Kang and Reuveny 2001. 15. Not all studies employ this practice. Some studies employ the “hostility level of dispute” as an ordinal measure (e.g., James, Solberg, and Wolfson 1999). Diehl and Goertz (2000) devise a method to convert the “hostility level of dispute” into a scale ranging from 0 to 214, computed based on the commutative frequency of MID actions in the population and the number of fatalities in the MID. 16. The popular net-conflict measure constructed from events data by adding weighted conflict and cooperation events in some period also is an aggregate. However, it is affected by all the reported actions in the period, as opposed to treating all the MID types as if they are the same phenomena. 17. For example, consider the “never-ending” debate on the proper methods to apply in the estimation of pooled time-series cross-section logit models. 18. See, for example, Mansfield and Pevehouse 2000 and Oneal and Russett 1998. 19. Reciprocity is the tendency of nations to play “tit for tat,” and inertia is the tendency of variables to depend on their own previous values. Summarizing, Ward (1981, 230) posits a golden rule that guides nations: “Do unto others what they have recently done unto you.” For a recent example of this approach see Goldstein and Pevehouse 1997. 20. For more extensive discussion, see Laurance 1990 and Davis 1994. 21. To my knowledge, some analysis of this type has already been done.

REFERENCES Azar, E. E. 1984. The Conflict and Peace Data Bank, 1948–1978. Study no. 7757. 2d ed. Ann Arbor, MI: Inter-university Consortium for Political and Social Research. Azar, E. E., and T. Havener. 1976. Discontinuities in the Symbol Environment: A Problem in Scaling. International Interactions 2:231–46. Davis, J. L. 1994. A New Generation of International Event-Data. International Interactions 20:55–78. Page 267 → de Vries, M. 1990. Interdependence, Cooperation and Conflict: An Empirical Analysis. Journal of Peace Research 27:429–44. Diehl, P. F., and G. Goertz. 2000. War and Peace in International Rivalry. Ann Arbor: University of Michigan Press. Dixon, W. J. 1986. Reciprocity in United States–Soviet Relations: Multiple Symmetry or Issue Linkage? American Journal of Political Science 30:421–44. Gasiorowski, M. J. 1986. Economic Interdependence and International Conflict: Some Cross-National Evidence. International Studies Quarterly 30:23–38. Gasiorowski, M. J., and S. W. Polachek 1982. Conflict and Interdependence: East-West Trade and Linkages in the Era of Detente. Journal of Conflict Resolution 26:709–29.

Gochman, C. S., and Z. Maoz. 1984. Militarized Interstate Disputes, 1816–1976: Procedures, Patterns and Insights. Journal of Conflict Resolution 28:585–615. Goldstein, S. J. 1992. A Conflict Cooperation Scale for WEIS Events Data. Journal of Conflict Resolution 36:369–85. Goldstein, S. J., and J. R. Freeman. 1990. Three Way Street: Strategic Reciprocity in World Politics. Chicago: University of Chicago Press. Goldstein, S. J., and J. C. Pevehouse. 1997. Reciprocity, Bullying, and International Cooperation: Time Series Analysis of the Bosnia Conflict. American Political Science Review 91:515–29. James, P., E. Solberg, and M. Wolfson. 1999. An Identified Systemic Model of the Democracy-Peace Nexus. Defence and Peace Economics 10:1–37. ______. 2000. Democracy and Peace: A Reply to Oneal and Russett. Defense and Peace Economics 11:215–29. Jones, D. M., S. A. Bremer, and D. J. Singer. 1996. Militarized Interstate Disputes, 1816–1992: Rational Coding Rules, and Empirical Patterns. Conflict Management and Peace Science 15:163–213. Kang, H., and R. Reuveny. 2001. Exploring Multi-Country Dynamic Interrelationships between Trade and Conflict. Defence and Peace Economics 12:175–96. Laurance, E. J. 1990. Events Data and Policy Analysis: Improving the Potential for Applying Academic Research for Foreign Defense Policy Problems. Policy Sciences 23:111–32. Leng, R. J. 1993. Automated and Machine-Assisted Coding of Events Data: The BCOW Approach. In International Event-Data Developments: DDIR Phase II, ed. R. L. Merritt, R. G. Mancuster, and D. A. Zinnes, 113–23. Ann Arbor: University of Michigan Press. Mansfield, E. D., and J. Pevehouse. 2000. Trade Blocs, Trade Flows, and International Conflict. International Organization 54:775–808. Mansfield, E. D., and B. M. Pollins. 2001. The Study of Interdependence and Conflict: Recent Advances, Open Questions, and Directions for Future Research. Journal of Conflict Resolution 45:834–59. Maoz, Z. 1999. Dyadic Militarized Interstate Disputes (DYMID1.1) Dataset. Tel Aviv: Tel Aviv University. Page 268 → McClelland, C. A. 1983. Let the User Beware. International Studies Quarterly 27:169–77. McClelland, C. A., R. G. Tomlinson, R. G. Sherwin, G. A. Hill, H. L. Calhun, P. H. Fenn, and J. D. Martin. 1971. The Management and Analysis of International Event Data: A Computerized System for Monitoring and Projecting Event Flows. Los Angeles: University of Southern California. McGinnis, M. D., and J. T. Williams. 1988. Change and Stability in Superpower Rivalry. American Political Science Review 83:1101–23. Merritt, R. L. 1994. Measuring Events for International Political Analysis. International Interactions 20:3–33. Oneal, J. R., and B. Russett. 1998. Assessing the Liberal Peace with All Dyads: 1950–1992. Department of Political Science, University of Alabama. ______. 2000. Comment: Why “an Identified Systemic Model of the Democratic-Peace Nexus” Does Not Persuade. Defence and Peace Economics 11:197–214.

Penubarti, M., and M. D. Ward. 2000. Commerce and Democracy. Working paper 6, Center for Statistics and the Social Sciences, University of Washington. Polachek, W. S. 1980. Conflict and Trade. Journal of Conflict Resolution 24:55–78. ______. 1997. Why Democracies Cooperate More and Fight Less: The Relationship between International Trade and Cooperation. Review of International Cooperation 5:295–309. Polachek W. S., and J. Robst. 1998. Cooperation and Conflict among Democracies: Why Do Democracies Cooperate More and Fight Less? In The Political Economy of War and Peace, ed. M. Wolfson, 127–54. Boston: Kluwer Academic Publishers. Pollins, B. M. 1989a. Conflict, Cooperation, and Commerce: The Effect of International Political Interactions on Bilateral Trade Flows. American Journal of Political Science 33:737–61. ______. 1989b. Does Trade Still Follow the Flag? American Political Science Review 83:465–80. Reuveny, R., and H. Kang. 1996a. International Conflict and Cooperation: Splicing COPDAB and WEIS Series. International Studies Quarterly 40:281–306. ______. 1996b. International Trade, Political Conflict/Cooperation and Granger Causality. American Journal of Political Science 40:281–306. ______. 1998. Bilateral Trade and Political Conflict/Cooperation: Do Goods Matter? Journal of Peace Research 35:581–602. Richardson, L. F. 1960. Arms and Insecurity. Pittsburgh: Boxwood Press. Sayrs, L. 1989. Trade and Conflict Revisited: Do Politics Matter? International Interaction 14:155–75. Schrodt, P. A., S. G. Davis, and J. L. Weddle. 1994. Political Science: KEDS, a Program for the Machine Coding of Event Data. Social Science Computer Review 12:561–88. Sherman, F. L. 1994. SHERFACS: A Cross-Paradigm, Hierarchical and Contextually Sensitive Conflict Management Data Set. International Interactions 20:79–100. Ward, M. D. 1981. Seasonality, Reaction, Expectation, Adaptation, and Memory in Page 269 → Cooperative and Conflictual Foreign Policy Behavior. International Interactions 8:229–45. Ward, M. D., and S. Rajmaira. 1992. Reciprocity and Norms in U.S.–Soviet Foreign Policy. Journal of Conflict Resolution 36:342–68. World Event Interaction Survey (WEIS). 1993. Coding Manual. 6th rev. Rev. Rodney G. Tomlinson. Department of Political Science, U.S. Naval Academy. Page 270 →

Page 271 →

Methodological Advances Page 272 →

Page 273 →

Temporal Dynamics and Heterogeneity in the Quantitative Study of International Conflict Janet M. Box-Steffensmeier, Dan Reiter, and Christopher J. Zorn Our discussion focuses on the importance of temporal dynamics in the study of international conflict and more specifically on the nature of those dynamics as they relate to interdependence. Our goal is to point out some areas in which the substantial achievements of international politics scholars may be further improved. In part because of the unique challenges posed by data on international relations, students of international politics have been among those at the forefront of developing innovative approaches to the study of political phenomena. For example, scholars studying international conflict have, in recent years, begun to go beyond the mere incidence of international disputes to study the process by which conflicts occur, including the stages through which nations typically pass as crises escalate to war (e.g., Reed 2000b; Reed and Clark 2000; Reed and Lemke 2001; Schultz 2000, 2001; Signorino 1999; Smith 1996, 1999) or fail to escalate to war (Gartzke this volume; Mansfield this volume). Such studies highlight a central characteristic of international conflict: the importance of time and temporal dynamics in the development of such disputes. These studies have been accompanied by ever more sophisticated analyses of the causes of conflict, many of which explicitly model the influence of those causes over time (e.g., Beck, Katz, and Tucker 1998; Beck 1999; Bennett and Stam 2000; Maoz and Russett 1993; Oneal and Russett 1999a, 1999b). We emphasize the importance of paying greater attention to the connection between a number of general theoretical expectations regarding time and Page 274 → conflict and the kinds of statistical tools used to model and test those expectations. Neither model assumptions (e.g., that of temporal parameter stability) nor basic political facts about the phenomenon of study (e.g., the repeated nature of conflict in the international arena) are innocuous. Different methodological treatments of repeated disputes, for example, imply fundamentally different understandings of the effects of those disputes on current conflicts. We concentrate on three main areas ripe for development: investigating temporal variation in the influence of covariates on conflict, evaluating the importance of repeated conflicts between nations over time, and incorporating corrections for unobserved heterogeneity (both temporal and otherwise) in international-conflict data.

Temporal Change in the Causes of Conflict Scholarship on the connection between interdependence and conflict has in recent years become more sophisticated. Earlier interdependence-conflict studies essentially ignored temporal dimensions, including research that assumed normally distributed (Polachek 1980; Gasiorowski 1986; Mansfield 1994) and categorical dependent variables (Maoz and Russett 1993; Barbieri 1996; Oneal and Russett 1997). In the late 1990s, studies of the interdependence-conflict connection and international relations more generally began to apply more sophisticated models that accounted for temporal dynamics. Notably, nearly all have adopted duration models as the primary tools of analysis. The most common of these is the Weibull model (e.g., Bennett 1997, 1998; Werner 1999), though both Cox models (e.g., Raknerud and Hegre 1997) and discrete-time approaches (e.g., Beck, Katz, and Tucker 1998; Crescenzi and Enterline 2001) have also been considered. Most of the recent interdependenceconflict scholarship has used the discrete-time approach (Beck, Katz, and Tucker 1998; Crescenzi and Enterline 2001; Gartzke, Li, and Boehmer 2001; Bennett and Stam 2000; Mansfield and Pevehouse 2000), though other methods accounting for temporal effects have been employed, including generalized estimating equations (GEEs) (Oneal and Russett 1999a, 1999b), Cox models (Hegre 2000), Granger causality models (Reuveny and Kang 1996), and fixed-effects models (Green, Kim, and Yoon 2001). A signature characteristic of most of these studies is the assumption that the impact of causal factors on the occurrence of a dispute remains constant over time. In the context of nearly all widely used duration models, this is referred to as the assumption of proportional hazards, so-called because the effect of a unit change in a covariate is to shift the hazard rate by a factor of proportionality. Page 275 → The result is that the marginal effect

of a covariate remains constant across time: a one-unit shift in, say, the balance of power between two nations affects the hazard of a dispute between them by the same (proportional) amount, irrespective of whether that change occurs in the first or the fortieth (or four hundredth) year of peace between them. For some influences on international disputes, the assumption of proportional hazards is likely to be a reasonable one. For others, however, international relations theories suggest a need to relax this requirement and allow the influence of factors known to affect the onset of international crises to vary over time. This is the case, for example, when the influence of some factor on the likelihood of conflict is theorized to wax or wane over time. Here, we briefly describe how theories about interdependence and conflict suggest that the relationship between them may vary across time and outline some methodological innovations to allow and test for that proposition. The idea that higher levels of trade and economic interdependence between states reduce the likelihood of conflict between them is one of the oldest in the study of international conflict. This eighteenth-century Enlightenment proposition has very recently received substantial empirical attention from international relations scholars. Oneal and Russett (1997, 1999a, 1999b; see also Gartzke, Li, and Boehmer 2001; Mansfield and Pevehouse 2000) exemplify this position, finding a statistically significant negative relationship between higher levels of trade and the likelihood of conflict. Critics have argued, on theoretical and methodological grounds, that there is no such relationship (Barbieri 1996; Beck, Katz, and Tucker 1998; Green, Kim, and Yoon 2001). The logic of the tradeconflict argument is that higher levels of economic interaction between two states provide subnational groups a stake in continued trade between the two countries; conversely, war is expected to substantially disrupt or even sever trade relations. As a result, these groups have a strong incentive to pressure the national government to maintain peace with trading partners (Papayoanou 1999; see also Gowa 1999). A logical extension emerges from the mitigating effects of economic ties on international conflict: that these effects should become more substantial as time passes. The passage of time permits the deepening and institutionalization of economic ties, as nations take a variety of actions to strengthen the interdependence between them; nations establish economic consulates, increase cultural and educational exchanges, and expand the scope of contact to new parts of the two economies, among other things (e.g., Mansfield this volume). Moreover, by lowering information costs and reducing the incentive to renege on contracts, longer-term relationships permit more efficient transfers of goods Page 276 → and services, a situation beneficial to both sides. Thus, as time passes, increasing interdependence provides more groups in a society with greater incentives to maintain peace (Keohane and Nye 1989). It is not simply higher levels of economic interdependence between states that mitigate conflict but also longer histories of economic interdependence. Put differently, the theory suggests that equal levels of trade will have a greater pacifying impact later in the relationship between two nations than earlier.1 The example outlined here is but one of many potential examples where theories of international politics predict significant temporal variability in the effects of those factors important in international disputes (see also BoxSteffensmeier, Reiter, and Zorn 2002). Testing such a theory, however, requires a model that allows the influence of economic interdependence on conflict to vary over the history of the dyad. Such considerations are important for two reasons. First, as has been widely shown in the statistics literature, estimation of proportional-hazards models when the covariate effects are nonproportional may result in biased estimates, incorrect standard errors, and faulty inferences about the substantive impact of independent variables (e.g., Kalbfleisch and Page 277 → Prentice 1980; Schemper 1992; Klein and Moeschberger 1997). Thus, it is important that researchers examine the extent to which the assumption of temporally constant covariate effects applies. Second, only models that permit covariate effects to change over time will allow for appropriate tests of these theories; nearly all such models in current use are not capable, without modification, of evaluating whether and how the factors influencing conflict vary over time. There is a wide range of methods for investigating nonproportional covariate effects in hazard-rate models (see, generally, Box-Steffensmeier and Zorn 2001). Statisticians have developed an array of both graphical and statistical tests for assessing the presence and nature of nonproportionality (e.g., Schemper 1992; Grambsch and Therneau 1994; Ng'andu 1997), nearly all of which are easily implemented in commonly used software for

estimating duration models (such as Stata™). As a result, the potential for wide adoption of such methods is great, provided that scholars are aware of the possible gains from such approaches and the risks of ignoring temporal dynamics.

Not Like Last Time: The Dynamics of Repeated Conflict Do conflicts beget future conflicts? Existing work on international conflict largely neglects the issue of past conflicts; such studies implicitly treat the first and the tenth disputes between two countries as exactly the same. These approaches thus run counter to much of what we know about international politics: that repeated conflicts often occur due to enduring rivalries (e.g., Bennett 1998; Diehl 1998; Goertz and Diehl 2000), that nations learn from previous conflicts (e.g., Reiter 1996), and that the information so gained is incorporated into future security decisions (e.g., Waltz 1979). Moreover, a number of competing theoretical perspectives on the nature of international conflict have real and potentially important implications for our expectations about repeated conflicts. On one hand, studies of the “security dilemma” (e.g., Jervis 1978) point out that a state's planning for war (e.g., by building up its military capabilities) can be interpreted as a sign of hostility by other nations, which in turn endeavor to cultivate their own capabilities. A related view holds that nations often go to war when the opposing state is perceived as hostile (e.g., Jervis 1976; Lebow 1981; Larson 1997); these hostile images of the opponent are exacerbated by conflicts, which serve to reinforce those images. Taken together, these perspectives suggest that, by feeding fears over security and fostering negative views of the enemy, the occurrence of an international conflict makes future conflicts Page 278 → that much more likely. On the other hand, rationalist perspectives on international conflict argue that “war is in the error term” (Gartzke 1999): that two states go to war when their perceptions about their relative capabilities diverge (Fearon 1995). To the extent that war itself reveals information about capabilities, then, it should have the effect of making future conflicts less likely. Thus, under this view, conflicts are self-restraining phenomena: because of what it reveals about the two states' power relationship, the occurrence of a dispute makes immediate future disputes less likely. Accounting for repeated events is likely to be especially important in the study of the interdependence-conflict relationship. The principal internal logic of the claim is that higher levels of trade make war less likely because of the expectation that war will diminish trade levels. If two nations fight and experience a corresponding disruption of trade, then the lesson that war prevents trade will be driven home, making the reoccurrence of war even less likely. If, however, war does not interrupt trade (Barbieri and Levy 1999), then states will reduce their estimates of the costs of war and fighting wars may make future wars more (or at least no less) likely. Additionally, if uncertainty is relatively low between trading states (Reed 2000a), then when trading states do fight it will likely be for reasons other than uncertainty, such as the indivisibility of issues or the inability of the two sides to commit to a settlement (Fearon 1995). Under these conditions, war is not caused by uncertainty, and therefore the reduction of uncertainty produced by war does not make future war less likely. While these competing views offer divergent predictions about the influence of previous disputes on future conflicts, the vast majority of extant analyses pay little attention to such considerations.2 By adopting approaches that ignore the influence of past events, these studies fail to address researchers' substantive interests in the effects of multiple disputes; moreover, and as a result, they likely also provide incorrect inferences about the estimated effects of covariates on those disputes. This omission is surprising, given both the centrality of concerns about enduring rivalries in theoretical discussions and the fact that the general issue of multiple events is one that has come to occupy center stage in statistical duration analysis research. Researchers have derived a host of methods, generally referred to as variance-corrected approaches, for making valid inferences about repeated events when those events are dependent and for assessing the impact of previous events on future ones (e.g., Andersen and Gill 1982; Prentice, Williams, and Peterson 1981; Wei, Lin, and Weissfeld 1989; for recent reviews, see Box-Steffensmeier and Zorn 2002; Kelly and Lim 2000). These approaches vary widely Page 279 → in their foundations: for example, some impose a sequential ordering on events, such that a subject is only “at risk” for a second event after a first event has occurred. Others impose no

such restriction, allowing events to develop in “parallel” fashion. The choice of which such model is appropriate for the subject at hand depends critically on the data-generating process in question. Similarly, some such methods (including all those used thus far in studies of international conflict) restrict the baseline hazard of the event in question to be constant over repeated events; in such instances, the conditional odds of a conflict are forced to be constant across disputes, whether that dispute is the first or the tenth between the states. By contrast, other methods allow for a relaxation of this restriction, thus permitting the researcher to estimate whether and to what extent the conditional odds of a dispute change over time. Such approaches are invaluable, for they provide a clear, easily interpretable way of assessing the various hypotheses outlined earlier: spirals of conflict imply steadily rising baseline hazards, while if disputes are self-regulating, those hazards should decline. More generally, we suggest that, at a minimum, the assumption that all sequential disputes are exactly the same, regardless of their order, needs to be tested.

One of These Dyads Is Not Like the Others: Unobserved Heterogeneity in International Relations Research The consideration of repeated events is very closely related to the issue of heterogeneity among pairs of nations. Implicit in all existing quantitative studies of international conflict is the assertion that observations are exchangeable; that is, conditional on the covariates considered in the model, the probability of any two nations engaging in a dispute is the same (King 2001). Such an assumption is, we believe, not supportable on either theoretical or empirical grounds: unmeasured and unmeasureable factors such as history, culture, and exogenous shocks clearly impact the probability of conflict.3 In addition, heterogeneity may also arise more directly from repeated events, since the correlation between repeated conflicts can be viewed as a specific form of heterogeneity. Because such heterogeneity amounts to a specific form of model misspecification, these influences make the conclusions drawn using standard statistical approaches problematic at best and simply incorrect at worst. Accordingly, in addition to the aforementioned variance-corrected approaches for addressing repeated events, we suggest the use of mixture models, including “cure” and “frailty” models, as a means of addressing unmeasured heterogeneity Page 280 → and discuss how such models both conform closely to our understanding of international disputes and provide superior assessments of the causes of those disputes. Previous work on international conflict has, on occasion, recognized the importance of such unmeasured factors in studying the onset of war. One significant example is the debate surrounding the definition and use of “politically relevant dyads”: pairs of nation-states that, because of geography or international status, have more than a trivial chance of coming into conflict (e.g., Maoz and Russett 1993; Bennett and Stam 2000; Lemke and Reed 2001; Barbieri this volume). The argument for examining only politically relevant dyads lies in the hundreds of thousands of dyads for which conflict is essentially impossible (e.g., Surinam-Malaysia); because such dyads offer little information about the roots of international conflict, scholars argue, they can and should be omitted from quantitative analyses of international disputes. Others question the politically relevant approach, arguing that focusing on pairs of states we know are seriously thinking about fighting each other because of a clash over a serious political issue is the correct approach (e.g., Huth 1996).4 This approach, however, leads directly to the problem of determining what dyads are seriously thinking about war over time. And still others advocate the analysis of all dyads in such studies, on the belief that their exclusion represents a potentially more serious threat to inference than any ill effects from their inclusion. The issue of heterogeneity is especially salient in studying the interdependence-conflict connection. For building large data sets covering the entire international system, trade data between states is available only for aggregate levels of exports and imports. However, within these aggregate measures, trade in different kinds of goods is likely to vary in impact. Trade in goods that are strategic is likely to have a greater pacifying effect, as a state may be less willing to accept an interruption in oil imports, for example, than in children's toys. Similarly, some domestic industries may be more politically powerful, meaning that they will be more effective in preserving peace if export-reducing conflict threatens. Busch and Reinhardt (1999), for example, found that geographically concentrated but politically dispersed industries are more effective in affecting foreign-trade policy. In short, among dyads that have apparently equal levels of trade the pacifying effect of trade is likely to vary in ways that

are practicably unobservable. The central issue with politically relevant dyads is the possibility of a dispute—that is, whether the nations in question are ever likely to experience a conflict. Cure (or “split-population”) models allow for the possibility that Page 281 → some observations in the data will never experience the event of interest (e.g., Maller and Zhou 1996; Schmidt and Witte 1989).5 Most often, this is accomplished by setting the hazard of the event equal to a mixture of a standard distribution and a point mass at zero; in this way, some observations are allowed to be “immune” from ever experiencing the event of interest. These models are thus an improvement on more standard approaches, for two critical reasons. First, in contrast to cure models, all other duration models assume that every observation in the data will eventually experience the event being studied. In the context of international conflict, this is equivalent to stating that all dyads will someday have a militarized dispute. Second, cure models allow the researcher to empirically assess the likelihood that an observation is “cured.” Thus, rather than relying on necessarily arbitrary ex ante distinctions about which dyads may or may not clash, cure models offer a means of determining both whether and when a conflict may or may not occur and of assessing the influence of covariates on both of those processes. Importantly, the covariates may have different effects for different parts of the model, that is, for whether versus when the event may occur. The effects of ignoring unobserved heterogeneity of any sort are as widely understood as they are pernicious: biased estimates of variable effects, incorrect standard errors, and a lack of consistency in parameter estimates. Heterogeneity thus represents a serious threat to our understanding of international conflict, a fact scholars are only now beginning to appreciate (e.g., Clark and Regan 2001; Beck and Katz 2001; Bennett and Stam 2000; Green, Kim, and Yoon 2001; King 2001; Zorn 2000; but see Bennett 1997 for an earlier exception). One prominent means for making valid inferences in the presence of such heterogeneity has been the use of frailty models (e.g., Manton, Stallard, and Vaupel 1981; Omori and Johnson 1993; Sastry 1997).6 Frailty models, also known as “random-effects” models in econometrics, allow for individual heterogeneity in the form of a subjectspecific term that captures that particular observation's unobserved propensity toward the event of interest. A general form of such models is Y i = f (Xiβ + αi + ε), where the α i represents individual- (here, dyad-) level heterogeneity. The most common approach is to assume that the αs are random draws from some known distribution, the parameters of which are then estimated along with the model coefficients. Such models provide a direct means of both assessing the Page 282 → extent of individual-level heterogeneity in the data and of making valid inferences about covariate effects in the presence of such heterogeneity. Because of these characteristics, frailty models have been highlighted as particularly promising models for event-history research. We argue here that a frailty approach to the study of international conflict offers significant improvements over conventional methods. By adopting frailty approaches, researchers can estimate the unobserved tendency toward conflict in a dyad, even after controlling for those factors we know contribute to the likelihood of such conflict. The importance of this fact is clear: such a flexible empirical approach is generally to be preferred over necessarily arbitrary distinctions about “political relevance” and methods requiring one to omit large amounts of data (such as fixed-effects models; e.g., Green, Kim, and Yoon 2001).7 In addition, frailty models offer an attractive solution to the problem of obtaining consistent estimates of variable effects in the face of such heterogeneity. These models are also straightforward to estimate and interpret and are easily implemented using standard software, making their potential for widespread adoption by substantively oriented political scientists high. Finally, frailty models may help with concerns over the dyadic focus in international relations, since one of the advantages of frailty models is their ability to allow for multilevel inferences, for example, to assess the influence of dyadic, nation-, alliance-, and regional-level variables (Jones and Steenbergen 1999; Sastry 1997). In the introduction to this volume Edward D. Mansfield and Brian M. Pollins highlight the theoretical importance of

identifying relevant actors at various levels. Multilevel duration models offer a promising methodological match for this problem. The underlying logic of frailty models is that some observations (or groups of observations) are intrinsically more or less prone to experiencing the event of interest than are others. We submit that it is important in all empirical endeavors to assess the potential biases that may result from unobserved heterogeneity in one's data and to employ methods appropriate for correcting those biases; the models offered here provide an attractive means of doing so.

Quantitative Methods and the Future of International-Conflict Research Advances in the quantitative study of interdependence and international conflict need to be melded with the increasing theoretical and data-gathering sophistication of this work. Our goal here has been to outline a number of promising directions for the continuing evolution of these methods. In particular, Page 283 → we focus on the issue of temporal dynamics and heterogeneity in international conflict and draw upon techniques developed in biostatistics, epidemiology, and demography to address those issues. While scholars have long recognized the importance of temporal dynamics in the study of international relations, there has not been enough attention to, let alone agreement about, the best way to address these dynamics empirically. We argue that allowing time-varying effects of covariates, statistical incorporation of facts such as repeated conflicts, and recognition of and accounting for heterogeneity will help align the theory and quantitative analysis. The solutions we offer here, while general, nonetheless require a good deal of hard substantive thinking on the part of scholars. Issues of case selection, measurement, and model specification are critical to the success of these or any other empirical endeavors. Our proposals must thus be considered a starting point, rather than the last word, about the directions scholars should take in further empirical analyses of international disputes. Nonetheless, we hope that they will inspire researchers to continue with ever more rigorous and informative work on this subject of critical importance.

NOTES We would like to thank the participants of the Interdependence and Conflict Conference, the editors of the volume, Bridget Coggins, and Andy Farrell for comments. 1. One study that accounts for nonproportionality found that trade was not significantly correlated with conflict (Box-Steffensmeier and Zorn 2001a). 2. Recent exceptions include Beck, Katz, and Tucker 1998; Jackman 2001; and Crescenzi and Enterline 2001. 3. More generally, Blossfeld and Rohwer (1995, 243) point out that, as social scientists, “we can usually be sure that we were not able to include all important covariates” in our models. 4. Alternatively, Lemke (1995) suggests “military reachability” as a more viable means for assessing which dyads have at least the potential for conflict. 5. Applications of cure models in political science include Box-Steffensmeier and Radcliffe 1996; Hettinger and Zorn 2002; and Clark and Regan 2001. 6. The term frailty comes from biostatistics and refers to the notion that some observations are more “frail” than others (due to unmeasured factors) and thus will experience the event of interest (often the recurrence of a disease) earlier. 7. In fixed-effects models, each of the individual αs are estimated along with the model's structural parameters. The literature on fixed effects points out a number of problems with these models (e.g., Andersen, Klein, and Zhang 1999). Among these is the Page 284 →“incidental parameters problem”: absent a truly “fixed” number of fixed effects, fixed-effects ML estimators are inconsistent (Lancaster 2000). Moreover, fixed effects prevent researchers from estimating the impact of variables that do not vary within units/dyads (e.g., contiguity), a significant disadvantage in studies of international conflict.

REFERENCES Andersen, Per Kragh, and R. D. Gill. 1982. Cox's Regression Model for Counting Processes: A Large Sample Study. The Annals of Statistics 10:1100–1120. Andersen, Per Kragh, John P. Klein, and Mei-Jie Zhang. 1999. Testing for Centre Effects in Multi-Centre Survival Studies: A Monte Carlo Comparison of Fixed and Random Effects Tests. Statistics in Medicine 18:1489–1500. Barbieri, Katherine. 1996. Economic Interdependence: A Path to Peace or a Source of Interstate Conflict? Journal of Peace Research 33:29–49. Barbieri, Katherine, and Jack S. Levy. 1999. Sleeping with the Enemy: Trade between Adversaries during Wartime. Journal of Peace Research 36:463–79. Beck, Nathaniel. 1999. Trade and Conflict in the Cold War Era: An Empirical Analysis. Paper presented at the 95th annual meeting of the American Political Science Association, September 25, Atlanta. Beck, Nathaniel, and Jonathan N. Katz. 2001. Throwing Out the Baby with the Bath Water: A Comment on Green, Yoon, and Kim. International Organization 55:487–95. Beck, Nathaniel, Jonathan N. Katz, and Richard Tucker. 1998. Taking Time Seriously:Time-Series-Cross-Section Analysis with a Binary Dependent Variable. American Journal of Political Science 42:1260–88. Bennett, D. Scott. 1997. Testing Alternative Models of Alliance Duration, 1816–1984. American Journal of Political Science 41:846–78. ______. 1998. Integrating and Testing Models of Rivalry Duration. American Journal of Political Science 42:1200–1232. Bennett, D. Scott, and Allan C. Stam. 2000. Research Design and Estimator Choices in the Analysis of Interstate Dyads: When Decisions Matter. Journal of Conflict Resolution 44:653–85. Blossfeld, Hans-Peter, and Götz Rohwer. 1995. Techniques of Event History Modeling. Mahwah, NJ: Lawrence Erlbaum. Box-Steffensmeier, Janet M., and Peter Radcliffe. 1996. The Timing of PAC Contributions: A Split-Population Survival Model. Paper presented at the 92nd annual meeting of the American Political Science Association, August 29–September 1. San Francisco. Box-Steffensmeier, Janet M., Dan Reiter, and Christopher J. W. Zorn. 2002. Modeling Unobserved Heterogeneity in International Conflict Data. Paper presented at the Annual Meeting of the Midwest Political Science Association, Chicago. Box-Steffensmeier, Janet M., and Christopher J. W. Zorn. 2001. Duration Models and Page 285 → Proportional Hazards in Political Science. American Journal of Political Science 45:972–88. ______. 2002. Duration Models for Repeated Events. Journal of Politics 64 (November):1069–94. Busch, Marc L., and Eric Reinhardt. 1999. Industrial Location and Protection: The Political and Economic Geography of U.S. Nontariff Barriers. American Journal of Political Science 43:1028–50. Clark, David, and Patrick Regan. 2001. Opportunities to Fight: A Statistical Technique for Modeling Unobservable Phenomena. Typescript, Binghamton, NY. Crescenzi, Mark J. C., and Andrew J. Enterline. 2001. Time Remembered: A Dynamic Model of Interstate

Interaction. International Studies Quarterly 45:409–31. Diehl, Paul F. 1998. The Dynamics of Enduring Rivalries. Urbana: University of Illinois Press. Fearon, James D. 1995. Rationalist Explanations for War. International Organization 49:379–414. Gartzke, Erik. 1999. War Is in the Error Term. International Organization 53:567–87. Gartzke, Erik, Quan Li, and Charles Boehmer. 2001. Investing in the Peace: Economic Interdependence and International Conflict. International Organization 55:391– 438. Gasiorowski, Mark. 1986. Economic Interdependence and International Conflict: Some Cross-National Evidence. International Studies Quarterly 30:23–38. Goertz, Gary, and Paul F. Diehl. 2000. War and Peace in International Rivalry. Ann Arbor: University of Michigan Press. Gowa, Joanne. 1999. Ballots and Bullets. Princeton: Princeton University Press. Grambsch, Patricia M., and Terry M. Therneau. 1994. Proportional Hazards Tests and Diagnostics Based on Weighted Residuals. Biometrika 81:515–26. Green, Donald P., Soo Yeon Kim, and David Yoon. 2001. Dirty Pool. International Organization 55:441–68. Hegre, Havard. 2000. Development and the Liberal Peace: What Does It Take to Be a Trading State? Journal of Peace Research 37:5–30. Hettinger, Virginia, and Christopher J. W. Zorn. 2002. Explaining the Incidence and Timing of Congressional Responses to the U.S. Supreme Court. Typescript, Emory University. Huth, Paul K. 1996. Standing Your Ground: Territorial Disputes and International Conflict. Ann Arbor: University of Michigan Press. Jackman, Simon. 2001. In and Out of War and Peace: Transitional Models of International Conflict. Typescript, Stanford University. Jervis, Robert. 1976. Perception and Misperception in International Relations. Princeton:Princeton University Press. ______. 1978. Cooperation under the Security Dilemma. World Politics 30:167–214. Jones, Bradford S., and Marco Steenbergen. 1999. Modeling Multilevel Data Structures.Working paper, University of North Carolina, Chapel Hill. Page 286 → Kalbfleisch, J. D., and R. L. Prentice. 1980. The Statistical Analysis of Failure Time Data. New York: John Wiley. Kelly, Patrick J., and Lynette L-Y. Lim. 2000. Survival Analysis for Recurrent Event Data. Statistics in Medicine 19:12–33. Keohane, Robert O., and Joseph S. Nye. 1989. Power and Interdependence. Glenview, IL: Scott Foresman. King, Gary. 2001. Proper Nouns and Methodological Propriety: Pooling Dyads in International Relations Data. International Organization 55:497–507.

Klein, John P., and Melvin L. Moeschberger. 1997. Survival Analysis. New York: Springer. Lancaster, Tony. 2000. The Incidental Parameter Problem since 1948. Journal of Econometrics 95:391–413. Larson, Deborah Welch. 1997. Anatomy of Mistrust: U.S.–Soviet Relations during the Cold War. Ithaca: Cornell University Press. Lebow, Richard Ned. 1981. Between Peace and War: The Nature of International Crisis. Baltimore: Johns Hopkins. Lemke, Douglas. 1995. The Tyranny of Distance: Redefining Relevant Dyads. International Interactions 21:23–38. Lemke, Douglas, and William Reed. 2001. The Relevance of Politically Relevant Dyads. Journal of Conflict Resolution 45:126–44. Maller, R. A., and S. Zhou. 1996. Survival Analysis with Long-Term Survivors. New York: Wiley. Mansfield, Edward D. 1994. Power, Trade, and War. Princeton: Princeton University Press. Mansfield, Edward D., and Jon C. Pevehouse. 2000. Trade Blocs, Trade Flows, and International Conflict. International Organization 54:775–808. Manton, K. G., E. Stallard, and J. W. Vaupel. 1981. Methods for Comparing the Mortality Experience of Heterogeneous Populations. Demography 18:389–410. Maoz, Zeev, and Bruce Russett. 1993. Normative and Structural Causes of Democratic Peace, 1946–1986. American Political Science Review 87:624–38. Ng'andu, N. H. 1997. An Empirical Comparison of Statistical Tests for Assessing the Proportional Hazards Assumption of Cox's Model. Statistics in Medicine 16:611–26. Omori, Y., and R. A. Johnson. 1993. The Influence of Random Effects on the Unconditional Hazard Rate and Survival Functions. Biometrika 80:910–14. Oneal, John R., and Bruce M. Russett. 1997. The Classical Liberals Were Right: Democracy, Interdependence, and Conflict, 1950–85. International Studies Quarterly 41:267–94. ______. 1999a. Assessing the Liberal Peace with Alternative Specifications: Trade Still Reduces Conflict. Journal of Peace Research 36:423–42. ______. 1999b. The Kantian Peace: The Pacific Benefits of Democracy, Interdependence, and International Organization, 1885–1992. World Politics 52:1–37. Page 287 → Papayoanou, Paul. 1999. Power Ties: Economic Interdependence and War. Ann Arbor: University of Michigan Press. Polachek, Solomon W. 1980. Conflict and Trade. Journal of Conflict Resolution 24:57–78. Prentice, R. L., B. J. Williams, and A. V. Peterson. 1981. On the Regression Analysis of Multivariate Failure Time Data. Biometrik a 68:373–79. Raknerud, Arvid, and Håvard Hegre. 1997. The Hazard of War: Reassessing the Evidence for the Democratic Peace. Journal of Peace Research 34:385–404.

Reed, William. 2000a. Information, Bargaining, and Economic Interdependence. Manuscript, East Lansing, MI. ______. 2000b. A Unified Statistical Model of Conflict Onset and Escalation. American Journal of Political Science 44:84–93. Reed, William, and David H. Clark. 2000. War Initiators and War Winners: The Consequences of Linking Theories of Democratic War Success. Journal of Conflict Resolution 44:378–95. Reed, William, and Douglas Lemke. 2001. The Relevance of Politically Relevant Dyads. Journal of Conflict Resolution 45:126–44. Reiter, Dan. 1996. Crucible of Beliefs: Learning, Alliances and World Wars. Ithaca: Cornell University Press. Reuveny, Rafael, and Heejoon Kang. 1996. International Trade, Political Conflict/Cooperation and Granger Causality. American Journal of Political Science, 40:943–70. Sastry, Naryan. 1997. A Nested Frailty Model for Survival Data, with an Application to the Study of Child Survival in Northeast Brazil. Journal of the American Statistical Association 92 (438): 426–35. Schemper, Michael. 1992. Cox Analysis of Survival Data with Non-Proportional Hazard Functions. The Statistician 41:455–65. Schmidt, Peter, and Ann D. Witte. 1989. Predicting Recidivism Using “Split-Population” Survival Time Models. Journal of Econometrics 40:141–59. Schultz, Kenneth A. 2000. Do Democracies Mean What They Say? Regime Type and Threat Implementation in International Crises. Typescript, Princeton University. ______. 2001. Looking for Audience Costs. Journal of Conflict Resolution 45:32–60. Signorino, Curtis S. 1999. Strategic Interaction and the Statistical Analysis of International Conflict. American Political Science Review 93:279–98. Smith, Alastair. 1996. To Intervene or Not to Intervene: A Biased Decision. Journal of Conflict Resolution 40:16–40. ______. 1999. Testing Theories of Strategic Choice: The Example of Crisis Escalation. American Journal of Political Science 43:1254–83. Waltz, Kenneth. 1979. Theory of International Politics. New York: Random House. Wei, L. J., D. Y. Lin, and L. Weissfeld. 1989. Regression Analysis of Multivariate Incomplete Failure Time Data by Modeling Marginal Distributions. Journal of the American Statistical Association 84 (408): 1065–73. Page 288 → Werner, Suzanne. 1999. The Precarious Nature of Peace: Resolving the Issues, Enforcing the Settlement and Renegotiating the Terms. American Journal of Political Science 43:912–34. Zorn, Christopher J. W. 2000. Modeling Duration Dependence. Political Analysis 8:367–80.

Page 289 →

Concerns with Endogeneity in Statistical Analysis: Modeling the Interdependence between Economic Ties and Conflict Richard J. Timpone In their respective and intersecting literatures, the level of economic interdependence among nations and the likelihood of their engaging in conflict are both treated as dependent or endogenous variables that are determined within the system of interest. While economic ties and conflict are each important dependent variables, as the contributions in this volume demonstrate, there remain a number of debates regarding the relationship between the two. Does economic interdependence decrease the likelihood of conflict between states? Do political-military power relations and conflict themselves affect economic ties? While traditional approaches to testing hypotheses such as these often treat the relevant factor of interest as the dependent variable and the other as an independent variable in an equation, the fact that both are endogenously determined in a larger system of equations has consequences for empirical models that deserve more consideration. What one theorizes about these causal relationships has important implications for the proper formulation of quantitative analyses. As will be discussed, the statistical concerns regarding the modeling of these forces are not isolated to theories of reciprocal causation but may require further justification even in “simpler” models because both economic ties and military conflict can be considered endogenous in a larger system. Including an endogenous factor as an explanatory variable in a regression model may create problems for accurately understanding the relationships of interest because of the possibility of biased Page 290 → results. This is true for ordinary least squares (OLS), probit, and logit models and their extensions. As William of Occam stated, “Non sunt multiplicanda entia praeter necessitatem”: we should not multiply the complexity of our models beyond what is necessary (Starfield, Smith, and Bleloch 1990). While this is often misinterpreted as mandating the simplest possible model, it actually requires that scholars develop models that both are simple and accurately reflect the system being examined. Some analyses may have cut too deeply with “Occam's Razor” by ignoring the fact that both economic and military relations are endogenous, and it may be necessary to add additional complexity for accurate understanding. In this chapter, I will discuss why simplifications may lead to bias in models testing theories that consider reciprocal causation as well as those that do not. Several extensions to traditional single-equation approaches (i.e., ordinary least squares, probit, and logit models) allow for different types of measurement of the endogenous factors. Some of these are relatively simple to implement while others are more complex and point to future directions that may prove fruitful for methodological attention. While technical matters regarding the choice of estimator will be discussed, it is important to acknowledge concerns that exist with the “solutions” used to address the concerns in practice as well. The chapter will therefore follow the discussion of different statistical estimators with caveats about the practical applications of these solutions. Given the properties of the estimators and concerns regarding the solutions, I will therefore discuss tests to determine if traditional single-equation approaches may be appropriate for specific analyses before concluding.

Concerns with Traditional Single-Equation Approaches Concerns That May Arise without Reciprocal Causation Much of the research that serves as the foundation for the debates addressed in this volume examines models of conflict where measures of economic interdependence are included as independent variables to test whether such economic ties have an effect on the likelihood of disputes. Ignoring the numerous types of economic links, I will simplify the discussion of the models by dubbing these simply as “trade.” In this way, a statistical model may be set up along the line of

conflict = f 1 (trade, X1, ε1).

(1a)

Page 291 → In equation (1a), conflict is a function of trade, a set of exogenous variables X 1, and the stochastic error term ε 1. Unlike endogenous variables, which are determined within the system, exogenous variables are assumed to be determined outside of it and uncorrelated with the error term. Traditional regression models assume all independent variables are exogenous, but as we will see there are reasons to be concerned when we use endogenous explanatory variables, like trade, in equation (1a). There are numerous debates in the literature regarding how to analyze equation (1a) in practice. These include choices in operationalizing “conflict” and “trade,” as well as the set of exogenous factors included and the functional forms of the relationships. The questions of the interdependence between trade and conflict that I focus on in this chapter deal with the assumption of statistical models that all independent variables are uncorrelated with the error term in the estimated equation. Concerns with the appropriateness of this assumption exist regardless of whether or not one theorizes there to be reciprocal causation when endogenous explanatory variables are included. Even though we treated trade as an independent variable in equation (1a), we know that it is itself a function of other factors. We can therefore set up a second equation for trade: trade = f2 (X2, ε2).

(1b)

In equation (1b), trade is a function of a set of exogenous factors X 2, as well as a stochastic error term ε 2. The sets of exogenous factors X 1 and X 2 are likely to overlap, as we know that numerous factors, such as geographic proximity, are clearly related to both trade and conflict. In this case, equations (1a) and (1b) are known as the structural or behavioral equations that together form model 1. If a researcher is only interested in modeling conflict, it may be appropriate to look at equation (1a) in isolation, but this is not always the case. Equations (1a) and (1b) form a triangular/hierarchical model. This simply describes how the endogenous factors, trade and conflict, are related within the larger system. At the top of the hierarchy, the systematic component of trade is only a function of exogenous factors, X 2, that are assumed uncorrelated with the errors in the model by definition. Going lower in the hierarchical system, conflict is a function of the earlier endogenous factor trade as well as the set of exogenous variables X1. In this hierarchical model, the equation for trade can be estimated in isolation without any corrections for the problems discussed here, since all of the Page 292 → independent variables in equation (1b) are exogenous and uncorrelated with the error. The key to whether or not equation (1a) can be estimated in isolation revolves around the relationship between ε 1 and ε 2, and by extension between ε 1 and trade. While the variables that constitute X 1 are not correlated with ε 1, does this also hold for trade? It is possible that ε 1 and ε 2 are uncorrelated, in which case a recursive system exists. If a system of equations is recursive (hierarchical with uncorrelated errors), each equation can be estimated in isolation (using OLS, logit, probit, etc., as appropriate) without concern over bias caused by the inclusion of the endogenous explanatory variables. Even if a researcher is only interested in modeling conflict, that is, equation (1a), a correlation between the error terms leads to bias. This is likely to occur if important factors systematically related to both trade and conflict are left out of their respective equations. These would then become part of the error terms and lead to a correlation between them. Since trade is a function of ε 2, the correlation between the errors will lead trade to also be correlated with ε 1. An alternative way to consider this is that, because trade is related to the omitted factors in equation (1a), it will therefore be correlated with the error term that includes the factors. For this reason, the potential problem for estimating equation (1a) in isolation (in this hierarchical case) is often discussed with issues of model specification and the general problem of omitted-variable bias. Many standard texts include discussions

of model specification, including the problems caused by omitted variables and ways to test for them (e.g., Greene 2000; Gujarati 2003; Pindyck and Rubinfeld 1998). While theory leads to concerns about endogeneity, these texts also discuss statistical tests that can be applied to determine if an explanatory variable is correlated with the error term under certain conditions. These will be reviewed in the fourth section, “Testing If Corrections for Endogeneity Are Needed.” The accuracy of these estimation results therefore is a function of model specification and the variables included for theoretic and control purposes. Estimating equation (1a) in isolation can only avoid potential bias due to correlation between trade and ε 1 if the set of exogenous variables X 1 incorporate all of those factors that are directly related to both trade and conflict. This is true regardless of whether the researcher is substantively interested in the full system or only in modeling conflict and ignores equation (1b) entirely. Concerns When Reciprocal Causation Is Theorized or Tested As was seen in the previous section, even if a scholar is only interested in modeling one of the areas and assumes that the direction of causality flows in one Page 293 → direction, consideration of the fuller system is needed to insure that the estimation of each individual equation is unbiased. In that case, the problem could in theory be handled with adequate statistical controls in the set of exogenous variables included in the estimation. Adopting theories of reciprocal causation always requires consideration of the fuller system for accurate estimation even if a researcher is primarily interested in what is causing one of the dependent variables, trade or conflict. In this case, the system of equations would not be hierarchical and would be represented by different functions than those discussed earlier. Now we would have conflict = g1 (trade, X1, δ1),

model 2

trade = g2 (conflict, X2, δ2). The two equations that constitute model 2 form a system where conflict and trade are reciprocally related to each other and are functions of exogenous factors X 1 and X 2 and of stochastic error terms δ 1 and δ 2 respectively. With the hierarchical assumption, bias exists if the error terms are correlated. The problem of bias is more serious in cases of reciprocal causation. Even if the errors in model 2 were uncorrelated, δ 1 is a component of conflict that is causally related to trade; thus trade will be correlated with δ 1. The same is true for conflict and δ 2 given its chain of causality. Even if the errors were uncorrelated across the equations in model 2, reciprocal causation leads to correlation between the endogenous factors on the right-hand side of each equation and their respective error terms. While such correlation was a potential problem with the hierarchical system in model 1 that could possibly be addressed with improved specification, it is endemic to nonhierarchical models and cannot be as simply corrected.

Addressing These Statistical Issues There are a number of different estimators that have been developed to address the issue of endogenous variables being correlated with the errors in models. Here, I will focus on the extensions of linear-regression and basic maximum likelihood models, which are the most common statistical tools used to test these hypotheses. While I introduce a number of approaches briefly in this chapter, many standard texts provide fuller discussions of the problems and basic techniques to address them (e.g., Greene 2000; Gujarati 2003; Pindyck and Rubinfeld 1998). Even with these regression extensions, differences and difficulties emerge depending on the nature of the measurement of the dependent Page 294 → variables, that is, whether they are continuous or categorical. All of the various approaches have certain commonalities, including the need for instrumental variables that provide additional information to obtain consistent estimates of the parameters of interest. These are variables that are uncorrelated with the error terms and allow for identification of the estimators. Given the centrality of these variables, special attention will be given to caveats that affect the alternative approaches in practice.

Extending traditional ordinary least squares regression and maximum likelihood approaches, several estimators have been developed to obtain consistent estimates for equations under these conditions. While full-information estimation of an entire system of equations simultaneously is possible, limited-information estimators are the most common, given their greater simplicity and desirable statistical properties (Greene 2000, 693–96, discusses fullinformation estimation of systems of equations). The limited-information estimators allow for the estimation of a single equation in a system, for instance, equation (1a), and correct for problems where the endogenous variables serving as regressors (trade in that case) are (or may be) correlated with the error term. Thus, these are equationby-equation estimators that are appropriate whether one is interested in modeling all equations or only one in a broader system. These approaches may be used for any nonrecursive system, that is, if the system is not hierarchical or is hierarchical and the errors are correlated. Problems Where Both Endogenous Variables Are Continuous Using the limited-information approach to obtain consistent estimates of a single equation underlies the frequently used statistical methods of instrumental-variable analysis or two-stage least squares (2SLS) for handling continuous variables. In these approaches, instrumental variables, variables uncorrelated with the error term, are used to provide information in place of the endogenous variables on the right-hand side of the equation. In 2SLS, the researcher first creates predicted values of right-hand-side endogenous terms using all of the exogenous variables in the system (the union of the sets of variables in X 1 and X 2 in the two models discussed earlier). These first-stage equations are known as the reduced-form equations. Because the exogenous variables are uncorrelated with the error terms, the predictions created from them will likewise be uncorrelated with the errors. The prediction from a reduced-form equation is plugged into the second-stage analysis in place of the actual endogenous variable. The results of this analysis will produce consistent estimates of the model parameters (as long as quality instrumental Page 295 → variables are available, as discussed in the third section of the chapter, “Caveats Regarding the Implementation of Corrections”), but the standard errors will be wrong because they are based on the predicted values of the endogenous explanatory factor. The instrumental-variable approach described in the next paragraph is preferable when all endogenous variables are continuous in an equation. However, the manual two-stage approach introduced here will serve as the foundation for extensions into problems where the nature of the measures creates additional complexity. While 2SLS is one approach to using additional information for consistent parameter estimation, this could also be handled in one step with an instrumental-variable estimator that folds the two stages into a single estimation by extending the matrix algebra used in ordinary least squares. While the parameter estimates of each approach are consistent, the advantage of this instrumental-variable estimator over the manual two-stage approach is accurate determination of the standard errors, whereas in the manual two-stage calculation these must be corrected. Many statistical textbooks provide overviews of 2SLS/instrumental-variable extensions to OLS (e.g., Hanushek and Jackson 1977; Greene 2000; Gujarati 2003; Pindyck and Rubinfeld 1998). Also, the formulas for correcting the standard errors in the manual application of 2SLS in this case are known and can be implemented as well (see, e.g., Achen 1986; Gujarati 2003). While the basic statistical concerns and the 2SLS/instrumental-variable corrections may be familiar to many with advanced statistical training, the situation is rarely so easily handled in the modeling of economic interdependence and conflict as discussed thus far. The reason for this is that the desirable statistical property of consistency assumes that we are dealing with continuous variables. While some economic data may fit this assumption, many common operationalizations of conflict do not (such as dichotomous measures of engaging in militarized disputes). Even the economic measures may violate this if we consider factors like involvement in trade agreements (this is true whether we use these as our sole measures of economic interdependence or build larger systems of equations with more complex interrelations between the factors). With discrete variables, parameter estimates may not be consistent even with large samples and quality instrumental variables. Unlike the 2SLS/instrumental-variable corrections for endogeneity with continuous variables, extensions for the types of variables common in studies of economic interdependence and conflict are not as straightforward. We will continue with the limited-information approach of estimating each equation in a broader system and

correcting for the problems caused by endogenous Page 296 → explanatory variables. While the earlier discussions explain the conceptual solution for two continuous variables, it is worthwhile to consider separately three distinct cases: (1) the case of the discrete dependent variable/continuous endogenous explanatory variable; (2) the case of the continuous dependent variable/discrete endogenous explanatory variable; and (3) the case where neither of the endogenous variables on the right-hand side (explanatory) or left-hand side (dependent variable in the equation) is continuous. For simplicity in the following discussions, and given some of the common operationalizations of variables of interest (militarized interstate disputes [MIDs], trade organizations, etc.), I will treat the discrete variables as dichotomous. The core extension will be two-stage probit regression in these conditions although other approaches are appropriate in some cases as well. Dichotomous Dependent Variable/Continuous Endogenous Explanatory Variable The simplest approach to dealing with any endogenous dichotomous variable (either explanatory or explained) is to extend ordinary least squares to a linear-probability model. In fact, given the complexity of preferable corrections in some of the conditions, this has often been the approach endorsed and implemented. Running ordinary least squares on a dichotomous variable results in inefficient estimates by definition, since the structure of the estimation is guaranteed to create heteroskedasticity (Aldrich and Nelson 1984; Long 1997). Just as weighted least squares can be implemented to adjust for the heteroskedasticity, in this case additional corrections can be made if one uses 2SLS. Achen (1986) details the necessary corrections to extend the previous logic to a generalized two-stage least squares framework. While it is the easiest solution to estimate, problems of functional form have been the key reason for moving from linear-probability models to logit and probit treatments of dichotomous dependent variables in general. Shifting from linear-probability models to these maximum likelihood estimators leads to added complexity when addressing issues of endogeneity. While extensions of the 2SLS approach have been developed, they require manual estimation of the multiple stages and corrections to produce consistent parameter estimates and correct standard errors. The first extension to the 2SLS/instrumental-variable approach that can deal with the case where the dependent variable is dichotomous and the endogenous explanatory variable is continuous is known as two-stage probit least squares, although I will treat this as one condition under the more general Page 297 → rubric of two-stage probit regression (2SPR). Again, this could apply to either of the equations explaining conflict in model 1 or 2 regardless of theories of reciprocal causation. Because the endogenous explanatory variable is continuous, the first stage is the same as in traditional 2SLS models. An OLS reduced-form equation for this factor is created using all exogenous variables in the system (the union of the sets of X 1 and X 2 in model 1 or 2). The predicted value is used in place of the original variable, and probit is then run on the second-stage dependent variable. An additional rescaling step is necessary to produce consistent parameter estimates. The statistical derivation and practical application of this technique are described in detail elsewhere (Achen 1986, 48–50; Alvarez and Glasgow 2000; Maddala 1983, 244–45). The main drawback of 2SPR is that the standard errors produced are biased and their correction is very difficult. In a later section, I discuss how this limitation may be addressed by using 2SPR with bootstrapped standard errors. The problems in calculating the standard errors provided one motivation for the Alvarez and Glasgow (2000) discussion of two-stage conditional maximum likelihood (2SCML), whose technical derivation and comparison with other estimators is provided by Rivers and Vuong (1988). Rivers and Vuong (1988) developed the 2SCML estimator as an alternative to 2SPR to allow for consistent estimation where the dependent variable is dichotomous and the independent variable is continuous (see Alvarez and Glasgow 2000 for a less technical introduction and application of this technique). While this is also a consistent estimator of the model parameters, Rivers and Vuong (1988) demonstrate that it has advantages in efficiency as well as calculation, although Alvarez and Glasgow's (2000) simulations show that the 2SPR approach is slightly superior in reducing the bias in the coefficient estimates caused by the endogeneity. In spite of this, the accuracy in the estimation of the standard errors leads Alvarez and Glasgow (2000) to support use of the 2SCML estimator. The 2SCML approach begins like the 2SLS and 2SPR approaches, with the estimation using the endogenous continuous explanatory variable as a dependent variable in a first-stage OLS reduced-form regression model using

all exogenous variables in the system. In the second stage, the residuals of the first-stage model are included in the probit model as an additional variable along with the original endogenous regressor. In addition to producing consistent estimates and accurate standard errors, this also allows for the statistical testing of endogeneity by comparing the log of the likelihood functions calculated with and without the residuals included in the second stage (again, Alvarez and Glasgow 2000 provides a fuller overview of this estimator). Page 298 → Continuous Dependent Variable/Dichotomous Endogenous Explanatory Variable The correction for endogenous explanatory variables (regardless of whether or not one treats the relationship as reciprocal) is also more complicated than the traditional instrumental-variable/2SLS approaches if the variable on the right-hand side of the equation is dichotomous. Again, the simplest approaches would treat the dichotomous variable estimation as a linear-probability model and extend the corrections in traditional 2SLS with weighted least squares. As discussed earlier, these further extensions do not address the problems of functional form that come from the use of the linear and additive specification of OLS with a dichotomous dependent variable (in the first stage in this case). The 2SPR approach for obtaining consistent parameter estimates under these conditions is similar to the dichotomous dependent variable/continuous endogenous explanatory variable condition, although the reducedform estimation and rescaling are different (Maddala 1983, 244–45). In the first stage, the reduced-form model is estimated for the dichotomous endogenous variable. Instead of OLS, a probit model is calculated using all exogenous variables in the system (again the union of the variables that constitute X 1 and X 2). There are different ways to derive the probit specification, and one way is to consider the dichotomous variable to be generated from an unobserved continuous index function, where a value of zero is observed if the value of this unobserved variable is below a threshold and a value of one is observed if it is above. Predicted values from probit models, as in the first stage here, produce “ z “ values on this unobserved continuum. While normally we transform these raw values into probabilities for substantive interpretation, the raw untransformed continuous z value predictions would be used in place of the endogenous variable in the second stage. The rescaling is different than the dichotomous dependent variable/continuous endogenous explanatory variable situation as the parameter estimates on the exogenous explanatory variables in the second stage do not need rescaling (unlike the dichotomous dependent variable case), although the parameter on the purged endogenous indicator would require a transformation for substantive interpretation. As in the earlier discussion of the 2SPR approach, the standard errors in the estimates in the second-stage equation of interest will be wrong and difficult to correct. Developing 2SPR estimators that incorporate bootstrapped standard errors would be even more useful in this case since there is not a simple approach to correct the results under these conditions (without moving to the assumptions of the linear-probability model). Issues that methodologists will Page 299 → need to consider when creating models of this sort will be discussed later, in the section “Extending 2SPR with Bootstrapped Standard Errors.” Problems Where Both Endogenous Variables Are Dichotomous Finally, a researcher may be dealing with cases where both the economic and conflict variables are measured dichotomously. This would be the case, for instance, if involvement in trade agreements and the emergence of militarized disputes were examined. The estimation of the 2SPR can be extended to the case where both endogenous variables are dichotomous as well (Maddala 1983, 245–47). Again, these solutions employ a limitedinformation equation-by-equation approach and do not require a theory of reciprocal causation. The first stage of estimation follows the previous example where the endogenous explanatory variable was dichotomous. In this case, the reduced-form equation is estimated using probit analysis, and the untransformed predicted z value from this stage is used in place of the variable of concern in the second stage. The dependent variable in the second stage is also dichotomous, and this is also estimated using probit analysis.

While these first steps are straightforward, the rescaling is more complex for both the exogenous and endogenous independent variables, and distinct adjustments are required. Even if this two-stage approach is adopted, and the consistent estimates for the second stage are obtained in a cleanly interpretable manner, the standard errors remain wrong, and their correction is extremely difficult in practice. Extending 2SPR with Bootstrapped Standard Errors Two-stage least squares/instrumental-variable estimation is appropriate when the dependent and explanatory endogenous variables are both continuous, and 2SCML can address the additional complexity caused by having a dichotomous dependent variable with continuous endogenous regressors. Throughout the different conditions, 2SPR is discussed as the logical extension of the 2SLS approach. While parameters require different rescaling transformations for substantive interpretation, the problems in estimating the standard errors have been discussed as the biggest limitation for this estimator. One solution is to use the consistent 2SPR parameter estimates along with bootstrapped standard errors. Researchers who note the problems with complex two-stage estimators and the need to correct the reported standard errors have started to use this approach (see, e.g., Mroz et al. 1999). Page 300 → Bootstrapping is a statistical technique where the sampling distributions for the parameter estimates of interest are simulated through an iterative process (see Mooney and Duval 1993 and Mooney 1996 for useful introductions to the topic). This approach allows for the creation of confidence intervals for statistics where the sampling distributions are unknown or, as in the case of 2SPR, very difficult to estimate. Bootstrapped estimates of standard errors are computationally demanding to produce but do not make the restrictive assumptions common in analytic estimates used for statistical inferencing. Bootstrapping the standard errors in 2SPR will require integrating a number of statistical issues. First, the uncertainty of the first-stage predictions will need to be incorporated into the second-stage bootstrapped results. A possibility may be to incorporate the logic of variable imputation strategies for the first stage along with the bootstrapping of the second to estimate the standard errors (King et al. 2001). While multiple imputation strategies may prove useful for addressing the uncertainty in the first-stage predictions, previous work that has explored the potential of using different data sources for the first and second stages may also provide guidance for developing accurate bootstrapped standard errors for the second stage (see, e.g., Franklin 1990; Gelman, King, and Liu 1998). The Gelman, King and Liu (1998) work on imputation in a hierarchical framework may be particularly useful in this case. In addition to the uncertainty, incorporating the appropriate scaling corrections and dealing with the nature of the measures will need to be handled as well. The statistical foundations that exist in these areas, along with the dramatic increase in computing power, lead to optimism that general solutions to these issues can be developed that will aid substantive investigations. Covariance Structure Models as an Alternative Approach Thus far, this chapter has focused on extensions of traditional regression models (OLS, probit, etc.) to handle the concerns raised when models contain multiple endogenous factors. It is worth briefly noting that a number of other approaches exist as well. Researchers building time-series models will likely incorporate techniques such as vector autoregression (VAR) and examine the nature of interdependence using Granger causality tests. While VAR models are relatively simple to estimate, they are sometimes criticized as being less theoretic, as they focus on forecasting and treat all variables as endogenous (simple introductions to these techniques are found in many texts, e.g., Gujarati 2003). I will leave the debates over these techniques for elsewhere but highlight the issue for scholars who are using time-series data in their research. Page 301 → Even without moving to different types of data, such as time series, other techniques can be used for investigation. One such approach uses structural equation models (SEMs), estimated by examining covariance structures as performed by programs like Lisrel, EQS, and Amos. These models often comprise two parts, a measurement

component that deals with the construction of latent variables and a structural model that examines the relationships among the latent variables and other indicators (Hoyle 1995). These models are likely to be most familiar to political scientists with some interest in political psychology, where these techniques are more common. While these SEMs are designed to explicitly examine systems of equations with potentially complex interrelationships, like simple OLS models they are linear and additive, and the problems found in 2SLS models due to functional forms may apply here as well. Extensions to traditional SEMs incorporate the logic presented earlier in describing one manner that the probit model could be derived. In the probit model, one could consider the dichotomous indicator a blunt measure of the unobserved continuous index function. In SEM extensions, this logic has been adopted for the specification of the latent variables that have been formulated to behave as the index functions that underlie models like probit. Xie (1989) details these extensions, and programs like Mplus (Muthén and Muthén 1998) can handle such operationalizations directly. Thus, researchers trained in structural equation model estimation using programs like Lisrel can also consider this alternative approach to address the concerns of analyzing a system of equations without the traditional limiting assumptions.

Caveats Regarding the Implementation of Corrections The critical issue for addressing the concerns of endogenous explanatory variables that are (or may be) correlated with the error term is that in order to correct the potential problems, adequate instrumental variables are necessary. In order to obtain consistent estimates of model parameters in these cases, we have thus far assumed that enough information that fulfilled fairly strict assumptions existed. This information is provided by instrumental variables that are uncorrelated with the error term. To aid estimation, instrumental variables are needed that are directly related to the endogenous explanatory variable but not the dependent variable. It is the leverage from these variables that allows for the consistent estimation in the instrumental-variable, the two-stage, and even the structural equation modeling approaches. If such information does not exist, unique solutions cannot be obtained. If variables that do Page 302 → not fulfill the necessary assumptions are used in their place, the desirable qualities of our estimators will not hold. Some scholars feel that, given the complexity of interrelationships in political science, true instrumental variables that are uncorrelated with the error terms may be so rare that estimators such as those discussed throughout this chapter may have limited utility in practice.1 Although this chapter has focused on the technical issues involved in creating models with desirable properties under different conditions, even after estimators are developed and widely available, researchers may still find that their research problem is intractable because of the difficulty in adequately identifying their model. While the caveats and difficulties involved in actually correcting the problems when explanatory variables are correlated with the errors in models must be acknowledged, simply ignoring the biases that these problems produce is not a viable solution. Bartels (1991) has demonstrated the consequences of failing to satisfy statistical assumptions. If the variables used to identify the estimation are only “quasi-instrumental,” and approximately uncorrelated with the error, the results of the second-stage/instrumental-variable estimation will not be consistent. While one may use a two-stage estimator and give the appearance of addressing the problems, if the instruments used for identification are correlated with the error, not only is the appearance deceptive, but the “corrected” results could be worse than a model that ignores the endogeneity in the first place. Bartels's conclusions reinforce the point that there are serious consequences to violating the assumptions of a statistical estimator and care is needed for identifying equations. The issue of identification requires not only that the indicators used as instruments are uncorrelated with the error term but also that they are meaningfully related to the endogenous variable for which they are serving as a proxy. Since the instruments are, in essence, standing in for the actual variables that are correlated with the error, variables that are very weakly related to the endogenous explanatory variable will not aid in estimation. In fact, Bound, Jaeger, and Baker (1995) show that, even if the instruments are truly exogenous and uncorrelated with the error, as the R2 in the first-stage estimation goes to zero, the bias in the “corrected” results approaches that of an uncorrected OLS model. Thus weak instruments can lead to inconsistency as well. The fit between the

instrumental variables that are providing the additional information and the original endogenous variable is critical. Bound, Jaeger, and Baker (1995) further show that the weaker the relationships between the instruments Page 303 → and the variable they are standing in for, the larger the consequences for even small amounts of “quasi-instrumentality” and correlation with the error. While the need for variables that are related to the endogenous variable for which they are serving as instruments makes obvious intuitive sense, the logical question is, how strong does this relationship need to be? Bollen and his colleagues provide a rule of thumb that a first-stage R2 below.1 will lead to poor 2SLS or 2SPR results in most cases with continuous endogenous regressors (Bollen 1996; Bollen, Guilkey, and Mroz 1995). Bartels (1991) demonstrates that this is not entirely accurate, though, as he points out that even a high R2 in the first-stage reduced-form equation is not adequate to insure that the instruments are up to the task they are fulfilling. The key to understanding the problems associated with the fit of first-stage estimates can be seen by emphasizing again the role that the instruments play in the reduced-form equation. Recall that the proxies created from the first stage for the endogenous indicator are functions of the sets of exogenous variables in all of the equations in the system. In the models used to illustrate the problem earlier in the chapter, a first-stage reduced-form equation of trade would include the union of variables in the sets of X 1 and X 2 (those directly causally related to conflict and trade, respectively). However, it is not enough to have a reduced-form equation that has strong predictive ability. What is critical is that those exogenous factors in the first stage that are not also in the second stage of the estimation predict well. In the example, this means that one needs to be concerned with how well the variables that constitute X 2, and that are not also in the set of X 1, predict trade. The problem of focusing on the overall model fit of the reduced-form equation can be seen clearly in the extreme case. If the sets of variables in X 1 and X 2 were the same, the predicted value for trade would be a perfect linear function of the exogenous variables in the second stage, and the model of conflict could not be estimated. In this extreme case, not enough information would be available to obtain unique estimates. What is needed are strong predictors of trade in X 2 that are not also in X 1. Bartels (1991) shows that the partial R2 based on these items is more important to determine if there is enough information for stable results, as opposed to rules of thumb based on the overall R2. Standard texts spend a substantial amount of time on the question of identification of the equations, which deals explicitly with the issue of how much information is available to obtain unique results. While this discussion has focused on creating an instrument for the endogenous variable of trade in an equation determining conflict, the same issues arise in the second model in creating a quality measure for the endogenous conflict Page 304 → variable in the second-stage equation for trade. The caveats illustrate the difficulty in implementing some of the techniques described earlier in the chapter in models of economic interdependence and conflict in general. What is needed is a set of variables that can be considered truly uncorrelated with the error in an equation and that are strongly related only to the first-stage dependent variable. Variables that are related to both endogenous factors do not provide additional information to aid in the estimation, and quasi-exogenous and weak indicators are also problematic for the estimation. In addition to other exogenous variables, lagged endogenous variables are often considered useful instruments in these first-stage equations. While such variables are common, one must still be cautious that these are truly uncorrelated with the error terms as well. This section has demonstrated the difficulty in finding quality variables to identify the system. However, it is still worth attempting to overcome the obstacles rather than ignoring the potential bias that results from running standard uncorrected single-equation estimators in models 1 and 2 that disregard the potential correlation between the endogenous explanatory variables and the error term in each equation.

Testing If Corrections for Endogeneity Are Needed As seen in the discussions of the solutions and their caveats, obtaining quality estimates using the various solutions may be extremely difficult. As Bollen, Guilkey, and Mroz (1995, 112) state, it is worth determining if a problem exists and a solution is actually necessary, given “the high price of correcting for endogeneity when it is not present.” Unfortunately, other than fully modeling the system, straightforward statistical tests only exist for continuous endogenous regressors. These approaches build upon the concepts already introduced regarding the reduced-form equations.2

The key to the tests can be linked to the nature of the problem set forth in the first section of the chapter. The reason that we are concerned with endogenous explanatory variables is that they may lead to biased results due to correlation with the error term in the equation being estimated. The tests of whether endogeneity is problematic follow the limited-information approach and question whether it leads to bias in the specific equation being examined. Hausman (1978) developed a test based on the fact that the estimators discussed in the second section of the chapter that address correlation with the error are consistent whether the problem exists or not. In contrast, the uncorrected estimators (standard OLS, probit, logit, etc.) will only be consistent if there is not a problem; they are inconsistent if one exists. The logic of the Hausman test of Page 305 → whether endogeneity is a problem in a specific equation focuses on whether results from the two types of approaches are statistically distinguishable. If they are statistically different, one concludes that correlation with the error is causing significant bias and inconsistency in the traditional estimator, and the solutions discussed in the second section of the chapter are necessary. Like the various two-step solutions, the tests of correlation with the error term are based on the reduced-form equations. As already stated, these tests are only appropriate if the endogenous explanatory variables are continuous. Since the reduced-form equations only use the exogenous variables in the system as regressors, in this case they can be estimated using OLS, since there will be no correlation with the error terms by definition. The predicted values of these reduced-form equations are linear combinations of the exogenous variables and will be uncorrelated with the error terms in the structural model. The residuals are the deviation from these estimates for each observation and constitute the stochastic portion of the model, which includes factors left out of the model, idiosyncratic randomness in behavior, and so on. The decomposition of the reduced-form equations into these two parts serves as the foundation for the tests, as it did for many of the solutions for problems of endogeneity. Examining different statistical texts, it can be seen that there are different ways that the components of the reduced-form equations can be used for conducting the test of whether the uncorrected single-equation results deviate significantly. When all endogenous variables are continuous, the nature of the endogeneity can be tested by revising the structural equation of interest by adding the reduced-form predicted value and residual in place of the actual endogenous measure, including the actual endogenous measure along with the reduced-form predicted value, or including the actual endogenous measure with the reduced-form residual (cf. Gujarati 2003; Greene 2000; Pindyck and Rubinfeld 1998). While the coefficients obviously change with these different specifications, Nakamura and Nakamura (1981) demonstrate that all three of these approaches yield the same F-test when comparing the test specification to the base structural equation of interest. Some confusion exists, though, when individuals simply look at the significance of specific coefficients across the various test specifications. Nakamura and Nakamura (1981) show how each of these approaches and tests can be transformed into the others, but in practice this is not necessary. If there is a single endogenous regressor, the simplest test is to add the residual from its reduced-form equation, since its statistical significance will yield the same results as the F-test comparing the test model to the structural equation itself. If there are multiple endogenous regressors, the residual from each of the relevant reduced-form equations is included, and a Page 306 → joint F-test of whether all are simultaneously indistinguishable from zero will determine whether problems of bias need to be resolved. The Hausman test and the alternative specifications discussed by Nakamura and Nakamura (1981) are appropriate when both the endogenous regressors and the dependent variable in an equation are continuous and can be estimated using OLS. Bollen, Guilkey, and Mroz (1995) extend this test to the case where the dependent variable in the structural equation of interest is discrete but the endogenous regressors are still continuous. This could be the case if one were trying to estimate the equation for conflict (e.g., equation [1a]), measured dichotomously as existence of a militarized dispute or not, and a continuous measure for economic interdependence was available. The test can be conducted in the same way as described earlier. First, OLS is conducted on the reduced-form equation for the endogenous trade variable. The residuals from this equation are then added to the conflict equation, (1a), and the appropriate maximum likelihood estimator, such as probit, is used. If the coefficient on the residual is statistically significant, the hypothesis that there is no bias in the simple single-equation approach is rejected, and correction is necessary. This extension to the Hausman test is closely related to the two-stage conditional maximum likelihood estimator that is appropriate for addressing problems caused by endogenous regressors under the circumstances discussed in the section “Dichotomous Dependent Variable/Continuous

Endogenous Explanatory Variable” (Alvarez and Glasgow 2000; Rivers and Vuong 1988). While the Hausman test and its variations are very useful when the endogenous right-hand-side variables are continuous and their reduced-form equations can be estimated with OLS, no equivalent tests have been developed for the discrete regressor case. Thus, if one were interested in involvement in trade agreements as the measure of economic interdependence, no simple tests would exist. This is true regardless of whether the dependent variable of conflict were itself measured continuously or discretely.

Conclusion In models examining the relationship between economic interdependence and conflict, the fact that measures of both types of factors can be treated as endogenous has troubling negative consequences for the single-equation models that are traditionally used to test our hypotheses. As shown in the first section, whether one theorizes the relationship to be one of reciprocal causation or not, the possibility exists that an endogenous factor included as an explanatory Page 307 → variable will be correlated with the error term and lead to biased estimates of the substantive parameters of interest. If continuous measures of both the economic and conflict factors are available, traditional instrumental-variable /2SLS approaches can address some of the statistical problems involved. Given that common measures often do not fit this measurement scheme, the problem becomes more complex. Approaches such as 2SPR and 2SCML are possible solutions, depending on the nature of measures used for analysis. Extensions to 2SPR that bootstrap the standard errors may prove to be very useful in the future. Also, while the discussion in this chapter has focused on the relationships among continuous and dichotomous variables, obtaining general solutions for ordinal and nominal variables will be fruitful for numerous research problems as well. The technical issues of how to obtain consistent estimators and accurate standard errors are often the focus of discussions dealing with systems of equations. The problems of identification for the actual application of these techniques also deserve serious consideration by political scientists. The need for sets of variables that are uncorrelated with the error terms and directly related to only one of the endogenous variables in a meaningful way may in the long run continue to be a serious hurdle even after the other technical issues have been addressed. Following up on the caveats introduced in this chapter will be useful for those interested in actually implementing these techniques in the future. While technical and practical difficulties exist in addressing the concerns raised by endogenous explanatory variables, simply ignoring the problems they cause is inappropriate. Given interest in understanding how economic interdependence and conflict are linked, as well as how other factors are related to each of these, scholars need to work to minimize the biases that result when basic statistical assumptions are violated. Thus, following the logic of William of Occam, the solution may be to produce more complex models that accurately reflect the true richness of the international system. This chapter points to some issues and possible solutions to move research further down this path.

NOTES 1. This is similar to arguments made by Liu (1960) regarding 2SLS models in economics over forty years ago, although Greene (2000) notes that these criticisms never gained wide acceptance in that discipline. While Liu's (1960) view of problems finding adequate instruments is similar to some arguments currently being made in political science, his solution was to focus on reduced-form equations rather than to ignore the problem itself. Page 308 → 2. The discussions in this chapter focus on modeling systems of equations with cross-sectional and pooled data. Scholars employing time-series analysis have other tests available to them, such as the Granger causality test mentioned earlier.

REFERENCES Achen, Christopher H. 1986. The Statistical Analysis of Quasi-Experiments. Berkeley: University of California Press. Aldrich, John H., and Forrest D. Nelson. 1984. Linear Probability, Logit, and Probit Models. Sage University Paper Series on Quantitative Applications in the Social Sciences, no. 07–045. Newbury Park, CA: Sage. Alvarez, R. Michael, and Garrett Glasgow. 2000. Two-Stage Estimation of Non-Recursive Choice Models. Political Analysis 8:147–65. Bartels, Larry M. 1991. Instrumental and “Quasi-Instrumental” Variables. American Journal of Political Science 35:777–800. Bollen, Kenneth A. 1996. An Alternative Two Stage Least Squares (2SLS) Estimator for Latent Variable Equations. Psychometrika 61:109–21. Bollen, Kenneth A., David K. Guilkey, and Thomas A. Mroz. 1995. Binary Outcomes and Endogenous Explanatory Variables: Tests and Solutions with an Application to the Demand for Contraceptive Use in Tunisia. Demography 32:111–31. Bound, John, David A. Jaeger, and Regina M. Baker. 1995. Problems with Instrumental Variables Estimation When the Correlation between the Instruments and the Endogenous Explanatory Variable Is Weak. Journal of the American Statistical Association 90:443–50. Franklin, Charles H. 1990. Estimation across Data Sets: Two-Stage Auxiliary Instrumental Variables Estimations (2SAIV). In Political Analysis, 1:1–23. Ann Arbor: University of Michigan Press. Gelman, Andrew, Gary King, and Chuanhai Liu. 1998. Not Asked and Not Answered: Multiple Imputation for Multiple Surveys. Journal of the American Statistical Association 93:846–57. Greene, William H. 2000. Econometric Analysis. 4th ed. Upper Saddle River, NJ: Prentice Hall. Gujarati, Damodar. 2003. Basic Econometrics. 4th ed. New York: McGraw-Hill. Hanushek, Eric A., and John E. Jackson. 1977. Statistical Methods for Social Scientists. San Diego: Academic Press. Hausman, J. A. 1978. Specification Tests in Econometrics. Econometrica 46:1251–71. Hoyle, Rick H, ed. 1995. Structural Equation Modeling: Concepts, Issues, and Applications. Thousand Oaks, CA: Sage. King, Gary, James Honaker, Anne Joseph, and Kenneth Scheve. 2001. Analyzing Incomplete Political Science Data: An Alternative Algorithm for Multiple Imputation. American Political Science Review 95:49–69. Page 309 → Liu, Ta-Chung. 1960. Underidentification, Structural Estimation, and Forecasting. Econometrica 28:855–65. Long, J. Scott. 1997. Regression Models for Categorical and Limited Dependent Variables. Thousand Oaks, CA: Sage. Maddala, G. S. 1983. Limited Dependent and Qualitative Variables in Econometrics. Cambridge: Cambridge University Press.

Mooney, Christopher Z. 1996. Bootstrap Statistical Inference: Examples and Evaluations for Political Science. American Journal of Political Science 40:570–602. Mooney, Christopher Z., and Robert D. Duval. 1993. Bootstrapping: A Nonparametric Approach to Statistical Inference. Sage University Paper Series on Quantitative Applications in the Social Sciences, no. 07–095. Newbury Park, CA.: Sage. Mroz, Thomas A., Kenneth A. Bollen, Ilene S. Speizer, and Dominic J. Mancini. 1999. Quality, Accessibility, and Contraceptive Use in Rural Tanzania. Demography 36:23–40. Muthén, Linda K., and Bengt O. Muthén. 1998. Mplus: The Comprehensive Modeling Program for Applied Researchers. Los Angeles: Muthén and Muthén. Nakamura, Alice, and Masao Nakamura. 1981. On the Relationships among Several Specification Error Tests Presented by Durbin, Wu, and Hausman. Econometrica 49:1583–88. Pindyck, Robert S., and Daniel L. Rubinfeld. 1998. Econometric Models and Economic Forecasts. 4th ed. Boston: McGraw-Hill. Rivers, Douglas, and Quang H. Vuong. 1988. Limited Information Estimators and Exogeneity Tests for Simultaneous Probit Models. Journal of Econometrics 39:347–66. Starfield, Anthony M., Karl A. Smith, and Andrew L. Bleloch. 1990. How to Model It: Problem Solving for the Computer Age. New York: McGraw-Hill. Xie, Yu. 1989. Structural Equation Models for Ordinal Variables. Sociological Methods and Research 17:325–52.

Page 310 →

Qualitative Research on Economic Interdependence and Conflict: Overcoming Methodological Hurdles Norrin M. Ripsman and Jean-Marc F. Blanchard With few exceptions, large-N statistical studies of the relationship between trade and conflict have dominated the economic interdependence and conflict literature. These analyses have yielded an array of provocative, albeit often conflicting, correlations. In the process, they have become increasingly sophisticated methodologically and have made great strides to overcome common methodological problems. Conversely, very few qualitative case studies have been performed to investigate how interdependence affects national security decision making. Those qualitative studies that do exist, moreover, are framed in different ways to address a host of different research questions employing different sets of variables; hence, it is difficult to view them together as a coherent whole or to build any cumulative knowledge from them. In addition, they are, as a whole, less self-conscious of the methodological challenges facing students of the interdependence-conflict relationship than the quantitative literature. In an effort to facilitate coordinated case-study research on commercial liberalism, this chapter will identify some of the methodological problems facing qualitative researchers when studying the economic interdependenceconflict relationship. In particular, we consider the challenge of conceptualizing and operationalizing the independent and dependent variables for use in case studies, the identification of crucial cases, and the determination of key decision makers. We offer our own answers to each of these questions, which we expand upon briefly below. In each case, we argue that the key to solving these methodological Page 311 → problems is to be aware of the comparative advantages and disadvantages of the case study method in order to avoid the pitfalls of conducting a case study as if it were a quantitative study. We begin with an overview of the existing qualitative research on economic interdependence and conflict in order to demonstrate the need for a more coherent research strategy.

The Existing Literature As the reference lists in this volume will attest, although the literature on economic interdependence and conflict is growing rapidly, qualitative research makes up only a very small portion of it. Not only are qualitative studies few and far between, but they are framed differently, with different variables and at different levels of analysis. Only two studies directly address the economic exchange-conflict relationship with few or no intervening or complicating variables. Focusing on the international system as a whole, Richard Rosecrance's (1986) study examines the possibility that, in the contemporary era, the growth of international trade might reduce the incidence of international conflict and foster cooperation. Our own research (Ripsman and Blanchard 1996–97) similarly seeks to isolate the effect of economic interdependence on national security decisions. The approaches taken in these two studies, however, vary considerably. Rosecrance assesses the systemic impact of “a new ‘trading world' of international relations” (1986, ix) across a broad sweep of history since the peace of Westphalia, with brief consideration of specific examples; we engage in two focused case studies that examine the interplay at the monadic/dyadic level between economic interdependence and the willingness of states to run the risks of war during crises. In addition, while Rosecrance conceptualizes interdependence as a host of economic ties, we focus primarily on vulnerability interdependence, defined in terms of a state's access to strategic materials essential for the maintenance of national security and prosperity. Anne Uchitel (1993) frames her study somewhat differently. Instead of inquiring about the effects of interdependence on conflict, she conducts case studies to assess the differing effects of different types of economic interdependence. Focusing on the monadic level, she concludes that dependence on the importation of strategic materials encourages great powers to behave aggressively, while dependence on access to export markets will have a moderating influence on states.

Other researchers investigate the effect of economic interdependence within a broader theoretical context, incorporating additional variables that may Page 312 → complicate the interdependence-conflict relationship. Paul Papayoanou's research (1996; Papayoanou and Kastner 1999), for example, studies whether the international balance of power and a state's domestic regime-type alter the effect economic interdependence has on decisions to go to war or not. Peter Liberman (1999–2000) investigates the interplay between the offense-defense balance and the effect that economic interdependence has on states' willingness to wage war. Dale C. Copeland's research (1996, 1999–2000) focuses on still another question: whether a state's expected future gains from trade affect the relationship between interdependence and the likelihood of war. Finally, two studies approach the commercial liberal argument from a different perspective, concentrating on the economic strategies and ideologies that states follow as the independent variable, rather than economic interdependence. Etel Solingen (1998) analyzes the impact of foreign economic policies on regional conflict, concluding that regions dominated by internationalist, liberal-oriented coalitions tend to be more peaceful than those dominated by statist-nationalist or confessional coalitions. In a similar vein, Jack S. Levy and Salvatore Ali (1998) consider the role of mercantilist policies in fueling the seventeenth-century Anglo-Dutch rivalry. While each of these studies asks worthwhile questions, it is difficult to draw any meaningful and reliable conclusions from them as a whole. We believe that this absence of a coherent, comprehensive qualitative literature on commercial liberalism stems from the myriad of methodological challenges and choices that qualitative researchers face when they embark on studies of interdependence and conflict and the lack of a clear blueprint for such research. To overcome this problem, the rest of this chapter will explore the methodological challenges and suggest a strategy for overcoming them in an attempt to build cumulative knowledge.

The Independent Variable At the outset, researchers confront key problems when defining and operationalizing an independent variable for a case study of commercial liberalism. In the first instance, they must choose whether to focus on openness or the economic opportunity costs of conflict (Stein 1993, 258–59; McMillan 1997; Mansfield and Pollins this volume). Since the classic definition of economic interdependence conceptualizes it as “ties that are costly to break” (Waltz 1970; Keohane and Nye 1977), we believe analysts should focus on the opportunity costs of conflict, which correspond more closely. Moreover, this conceptualization would allow researchers to test the more powerful claims of the commercial Page 313 → liberal argument—what Arthur Stein (1993) refers to as the binding argument, emphasizing that economic interdependence restrains states from using force to resolve disputes even in the presence of a casus belli—which stresses the constraint posed by potential economic losses. Subsequently, they must decide how to measure these opportunity costs. This critical task presents researchers with a number of important methodological quandaries that we detail. First, analysts must decide what indexes to use to gauge the level of economic interdependence. While quantitative studies have the advantage of comparing the extent of bilateral interdependence based on the magnitude of a predetermined ratio (such as bilateral trade as a percentage of national gross domestic product [GDP]), a researcher studying one or two cases cannot do so, since he or she could not meaningfully correlate small differences in the level of interdependence to changes in the dependent variable.1 Instead, case studies are more appropriately based upon categorical differences in the level of interdependence (e.g., highly dependent, moderately dependent, and minimally dependent) to ensure variation that can credibly explain variations in conflict patterns across a few cases. These categorical measures could certainly be based upon such a ratio, but that would squander the comparative advantage of qualitative research to employ measures that capture more of the context and complexity of a bilateral economic relationship than the indices that large-N studies must employ in the interests of efficiency. Given the lack of adequate measures for use in case study research, we devised benchmark measures of both vulnerability and sensitivity interdependence (Keohane and Nye 1977). Our strategic goods test (SGT) evaluates the vulnerability of states to a trade disruption by examining the importance of the goods traded for each state's economy and national security establishment, as well as the potential for each country to mitigate the costs of a cut-off by considering the availability of alternative suppliers, the prospects of increasing domestic production, the potential for substitution, and the prospects for conservation (Blanchard and Ripsman 1996). Such a measure is

especially important since existing studies of economic interdependence focus primarily on the strategically less significant sensitivity interdependence.2 We recognize, though, that researchers also need a measure of sensitivity since it, too, might be politically consequential. After all, while states should be more likely to alter their national security policy if they are threatened with a disruption of critical, defense-related materials, they may also do so in response to anticipated market effects that would affect GDP. Hence, we devised our contextual sensitivity estimator (CSE). The CSE not only examines the nature of trade relations between states but also looks beyond trade to Page 314 → examine sensitivity in the realm of investment and monetary relations. Furthermore, it considers the context of economic exchange by examining the importance of the goods traded, the stability of both the currency and the banking systems, and other factors that affect the importance of a given bilateral economic relationship (Blanchard and Ripsman 2001; cf. Uchitel 1993; Crawford 1993; Gartzke, Li, and Boehmer 2001). Our qualitative measures allow researchers to take advantage of the case study method's ability to study the context of economic relations and, therefore, determine whether or not they are meaningful to the states involved. Second, like statistically oriented scholars, qualitative researchers also must decide whether to measure interdependence at the level of the system, the dyad, or the state (Tetreault 1980; Stein 1993). Since the logic of commercial liberalism is that the opportunity costs of violent conflict in terms of economic gains forgone will make war unreasonable (see Buzan 1984; Keohane 1990, 177–79; Blanchard, Mansfield, and Ripsman 1999–2000), we argue that a bilateral measure is more appropriate because it focuses exclusively on the specific costs that states will incur in the event of war with a specific partner. Since states run the risk of disruptions with more than simply a single adversary, however, we advocate a modified dyadic measure, which takes into account economic ties between two states and their respective allies, or likely allies. Moreover, in light of our view that decision makers would be far more concerned with the constraints they face than the constraints others face, we recommend that researchers focus on each side's dependence on the other and its allies rather than a simple measure of the dyad's level of interdependence (Blanchard and Ripsman 1996).3 A third independent variable problem concerns whether to focus on objective or subjective interdependence. While our objective measures of interdependence allow us to judge the costs that would actually accrue to states in the event of a disruption of normal economic relations, they do not, on their own, tell us whether national leaders are aware of these costs. Nor do they tell us whether leaders expect the benefits of the bilateral economic relationship to continue, which Copeland (1996, 1999–2000) argues is a more powerful determinant of a state's willingness to use force. Since subjective perceptions can be at least as important as objective realities in international politics (Jervis 1976; Wendt 1992; Finnemore 1996; Checkel 1998), in our own study (Ripsman and Blanchard 1996–97) we supplemented our measurement of interdependence with a primary- and secondary-source analysis of whether or not leaders understood the extent to which they were dependent.4 Again, this is something that would be impracticable for large-N studies but is imperative for case studies Page 315 → to ensure that observed results are not caused by faulty perceptions of objective conditions.

The Dependent Variable Another important choice that researchers have to make in qualitative studies of commercial liberalism is what dependent variable best represents the claims of the binding commercial liberal argument. One could, perhaps, simply record whether the states in the study go to war with each other or not. This would be a poor selection, however, since such a research design would undermine the comparative advantage of case studies—their ability to shed light on the decision-making process and, therefore, on how the independent variable affects outcomes. Furthermore, since the war or peace outcome would already be known at the time of case selection, the choice of cases would bias the study. We recommend that researchers focus on the willingness of decision makers to run the risks of war in moments of heightened crisis. This emphasis allows researchers to engage in a process-tracing exercise, tracing the effect that interdependence has on decision making, irrespective of the outcome of the crisis (George and McKeown 1985; George 1997; Van Evera 1997, 58–67). Thus, even in the event of war, if the constraints of economic

interdependence moderated state behavior or reduced the government's willingness to take risks, the case might provide support for commercial liberalism. Conversely, in a crisis where war was averted, if interdependence played no significant role in national security decision making and the state's willingness to take risks, then a peaceful outcome could not be used as support for commercial liberalism (Ripsman and Blanchard 1996–97).

Which Cases Should We Choose? Since it is extremely time-consuming to conduct thorough case studies that explore the entirety of a bilateral economic relationship as well as the available primary and secondary sources on decision making, qualitative researchers need to exercise judgement in their choice of appropriate cases to study. There are, in fact, four important case-selection decisions that researchers must face: the scope of the study, the era, the nature of the states involved, and the level of variation in the independent variable. Should we study relations between pairs of states over a sweep of history or focus on national decision making during a relatively narrow episode, such as a political/military crisis? This important choice should fit the specific claims Page 316 → being tested. Since the most powerful claim of commercial liberalism is that the potential costs of disrupting economic interdependence restrain states from using force when they would otherwise do so for strategic reasons (Stein 1993), we argue that the best test of the systemic logic of commercial liberalism is an analysis of crisis decision making, when the threat of war is real. Indeed, the existence of a crisis or a casus belli would be an appropriately difficult test for the theory (Eckstein 1975; Rogowski 1995) because the incentives to accept the costs of using force are present; thus we can test the more powerful claims of the commercial liberal argument. Conversely, in normal times (absent a crisis) the unwillingness of leaders to run the risks of war would be overdetermined and, therefore, difficult to attribute to the influence of economic interdependence. Furthermore, it would be difficult to isolate the impact of economic interdependence on national security decisions over a broad sweep of history since we would not be able to scrutinize the documents on every policy decision as carefully over a multiyear period as easily as we could during a two- to three-week crisis. In a crisis situation, we could utilize the process-tracing method most effectively to isolate the effect of economic considerations on national security decisions. Finally, during the two- to three-week crisis period, the level of at-risk economic interdependence is not likely to fluctuate as much as it could over a longer time horizon. The second decision is whether to choose recent cases or historical case studies. Theoretically, the logic of commercial liberalism should operate whenever states are highly interdependent. Thus, provided that adequate sources are available, the historical period should matter little. As a practical matter, the advantage of pre–World War II case studies—at least where states in the Western world are involved—is that there are sufficient documentary materials, memoirs, and secondary sources available to give the researcher a window into decision making, which facilitates process tracing (Ripsman and Blanchard 2000). Moreover, such cases allow the researcher to avoid the peculiarities of the Cold War and post–Cold War era, which might have complicated decision making. Specifically, in the bipolar Cold War world, trade flowed primarily between allies, who were themselves restrained from conflict by the superpowers, the fear of a common enemy, and the existence of nuclear weapons (Buzan 1984; Gowa 1995). After the Cold War, American hegemony has acted as a break on violent conflict in much of the world (Kupchan 1998). Thus, since World War II the absence of violent conflict between interdependent states is overdetermined. For these reasons, we chose pre–World War II cases for our initial study. There is, however, a compelling reason to study more recent cases. Rosecrance Page 317 → (1986) argues that in the contemporary era leaders must be more concerned about providing economic benefits to their constituents than ever before. Consequently, Rosecrance's variant of commercial liberalism, emphasizing “trading states,” has a distinctly different causal mechanism that must be investigated differently. If he is correct, then we should expect that economic interdependence has had a greater impact on national security decision making since World War II than before.5 Especially in light of our initial studies, which showed no relationship between economic interdependence and leaders' willingness to run the risks of war in the July Crisis of 1914 and the Rhineland Crisis of 1936 (Ripsman and Blanchard 1996–97), it would be appropriate for researchers to conduct more contemporary case studies to investigate Rose-crance's claim.

Regarding the dyads to choose, researchers can select great power dyads, smaller power dyads, or mixed dyads. As the introduction to this volume, by Edward D. Mansfield and Brian M. Pollins, points out, the existing qualitative research has a very heavy great power bias (Uchitel 1993; Copeland 1996, 1999–2000; Papayoanou 1996; Ripsman and Blanchard 1996–97; Levy and Ali 1998). This is reasonable since, if commercial liberalism is to be judged a powerful theory of international politics, it should explain the behavior of the most important international actors. Nonetheless, the logic of commercial liberalism should apply to the smaller powers, which tend to be more dependent on international economic exchange (Katzenstein 1985), to an even greater extent than the more independent great powers. Therefore, we agree with Mansfield and Pollins that researchers should supplement great power case studies with those of smaller powers. A final case-selection problem relates to the extent to which researchers want their independent variable to vary. To ensure that there is adequate variation, researchers could choose cases with wide ranges of variation, from states that are independent of their adversary to those that are highly dependent. In this manner, comparison is possible between the willingness of states at both ends of the spectrum to take risks, provided that other aspects of the interaction are held constant. Once again, however, this approach is best suited to a large -N analysis that seeks to determine if, on average, states with higher levels of interdependence are less likely to enter into violent conflicts. Not only is it too constraining for researchers to find cases that are alike in all respects except for the level of interdependence, but such conclusions on the basis of a handful or fewer cases are also not meaningful unless the researcher is analyzing hard test cases. Instead, since the comparative advantage of case studies is their ability to Page 318 → trace the effect of the independent variable at multiple stages of a decision, we recommend a focus on states that, according to our benchmark measures, are meaningfully interdependent. In our own research (Ripsman and Blanchard 1996–97), we selected cases of crisis decision making between interdependent states (in one case vulnerable and in the other case sensitive) to see whether leaders were aware of the economic costs of conflict and whether that affected their willingness to accept the risks of war. We felt this was an appropriate choice, since, in the absence of interdependence, a process-tracing approach would tell us little about how economic interdependence shapes crisis decision making. Of course, this does not allow for a comparison with decision making in cases where interdependence is insignificant, but that task is not well suited to such a small-N qualitative research design.

Which Actors Should We Study? Once we have decided upon the variables and case(s) that we wish to study, we must bound our study by determining which national actors to focus on. Should we cast our nets broadly or narrowly? Given our interests in the effect of economic interdependence on national security decision making, we obviously need to consider the statements and perceptions of the chief of government and key actors charged with foreign affairs and defense, such as the foreign minister, the minister of defense, the chief of the general staff, and any other actors with privileged positions in crisis decision making. For if the economic opportunity costs of the use of force exert a restraining influence on states, it is these key decision makers of the foreign security policy executive (Ripsman 2002) who would have to be restrained and persuaded by the logic of commercial liberalism. But what about other governmental and societal actors, such as parliament, agencies having economic responsibilities (e.g., finance ministries), the banks, business elites, and workers in export sectors, all of which might encourage caution due to the potential economic costs of conflict? Indeed, it is precisely these actors who are likely to feel the pinch of interdependence the most, as the banking and commercial sectors would bear the brunt of an economic disruption and parliament and the bureaucracy would be the most direct access point for commercial and financial entities to influence government policy. American business groups lobbying members of Congress and the Commerce and Agriculture Departments every time permanent normal trading relations for China were up for renewal shows this dynamic in action. We contend that it is not necessary to delve into the attitudes of these other Page 319 → interested societal actors, since if their attitudes are influential, then their arguments should be reflected in discussions taking place at higher governmental levels and there should be evidence that these economic considerations swayed the foreign security policy executive. If not, then the mere presence of societal concerns in parliament, the press, or the business

community does not constitute evidence in support of the causal logic of commercial liberalism. Thus, we advocate treating the key governmental actors, who actually make the decisions, as gatekeepers for other societal interests and ideas. Others, such as Papayoanou (1996), however, disagree and maintain that the normal causal path of commercial liberalism is through societal economic and political interest groups and, therefore, that we must study the attitudes of such groups. We are skeptical, however, because business interests generally tend to be divided (Skidmore-Hess 1996); represent only one of the constituencies that the government must satisfy; and, most importantly, may not have either sufficient access or influence over national security decision makers. Nonetheless, this is an important area of disagreement, which can potentially affect whether the study overestimates (if these groups are included) or underestimates (if their preferences are excluded) the influence of interdependence.

Final Observations In this chapter, we have outlined some of the major methodological challenges that qualitative researchers face when they conduct case studies of the commercial liberal argument. Some of them, like the problems of conceptualizing and operationalizing the study's variables, are hurdles that all researchers must confront, regardless of their methodological approach, but that must be resolved differently by quantitative and qualitative scholars. Others, such as the choice of cases and the actors to concentrate on, are unique to qualitative research. We have also suggested our own answers to these problems and explained our choices. In each case, our choices reflect not only the specific theoretical claims being tested but also the specific comparative advantages and disadvantages of the case study method. We should not expect case studies to supplant statistical methods as a means of identifying correlations between independent and dependent variables or in allowing us to generalize freely from our sample to other cases. They generally are not well suited to this task because they cannot yield a sufficient number of observations. Instead, we contend that we should design case studies so that we can profit from their greater precision and Page 320 → rigor as well as their ability to get inside the black box of the state to link the independent and dependent variables and establish causation in particular cases (Russett 1970; George and McKeown 1985; Yin 1989; Tarrow 1995, 472–73). Our recommendations are an attempt to exploit this advantage. We recognize, of course, that other answers are possible. Nonetheless, our approach represents a coherent approach to testing the commercial liberal theory and, moreover, can provide the blueprint for a collaborative series of case studies on commercial liberalism, which we believe is essential to advance the state of qualitative research in this area and build a coherent body of qualitative research. In this manner, case study research will complement the increasingly sophisticated quantitative research on the commercial liberal argument.

NOTES 1. Of course, as we argue elsewhere, these ratios themselves may not be good indicators of interdependence (Blanchard and Ripsman 1996, 2001). 2. Most analysts represent interdependence with reference to the ratio of trade or imports as a percentage of GDP (Mansfield and Pevehouse 2000; Oneal and Russett 1997, 1999; Barbieri 1996; de Vries 1990). This ratio does not consider the importance of the goods traded to the state (the material context) or the availability of other sources of supply, including domestic production, conservation, or substitution. Therefore, it represents measures of sensitivity, rather than the long-run costs that would accrue to states if the trading relations were to be disrupted (Blanchard and Ripsman 1996, 2001). 3. A valid counterargument could be made in favor of measuring bilateral dependence in order to consider the effects of both symmetrical and asymmetrical dependence—that is, situations in which both countries are comparably dependent on each other and cases in which one state is considerably more dependent (Keohane and Nye 1977; McMillan 1997). Symmetric interdependence might affect political dynamics by

leading a state to play a brinkmanship game with its economic partner, which it might conclude is as reluctant as it is to cut bilateral economic ties. Conversely, asymmetric interdependence might lead a state to capitulate since it would have to contemplate economic losses while its economic partner faced no similar constraints. We agree that it could be fruitful to investigate this possibility in future case research. 4. We did not, however, focus on leaders' future expectations for the simple reason that most leaders are very sensitive to short-term costs while they are less motivated by uncertain long-run benefits. 5. We should note, though, that the causal argument that Rosecrance is making is different from the commercial liberal claim. He does not argue that economic interdependence necessarily constrains states from using force, but that certain types of states—modern states—will be especially attuned to the costs of disrupting economic relationships. Page 321 →

REFERENCES Barbieri, Katherine. 1996. Economic Interdependence: A Path to Peace or a Source of Interstate Conflict? Journal of Peace Research 33:29–49. Blanchard, Jean-Marc F., Edward D. Mansfield, and Norrin M. Ripsman. 1999–2000. The Political Economy of National Security: Economic Statecraft, Interdependence, and International Conflict. Security Studies 9:1–15. Blanchard, Jean-Marc F., and Norrin M. Ripsman. 1996. Measuring Vulnerability Interdependence: A Geopolitical Approach. Geopolitics 1:225–46. ______. 2001. Rethinking Sensitivity Interdependence: Assessing Trade, Financial, and Monetary Linkages between States. International Interactions. 27:95–127. Buzan, Barry. 1984. Economic Structure and International Security: The Limits of the Liberal Case. International Organization 38:597–624. Checkel, Jeffrey T. 1998. The Constructivist Turn in International Relations Theory. World Politics 50:324–48. Copeland, Dale C. 1996. Economic Interdependence and War: A Theory of Trade Expectations. International Security 20:5–41. ______. 1999–2000. Trade Expectations and the Outbreak of Peace: Détente, 1970–1974, and the End of the Cold War, 1985–1991. Security Studies 9:15–58. Crawford, Beverly. 1993. Economic Vulnerability in International Relations: The Case of East-West Trade, Investment, and Finance. New York: Columbia University Press. de Vries, Michael S. 1990. Interdependence, Cooperation, and Conflict: An Empirical Analysis. Journal of Peace Research 27:429–44. Eckstein, Harry. 1975. Case Study and Theory in Political Science. In Handbook of Political Science, ed. Fred I. Greenstein and Nelson Polsby, 7:79–138. Reading, MA: Addison–Wesley. Finnemore, Martha. 1996. National Interests in International Society. Ithaca: Cornell University Press. Gartzke, Eric, Quan Li, and Charles Boehmer. 2001. Investing in the Peace: Economic Interdependence and International Conflict. International Organization 55:391– 438. George, Alexander L. 1997. Knowledge for Statecraft: The Challenge for Political Science and History. International Security 22:44–52.

George, Alexander L., and Timothy J. McKeown. 1985. Case Studies and Theories of Organizational Decision Making. In Advances in Information Processing in Organizations, ed. Robert Coulam and Richard Smith, 43–68. Greenwich, CT: JAI Press. Gowa, Joanne. 1995. Allies, Adversaries, and International Trade. Princeton: Princeton University Press. Jervis, Robert. 1976. Perception and Misperception in International Politics. Princeton: Princeton University Press. Katzenstein, Peter J. 1985. Small States in World Markets: Industrial Policy in Europe. Ithaca: Cornell University Press. Page 322 → Keohane, Robert O. 1990. International Liberalism Revisited. In The Economic Limits to Modern Politics, ed. John Dunn, 165–94. Cambridge: Cambridge University Press. Keohane, Robert O., and Joseph S. Nye Jr. 1977. Power and Interdependence: World Politics in Transition. Boston: Little, Brown. Kupchan, Charles A. 1998. After Pax Americana: Benign Power, Regional Integration, and the Sources of a Stable Multipolarity. International Security 23:40–79. Levy, Jack S., and Salvatore Ali. 1998. From Commercial Competition to Strategic Rivalry to War: The Evolution of the Anglo-Dutch Rivalry, 1609–1652. In The Dynamics of Enduring Rivalries, ed. Paul F. Diehl, 29–63. Urbana/Champaign: University of Illinois Press. Liberman, Peter. 1999–2000. The Offense-Defense Balance, Interdependence, and War. Security Studies 9:59–91. Mansfield, Edward D., and Jon C. Pevehouse. 2000. Trade Blocs, Trade Flows, and International Conflict. International Organization 54:775–808. McMillan, Susan M. 1997. Interdependence and Conflict. Mershon International Studies Review 41:33–58. Oneal, John R., and Bruce M. Russett. 1997. The Classical Liberals Were Right: Democracy, Interdependence, and Conflict, 1950–85. International Studies Quarterly 41:267–94. ______. 1999. The Kantian Peace: The Pacific Benefits of Democracy, Interdependence, and International Organization, 1885–1992. World Politics 52:1–37. Papayoanou, Paul. 1996. Interdependence, Institutions, and the Balance of Power: Britain, Germany, and World War I. International Security 20:42–76. Papayoanou, Paul, and Scott Kastner. 1999–2000. Sleeping with the (Potential) Enemy:Assessing the U.S. Policy of Engagement with China. Security Studies 9:157–87. Ripsman, Norrin M. 2002. Peacemaking by Democracies: Domestic Structure, Executive Autonomy and the Post–World War Settlements. University Park: Pennsylvania State University Press. Ripsman, Norrin M., and Jean-Marc F. Blanchard. 1996–97. Commercial Liberalism under Fire: Evidence from 1914 and 1936. Security Studies 6:4–50. ______. 2000. Contextual Information and the Study of Trade and Conflict: The Utility of an Interdisciplinary Approach. In Beyond Boundaries? Disciplines, Paradigms, and Theoretical Integration in International Studies, ed. Rudra Sil and Eileen M. Doherty, 57–85. Albany: SUNY Press.

Rogowski, Ronald. 1995. The Role of Theory and Anomaly in Social-Scientific Inference. American Political Science Review 89:467–70. Rosecrance, Richard. 1986. The Rise of the Trading State: Commerce and Conquest in the Modern World. New York: Basic Books. Russett, Bruce M. 1970. International Behavior Research: Case Studies and Cumulation. In Approaches to the Study of Political Science, ed. Michael Haas and Henry S. Kariel, 425–53. San Francisco: Chandler. Page 323 → Skidmore-Hess, Thomas. 1996. Business Conflict and Theories of the State. In Business and the State in International Relations, ed. Ronald W. Cox, 199–216. Boulder: West-view Press. Solingen, Etel. 1998. Regional Orders at Century's Dawn: Global and Domestic Influences on Grand Strategy. Princeton: Princeton University Press. Stein, Arthur A. 1993. Governments, Economic Interdependence, and International Cooperation. In Behavior, Society, and International Conflict, ed. Philip Tetlock, Jo Husbands, Robert Jervis, Paul Stern, and Charles Tilly, 241–324. New York: Oxford University Press, for the National Research Council of the National Academy of Sciences. Tarrow, Sidney. 1995. Bridging the Quantitative-Qualitative Divide in Political Science. American Political Science Review 89:471–74. Tetreault, Mary Ann. 1980. Measuring Interdependence. International Organization 34:429–43. Uchitel, Anne. 1993. Interdependence and Instability. In Coping with Complexity in the International System, ed. Jack Snyder and Robert Jervis, 243–64. Boulder: Westview Press. Van Evera, Stephen. 1997. A Guide to Methodology for Students of Political Science. Ithaca: Cornell University Press. Waltz, Kenneth N. 1970. The Myth of National Interdependence. In The International Corporation, ed. Charles P. Kindleberger, 205–23. Cambridge, MA: MIT Press. Wendt, Alexander. 1992. Anarchy Is What States Make of It: The Social Construction of Power Politics. International Organization 46:391–426. Yin, Robert K. 1989. Case Study Research: Design and Methods. Rev. ed. Beverly Hills: Sage.

Page 324 →

Computer Simulations of International Trade and Conflict David H. Bearce and Eric O'N. Fisher There is a vast literature on the relationship between trade and military conflict. But in recent years, the empirical work on this subject has seen a decline in the use of computer simulations and a growing reliance on large-N statistical analysis.1 In this chapter, we briefly examine the broad utility of computer simulations and then introduce an agent-based model that explicitly considers the effect of economic geography. Our definition of economic geography encompasses both physical geography and the economic infrastructure that facilitates trading. Different regions in the international system have idiosyncratic economic geographies and disparately shaped trading networks. These differences have to do with both natural features of geography and technology. The economic geography in Western Europe, for example, is relatively compact as compared with the distended geography in South America. The invention of the railroad in the nineteenth century brought the markets of Continental Europe even closer together, but it did very little to make trade between Chile and Peru easier. An important feature of Bearce and Fisher's (2002) model is that economic geography has an ambiguous effect on the relationship between trade and war. Economic closeness is likely to facilitate international trade, but it can also make military conquest easier to achieve. This understanding may help explain the historically high levels of both trade and war in Western Europe. What is the value of employing computer simulations when there is so much existing data on bilateral trade and conflict? Despite the vast number of annual observations, we should not treat these data as truly independent. These data are not independent arguably because of historical contingency. History Page 325 → has unfolded along one particular path; some trade relationships were established, allowing economic exchange to flourish, while other potential trade relationships never emerged, perhaps due to a distended economic geography. If the initial economic geographies and concomitant trade networks had been otherwise, history could have developed along different paths. Likewise, historical outcomes, especially the occurrence and winners of wars, can be treated as random (Fearon 1995, 386–87; Gartzke 1999). Consequently, we have much fewer real observations of trade and war than is often acknowledged, and it may be worthwhile to examine counterfactual histories within a tightly structured analytical framework. Computer simulation allows us to rerun “history” many times. If one believes that history is a random process, then one must take seriously the possibility that even with the same set of initial conditions there may be many possible future worlds. Furthermore, small changes in initial conditions may lead to entirely different historical paths. This insight underlies the new interest among political scientists in counterfactual analysis (Tetlock and Belkin 1996). Had a road through the mountains separating two states been constructed, bilateral trade might have developed more quickly, perhaps contributing to a reduction in military tensions in the region. Alternatively, this same mountain road could have facilitated troop movements, making military conquest and the subsequent extraction of resources a more attractive option than voluntary economic exchange. One can perhaps best explore these possibilities by computer simulations.

Simulations in the Social Sciences It is rare for a formal model in the social sciences to have an analytical solution. In its most elegant form, a formal model reduces to a system of equations, and an equilibrium is just one of its solutions. For example, Walras (1874) used this level of mathematical abstraction to describe economic general equilibrium, and it remains a computationally daunting task to solve for market-clearing prices almost a century and a half later. In recent years, there has been a growing movement in macroeconomics using model-based computer simulations as a way of interpreting economic data. This trend gained momentum in part because of the failures of an earlier generation of atheoretical statistical studies to explain important changes in the U.S. economy that occurred during the 1970s and continue until this day.2 Although there is a continuing debate in economics concerning how best to

conduct simulations and interpret them,3 it is perhaps safe to say that there is Page 326 → widespread agreement among economists that formal models provide the discipline necessary for scientific advancement.4 Several computer models of international trade and war have been created. However, few of these models have become benchmarks for further research by other scholars since every model is designed to answer a specific research question. Pollins (1982) reviewed the early trade-simulation models. More recently, scholars have modeled war in the international system. Some impressive examples include Cederman 1997; Cusack and Stoll 1990; and Bremer and Mihalka 1977. Relatively few such computer models include both trade and war; although Epstein and Axtell (1996) built a notable recent exception. Kydland and Prescott (1996) identify the five key steps in building and simulating a model, calling this activity a computational experiment. The first step is to pose a research question. In our case, we explore how economic geography affects the relationship between trade and war. Second, it is important to build the model from welldeveloped theory. The model that we develop is based firmly upon economic theory and the best current theory of interstate conflict. Third, one actually constructs a model—really just a set of equations— describing a social-scientific system. We have tried assiduously to develop the simplest possible model to analyze the question at hand. It has two elements: a model of economic exchange and a model of conflict. The general model is an applied theoretical analysis of the economics of trading networks.5 Our particular model of exchange draws heavily upon a key feature of Jodhimani's (1999) work: the trade-off between the fixed costs of setting up and maintaining trade links and the benefits of increased diversity in consumption. Our specific model of conflict applies the insights of formal theorists like Fearon (1995) and Morrow (1989), who argue that war tends to result from a breakdown of interstate bargaining due to incomplete information. The fourth step is the calibration of the model. This activity entails choosing values for the key parameters. Finally, the fifth step is to run (and rerun) the computational experiment. Since most model-based outcomes are random processes, one must repeat the experiment a number of times without varying the model parameters. Once stable results emerge, parameter values should be examined for robustness of outcomes. Thus one often presents a range of results across different parametric specifications of the model. We now turn our attention to a description of the particular agent-based model of trade and military conflict that we have built and simulated. We will examine a computational experiment exploring the effect of different initial economic geographies. Page 327 →

Static Elements of the Model The presentation of the model in this section will be discursive, avoiding mathematical expressions entirely. The interested reader is urged to consult Bearce and Fisher (2002) for a fuller exposition of the model and its analytical details. Also, its exact formulation can be recovered from the computer programs used for the simulations. These are available at . Our model is a world consisting of several different state actors, each endowed with a unique good. A state's endowment depends upon time because it may suffer from the vagaries of war. Having a state-specific composite commodity captures the idea that each economy has comparative advantage in producing some good for international exchange. Each state has preferences summarized by a simple utility function whose rule serves as the foundation for an expected utility indicator. Each state is risk neutral and values every good symmetrically. We use a measure known to economists as the elasticity of substitution to capture the concept of “revisionism” in international relations theory (Morganthau 1948; Kissinger 1957; Carr 1946). This measure summarizes a state's desire to acquire the resources of other states. A diverse resource base, once acquired, translates into greater military power

for a revisionist state (Schweller and Preiss 1997). One can also think of this “utility” function as an aggregate production function; each state has its own distinct ability to turn a resource base of intermediate inputs into a homogenous final output called gross domestic product (GDP). Thus a highly revisionist state will have a high level of material well-being if it secures a diverse array of resources used in national production. A trading network is a list of sets whose union contains the set of all the states. This collection of sets satisfies two consistency properties. First, every state is its own trading partner. Second, being a trading partner is reflexive; thus, if one state trades with another, then the latter trades with the former. Two simple examples are autarky and a completely integrated world economy. Another plausible example is the world being broken into two trading blocs, with no trade between them. Maintaining commercial relations with another state requires paying a history-dependent fixed cost based upon the current economic distance between trading partners. The notion of economic geographic distance also has two simple properties. First, the distance from a state to itself is zero. Second, distance is also reflexive, although we do not impose the triangle inequality. Consider, Page 328 → for example, the three states Algeria, England, and France. It is entirely plausible that the sum of the economic distances from England to France and then from France to its former colony is less than the economic geographic distance between England and Algeria itself. We follow the literatures in international trade and economic geography and impose that economic distance entails iceberg costs à la Samuelson (1954). This analytical artifice models transport costs as a fraction of real resources that melts away during the process of trade. In our model, maintaining a link in the trading system costs a fraction of a state's resources, and this fraction depends upon the current economic distance inherent in that link. The initial economic geography is given, and economic distance evolves according to a simple rule that depends upon two factors. First, there is a parameter capturing the decrease in trading costs if a commercial link between two states is extant. This term measures the rate of learning by trading. Second, there is another parameter for the rate of exogenous technological progress that makes the global economic geography more compact in every period even if two states are not trading. We think of this parameter as capturing the process of globalization; the distance between every pair of states in the world may shrink, even if there is no current trade link between them. The real cost that a state pays to maintain its trading links is described by the total current cost of its trading links. It loses a fraction of its endowment to pay this cost, and the rest is available for trade. The most important aspect of this formulation is that bilateral trade costs become lower if countries establish a link, and they do not return to higher levels if that link is temporarily abandoned. The economic equilibrium in the world economy is a “price vector” and corresponding list of allocations such that each state maximizes utility (or the value of national output), taking its own terms of trade, the current economic geography, and the trading structure as given. Also, the demand for each good does not exceed its supply. The phrase “price vector” is in quotes because not every state trades for every good. These numbers then do not describe the marginal rates of substitution between pairs of commodities for representative agents in the world economy. Instead, they are akin to state-specific terms of trade. At the current terms of trade, the demand for each country's resources does not exceed its supply, even after the fixed costs of maintaining the trading network have been paid. We have yet to describe conflict between states. There are two basic elements to the model of conflict. First, we follow Fearon (1995) and model war as a breakdown in interstate bargaining due to incomplete information. Gartzke's (1999) insight is telling: wars occur only randomly. Indeed, a war will occur Page 329 → only when an aggressor makes an onerous demand for resources that is rejected by the defender. Second, the outcome of a war is random. Although each state's power is given by a measure of its gross national product (GNP) that includes the gains from trade, it is possible for a strong state to lose a war. Thus a rich and varied group of trading partners bolsters the power of any state, but it does so only probabilistically. In any particular conflict, the odds are heavily with the Goliath, but every once in a while a small David will win. Let us now explore in more depth how incomplete information matters in the model of conflict. First of all, an

aggressor state has to decide if it is content with the status quo or if it prefers to make a threat for additional resources. Conflict is costly for any aggressor; it pays a fixed cost that depends naturally on its geographic distance from the defender. Since it pays the fixed cost only if it makes a threat, any threat will be credible. The aggressor is fully aware that it is dealing with two types of states: tough ones and conciliatory ones. A tough state is very capable of defending itself; in particular, it has a high defensive advantage in war. A conciliatory state is relatively weak, and its defensive advantage in war is much lower than that of its tough counterpart. However, whether a defender is tough or weak is unobservable, consistent with the logic that all states have incentives to present themselves as strong types (Fearon 1995, 395–401). Thus, while the aggressor knows the distribution of weak and tough defenders, it cannot determine the particular type of defender with which it is dealing when it makes a resource demand. Hence, it may make a demand that is based upon the average type, and if it makes a very large demand, then it may be confronted occasionally with the rude surprise of having to fight a tough defender unwilling to make an onerous resource concession. Let us now explore the stochastic nature of war's outcome in greater detail. The probability of winning a war depends upon a state's relative economic size and the variable describing the defender's advantage in warfare.6 An economically large state attacking a smaller one has a high probability of winning a war. Thus the probability of winning a war depends upon the evolution of trading patterns and the value of a state's resource base. A country with a rich array of trading partners will have a higher level of gross domestic product because it is benefiting more fully from the analog of gains from specialization, namely the benefits of increased diversity in its resources. If the attacker wins, then it expropriates all but a subsistence level of resources from the unsuccessful defending state. (This subsistence level is a predetermined parameter in the model.) The resources taken include the current net trades of the conquered state. Two comments are in order. First, Page 330 → although economic geography does not enter explicitly into the technology of war, it does enter into the calculation of the spoils of war. In particular, if two states are far apart, then the attacker reaps relatively small benefits from usurping the losing defender's net trades. Second, the benefits from winning and attack are myopic; they represent a one-time gain that depends on the richness of the defender's trading pattern. If the attacker loses, then it uses up all but a subsistence level of its resources in a futile attack. If the defender loses, it consumes only a subsistence level of its former economy, including net trades from abroad. Also, its endowment remains permanently lower at a subsistence level. On the other hand, if the defender wins, then its utility remains unchanged, and so does its endowment. How does a war affect the trading network? In the event of a demand for resource concessions (a possible prelude to war), we impose that the relevant dyadic trade link be severed, consistent with the existing statistical evidence in favor of the liberal hypothesis (Oneal and Russett 1997, 1999). If war does break out, it is permanently destructive. Since either the attacker or the defender loses, one of these two states will have a subsistence level of endowment permanently. Having a subsistence level of endowment makes it impossible ever to be an attacker again and often precludes any future trade. Indeed, the only way that such a state can afford to pay the fixed cost of maintaining its trading links is if the economic geography has already evolved so that the economic distance to its nearest neighbor is sufficiently small. This description of the outcome of war assumes that third parties in the world economy are neutral, an assumption that could be relaxed in a more complex model. In particular, in any period year when there is a war, both belligerent states consider their trades with third parties as given. Still, third parties will be affected deleteriously in the future because the aggregate resources in the world economy become lower, no matter who wins. This specification of war's outcome is draconian, but it has the virtue of analytical simplicity. It is also consistent with Liberman's (1996) finding that only the most ruthless conquerors can extract a large net economic surplus from the states they vanquish.

How the Model Evolves

The international system begins in autarky and a state of peace, with the initial economic geography and distribution of resources given. In each period, two states are chosen randomly, and the pair considers changing the current trading network. They can create a new trading link or sever one if they are currently trading partners. If, after paying the fixed cost of creating the new link, Page 331 → the two states experience a strict Pareto improvement, then the new network becomes a possibility for the next period's status quo. Or if after saving the fixed costs of the current link and forgoing the increased diversity in consumption that it entails, both states are strictly better off, then a new (less expansive) trading network becomes the possible future status quo. Otherwise, the current network remains unchanged. Both these decisions are myopic, and they ignore the effects that they have upon third parties. But they are based upon the equilibrium terms of trade that occur if the new link is established or if an old one is severed. The first state in the random pair is arbitrarily labeled the aggressor. Having explored the possibilities of trading peacefully, the aggressor state makes the unilateral decision of whether to pay the requisite fixed costs and then make a credible threat for resource concessions from its counterpart. As long as the aggressor is not at a subsistence level, there is the potential for war. The aggressor calculates the optimal resource concession that it will try to extract. It is worth emphasizing that economic distance enters into this calculation, and a state will find the resources of a nearby neighbor more attractive simply because it is easier to extract them. There are two types of threats: demands for nickel and dime concessions and demands for large concessions. A nickel and dime demand is so small that even a strong defender will not find it worthwhile to fight. A demand for a large concession is much more onerous. It is designed so as to extract the greatest possible surplus from a weak state, pushing the weak state to the very limit beyond which it will certainly fight. It may well be the case that the optimal demand for a resource concession is not worthwhile. In this case, no threat is made, and peace prevails. Thus the initial economic distances between states and the random process of starting wars or seeking trading partners give rise to a dynamical system. The steady state of this system describes the long-run trading network. If the initial economic distance between states is large enough and the endowments are roughly symmetric, then the unique long-run outcome will be autarky. War would never erupt between two roughly equal states because the downside of losing would deter any potential attacker from demanding large concessions. On the other hand, if all states are sufficiently close and there are sufficiently high defense parameters acting as deterrents to war, then the long-run outcome may be a cycle of nickel and dime threats among near neighbors. However, asymmetric terms of trade or an asymmetric distribution of resources may make some states temporarily powerful. This imbalance of power will have random effects on the outbreak of war and the subsequent distribution of resources in the world economy. Page 332 → If the minimum of the pair-wise distances between states is sufficiently large, then no state will trade with or enter into conflict with any other state. A state will not trade because the increased diversity in resources is not worth the large cost of setting up bilateral economic relations. And it will not make a threat because the expected utility of making the optimal threat will never be as large as the utility of staying in autarky. Thus the model is only interesting if some pair-wise economic distance is sufficiently small so that at least one pair of states is willing to trade. If the maximal pair-wise distance is sufficiently small, then every state will be willing to trade with every other since the fixed cost of setting up bilateral exchange is small for every partner. Also, any state may go to war, especially if it demands a large resource concession from a (surprisingly) tough state that has an attractive array of net trades in the current period. What determines whether a state trades or goes to war in this model? It is possible to show that states that are treated identically in the world trading system will not fight one another under normal circumstances because a defensive advantage in war—no matter how slight—entails that equally powerful states will stay at peace. In fact, we can achieve an even deeper insight. Consider two states that have the same set of trading partners. In this case, they consume the same set of goods and face the same terms of trade. Thus, their relative military

powers are equal to the relative sizes of their gross domestic products. If the share of weak defenders in the population is very small and the fixed cost of making a threat is sufficiently large, then an aggressor will not demand resources from the other. Again, the intuition is that the likelihood of winning a war is less than the expected gain accrued to the victor because both parties are already consuming the same bundle of goods. This serves as an important insight: military conflict emerges in this model only between states that do not have the same trading partners or when the share of weak defenders is sufficiently high. This insight accords with the statistical finding by Mansfield, Pevehouse, and Bearce (1999–2000) that military conflict is less likely to emerge between states that are party to the same preferential trading arrangement, thus having a similar set of trading partners. What happens to the likelihood of war if states are farther apart or if the subsistence level is made more severe? The gains from war are decreasing in the economic distance between two countries. In this case, higher fences do make better neighbors. In essence, the model's specification of war puts a “transportation tax” on the aggressor. It must pay the fixed cost to haul the spoils of victory back home. Also, it is easy to show that the likelihood of war is decreasing Page 333 → in the subsistence level, so less severe conquest decreases the likelihood of conflict. When is conflict likely? Consider two equally powerful countries that consume very different bundles of commodities. Then the gains from conquest look very favorable indeed. This situation could easily occur if the attacking state were large but isolated from the world trading system and if the defending state were small but well integrated into it. In essence, the small neighbor is consuming a wide array of complementary goods that are coveted by the attacking state. These observations have important implications for the long-run behavior of the world system. The long-run behavior of the world trading system is essentially summarized by the absorbing states of the random process of trade or conflict. Since a state that loses a war reverts to subsistence level, it can never attack another state in the future. Also, for realistic subsistence levels, it may revert to autarky and stay there unless there is exogenous technological progress in the economic geographic infrastructure. We have already seen that perpetual autarky is an absorbing state if the initial economic geography is sufficiently disjointed so that neither war nor trade can occur.7 Likewise, an integrated world economy in which every country is treated symmetrically and there are perpetual cycles of nickel and dime threats is also a long-run outcome, although it may not be possible to reach that situation from the initial state of autarkic economies. Since the model is highly nonlinear, it is difficult to derive analytical properties about its long-run behavior. Thus one can perhaps best understand its long-run behavior by repeated simulations for different parameter values of interest.

How the Computer Simulations Work The computer code for these simulations was written in Gauss Programming Language; this is a matrix-oriented language that has achieved widespread acceptance among economists for statistical and numerical analyses. Any researcher interested in simulating a model must sink the fixed costs of learning at least one computer language. An important consideration for detailed simulations like those reported in this chapter is the computational efficiency of the algorithms describing the model. Matrix-oriented languages can be quite efficient if the problem at hand lends itself to a recursive formulation. The flowchart for a single run of the simulation is shown in figure 1. A simulation consisted of one hundred years, allowing for one hundred cycles of trade and war. There are three random elements in any one run. First, the pair of states chosen in any period is random. Second, if there is conflict, then the type of defender is chosen randomly. Third, if there is war, then its outcome is random. Thus the result of any one run is stochastic and history dependent. Hence, for each configuration of conditions—describing an initial economic geography and the parametric specifications for revisionism, defense, and subsistence—the model is run for five hundred different trials.8 The summary statistics are recorded after each run, and averages (across all the runs for any particular initial configuration) are reported. Page 334 → Page 335 →

The most time-intensive part of the simulations is the algorithm that calculates the terms of trade for any possible trading pattern. The subroutine that performs this task is based upon a calculation described in Judd (1998, 188); it incorporates the typical elements of a fixed-point mapping that are at the heart of most computations of general equilibrium. These calculations are tantamount to finding the root of a highly nonlinear equation.9 The program can use a random number generator twice in one period, first to determine the type of defender and second to determine the outcome of war. Since the outcome of any one run is random, it is imperative to run the calibration a large number of times for each configuration of parameters. Then the researcher can be sure to find statistical regularities and record occasional extreme outcomes. Finally, it is worth mentioning that one should consider carefully the kind of statistics that the simulation will record. We kept track of the number of wars, the average number of trade links, the correlation between trade and war, and the timing of conflict. We then averaged these statistics across the five hundred trials that were run for each set of initial conditions.

Simulation Results This section reports some specific simulation outcomes. We answer the research question posed earlier: how does the initial economic geography affect the relationship between trade and war? As a working hypothesis, we expect the dyadic correlation between trade and war to vary with the initial economic geography of the system. Since economic distance reduces the possibilities for both economic exchange and military conflict, we further expect that the dyadic correlation between trade and war will be more positive in the distended economic geographies. Before we can run the simulation, we first need to calibrate the model, choosing values for the parameters. It is typical in applied macroeconomics to use parameters from microeconomic studies for some important parameters. Also, the researcher needs to explore the robustness of simulation outcomes, Page 336 → understanding that the model itself may be suspect if the results are very sensitive to a specific value for one of the parameters. For our simulations of the base case, we chose a moderate level of revisionism, applying it symmetrically to all states in the system. The model is homogeneous with respect to the initial endowments, so for simplicity we let each state have the same endowment, set at unity. We also set the share of initial resources required to open trade with one's nearest neighbor as 50 percent; this figure is consistent with some estimates of the share of resources in a modern economy that are devoted to transactions, not primary production. The subsistence level was set at 40 percent, meaning that a vanquished state loses 60 percent of its resources, a figure that is consistent with what happened to Japan and Germany after World War II. Building from the insight that nondemocratic states lose a disproportionate share of wars (Lake 1992; Reiter and Stam 1998; Bueno de Mesquita et al. 1999), we identify the proportion of weak states in the international system as 72 percent, according to the Polity III data set of Jaggers and Gurr (1995).10 The average defensive advantage in war is commonly taken to be 3 (Mearsheimer 1989). We assume that weak states have no defensive advantage in war; thus, the defensive advantage for strong states must be set as 8.14. We specify a 4 percent annual reduction in trading costs when a trading link is extant. In the base case, there is no exogenous technological change since we wish to focus on the endogenous evolution of trading networks. Thus, states cannot become closer unless they engage in trade. We examine a system of six states, configured into five different initial economic geographies. These are a line, a circle, a hub-and-spoke configuration, a globe, and a bimodal configuration. Figure 2 is a pictorial representation of initial geographic configurations. The distances have all been normalized so that the closest state is only one unit away. The hub-and-spoke configuration shows that the (arc) distance between two neighbors on the periphery is greater than 2π / 5 ≍ 1.26 the distance from any peripheral state to the hub. The line and circle represent distended economic geographies, as compared to the more compact hub-and-spoke and globe configurations. Bimodal is a hybrid geography, which is very compact within each bloc but distended across the two blocs. For each initial economic geography, we ran the model for a century, stopping after one hundred periods or when

all states were reduced to subsistence, whichever came first. Since pairs of states are drawn randomly in each period, we replicate each run five hundred times to assure that our results are stable. Table 1 reports summary statistics based on each geographic configuration. We report two different measures of military conflict. The first is the average number of wars occurring per replication. This measure tends not to be greater than six since we halt a run when all six states have lost a war and reached subsistence. The second is the percentage of replications ending before one hundred periods, when all states have been reduced to a subsistence level. Both measures tell a similar story: the compact configurations (hub-and-spoke and globe) are more war-prone than the distended ones, suggesting that economic closeness makes military conquest and resource extraction easier to achieve. The bimodal geography shows the least systemic conflict. Since the two blocs are far apart, states are able to fight only within their own blocs. Page 337 → Page 338 → The average number of trade links per replication measures the economic activity within the system. In a replication lasting one hundred periods with six states, there can be a maximum of thirty-six hundred trade links if every state traded with every other (including itself) in each period. Likewise, the minimal number of trade links is six hundred. The results show, not surprisingly, that economic distance functions as a barrier to trade. The distended geographies (line and circle) show fewer trade links than the compact ones. In the bimodal geography, there is no trade across the two blocs, and all economic activity occurs among states residing within the same bloc. After each replication, we also computed the dyadic correlation between trade and war. The last column in table 1 shows the average correlation over five hundred replications. Consistent with our working hypothesis, the dyadic correlation between trade and war is highly dependent on the initial economic geography of the international system. This correlation becomes more negative as the initial economic geography becomes more compact. The hub-and-spoke and globe configurations exhibit a negative correlation, while the more distended line and circle show a positive correlation. The correlation between war Page 339 → and trade in the hybrid bimodal geography is strongly positive since each state is able to trade and fight only with the two other states located within its bloc. The initial economic geography also has important effects on the timing of conflict in the international system. Figure 3 shows the average number of wars in each period for three configurations. Wars occur at a relatively uniform rate in the comparatively peaceful line configuration. This contrasts with the timing of conflict in the globe, where the early periods are marked by heavy conflict, which then tapers off in later periods. In this configuration, states quickly establish different trading networks, and military conflict with states not having the same set of trading partners becomes immediately attractive. Military conflict then tapers off as aggressors inevitably lose wars and revert to a subsistence level. The bimodal configuration exhibits little conflict in early periods, and wars tend to occur more frequently later on. Symmetric trading networks quickly emerge within each bloc, making conflict less attractive in the early periods due to the implicit balance of power. However, when a unilateral demand for additional resources does occur, trading relationships are cut. These broken links create asymmetric trade patterns and a local imbalance of power, making military conflict more attractive in later periods.

Conclusion This chapter has shown the potential value of using computer simulation models to explore the relationship between international trade and conflict. We illustrate this value by presenting an agent-based model where trade and war become emergent properties within a system of states configured in different economic geographies. The simulation results show that the initial economic geography has important effects on systemic trade and war, the dyadic correlation between trade and war, and the timing of conflict in the international system. Compact

geographies exhibit greater economic exchange and military conflict but also a more negative correlation between these two variables. This finding has important implications for future studies of international trade and war. Scholars are more likely to find a negative relationship—supporting the liberal hypothesis—when they focus their empirical analysis on proximate pairs of states, where the economic geography is likely to be relatively compact. When examining a broader sample of state pairs, it will be important to control carefully for economic geography. In this regard, it is worth remembering that economic geography includes not only the physical distance between two states but also the economic technology and infrastructure connecting them; hence, this factor can be expected to vary across both space and time. But without proper consideration for the effects of economic geography, the true relationship between trade and war may be very hard to identify. Page 340 →

NOTES The authors are grateful for a grant from the Mershon Center, which made this research possible. They have also benefited from the comments of the editors and the participants at the 2000 Mershon Conference on Interdependence and Conflict. 1.Barbieri and Schneider (1999, 395) provide a concise summary of the large-N statistical studies of international trade and conflict. They (1999, 391) also offer a quick review of the formal work on the subject. We have no knowledge that any of these formal models have been extended into computer simulations of trade and war. 2.Lucas (1976) made an even more radical critique of econometric analyses that Page 341 → were not based upon an underlying model. He argued that most statistical analyses were useful for short-term forecasting only and that they could be very misleading for policymakers if the underlying structure of the economy had changed. 3.The articles by Hansen and Heckman (1996) and Sims (1996) are very illuminating. 4.We recognize that there is much less agreement among political scientists regarding the value of formal models. See Brown et al. 2000. 5.Rauch 1999; Bala and Goyal 2001; and Kranton and Minehart 2001 are examples of a growing literature on networks in economics. 6.Many studies, such as Organski and Kugler 1980, use GNP as a rough measure of a state's military power. Others, like Singer, Bremer, and Stuckey 1972, use the Correlates of War project's composite index of national capabilities. 7.Indeed, this insight can be generalized: any country that is permanently sufficiently distant from all its neighbors will never become a part of the world trading system, nor will it ever be involved in war. 8.A typical session with five hundred runs might take between five and ten minutes on a PC with a very fast Pentium processor. This is typical for simulations of even a simple model like this one. That is why these techniques have become available to the typical researcher only in the last decade, and there is always a premium on efficient programming for the solution of any nonlinear equation. 9.Economists often normalize prices in general equilibrium so that they add to unity. Thus the algorithm for finding an economic equilibrium searches over a simplex whose dimension is one less that the number of states in the model, since each country has its own commodity. 10.This number is of course sensitive to the exact definition of democracy. We used all country-year observations for which there were no missing data and subtracted the autocracy measure from the democracy measure. This yields 13,911 observations on an index ranging from 10 to 10. Following Russett 1993, we treat the 9,958 country-year observations in the bottom two-thirds of this range as nondemocratic. We deliberately use a generous definition of democracy to avoid undercounting the number of strong states, some of whom may be only semidemocratic.

REFERENCES Bala, Venkatesh, and Sanjiv Goyal. 2000. A Non-Cooperative Model of Network Formation. Econometrica 68:1181–1229. Barbieri, Katherine, and Gerald Schneider. 1999. Globalization and Peace: Assessing New Directions in the Study of Trade and Conflict. Journal of Peace Research 36:387–404. Bearce, David. H., and Eric O'N. Fisher. 2002. Economic Geography, Trade, and War. Journal of Conflict Resolution 46:365–93. Page 342 → Bremer, Stuart A., and Michael Mihalka. 1977. Machiavelli in Machina: Or, Politics among Hexagons. In Problems of World Modeling, ed. Karl W. Deutsch, Bruno Fritsch, Helio Jaguaribe, Andrei S. Markovits, 303–37. Cambridge, MA: Ballinger. Brown, Michael E., Owen R. Cote Jr., Sean M. Lynn-Jones, and Steven E. Miller. 2000. Rational Choice and Security Studies: Stephen Walt and His Critics. Cambridge, MA: MIT Press. Bueno de Mesquita, Bruce, James D. Morrow, Randolph M. Siverson, and Alastair Smith. 1999. An Institutional Explanation of the Democratic Peace . American Political Science Review 93:791–807. Carr, Edward Hallett. 1946. The Twenty Years' Crisis, 1919–1939: An Introduction to the Study of International Relations. New York: Harper and Row. Cederman, Lars-Erik. 1997. Emergent Actors in World Politics: How States and Nations Develop and Dissolve. Princeton: Princeton University Press. Cusack, T. R., and Richard Stoll. 1990. Exploring Realpolitik: Probing International Relations Theory with Computer Simulations. Boulder: Lynne Rienner. Epstein, Joshua M., and Robert Axtell. 1996. Growing Artificial Societies: Social Science from the Bottom Up. Washington, DC: Brookings Institution Press. Fearon, James D. 1995. Rationalist Explanations for War. International Organization 49:379–414. Gartzke, Erik. 1999. War Is in the Error Term. International Organization 53:567–87. Hansen, Lars P., and James J. Heckman. 1996. The Empirical Foundations of Calibration. Journal of Economic Perspectives 10 (winter): 87–104. Jaggers, Keith, and Ted Robert Gurr. 1995. Tracking Democracy's Third Wave with the Polity III Data. Journal of Peace Research 32:469–82. Jodhimani, Albert. 1999. Transaction Costs and the Evolution of Market Completeness. Ph.D. diss., Ohio State University. Judd, Kenneth L. 1998. Numerical Methods in Economics. Cambridge, MA: MIT Press. Kissinger, Henry A. 1957. A World Restored: Castlereagh, Metternich, and the Problem of Peace, 1812–22. Boston: Houghton Mifflin. Kranton, Rachel E., and Deborah F. Minehart. 2001. A Theory of Buyer-Seller Networks. American Economic Review 91:485–508.

Kydland, Finn E., and Edward C. Prescott. 1996. The Computational Experiment: An Econometric Tool. Journal of Economic Perspectives 10 (winter): 69–85. Lake, David A. 1992. Powerful Pacifists: Democratic States and War. American Political Science Review 86:24–37. Liberman, Peter. 1996. Does Conquest Pay? The Exploitation of Occupied Industrial Societies. Princeton: Princeton University Press. Lucas, Robert. E. 1976. Econometric Policy Evaluation: A Critique. In The Phillips Curve and Labour Markets, ed. Karl Brunner and Allan H. Meltzer, Carnegie-Rochester Conference Series on Public Policy, 1:19–46. Amsterdam: North-Holland. Mansfield, Edward D., Jon C. Pevehouse, and David H. Bearce. 1999–2000. Preferential Trading Arrangements and Military Disputes. Security Studies 9:92–118. Page 343 → Mearsheimer, John J. 1989. Assessing the Conventional Balance: The 3:1 Rule and Its Critics. International Security 13:54–89. Morgenthau, Hans J. 1948. Politics among Nations: The Struggle for Power and Peace. New York: Knopf. Morrow, James D. 1989. Capabilities, Uncertainty, and Resolve: A Limited Information Model of Crisis Bargaining. American Journal of Political Science 33:941–72. Oneal, John R., and Bruce M. Russett. 1997. The Classical Liberals Were Right: Democracy, Interdependence, and Conflict, 1950–85. International Studies Quarterly 41: 267–94. ———. 1999. Assessing the Liberal Peace with Alternative Specifications: Trade Still Reduces Conflict. Journal of Peace Research 36:423–42. Organski, A. F. K., and Jacek Kugler. 1980. The War Ledger. Chicago: University of Chicago Press. Pollins, Brian M. 1982. Modeling International Trade Flows: A Survey and Comparison of Simulation Approaches. International Political Science Review 3:503–33. Rauch, James E. 1999. Networks versus Markets in International Trade. Journal of International Economics 48:7–35. Reiter, Dan, and Alan Stam III. 1998. Democracy, War Initiation and Victory. American Political Science Review 92:377–89. Russett, Bruce M. 1993. Grasping the Democratic Peace: Principles for a Post–Cold War World. Princeton: Princeton University Press. Samuelson, Paul A. 1954. The Transfer Problem and Transport Costs, II: Analysis of Effects of Trade Impediments. The Economic Journal 64:264–89. Schweller, Randall L., and David Preiss. 1997. A Tale of Two Realisms: Expanding the Institutions Debate. Mershon International Studies Review 41:1–32. Sims, Chris A. 1996. Macroeconomics and Methodology. Journal of Economic Perspectives 10 (winter): 105–20. Singer, J. David, Stuart Bremer, and Jonathan Stuckey. 1972. Capability Distribution, Uncertainty, and Major Power War, 1820–1965. In Peace, War, and Numbers, ed. Bruce M. Russett, 19–48. Beverly Hills: Sage.

Tetlock, Philip E., and Aaron Belkin. 1996. Counterfactual Thought Experiments in World Politics. Princeton: Princeton University Press. Walras, Léon. 1874. Eléments d'Economie Politique Pure. Lausanne: Corbaz. Page 344 →

Page 345 →

Contributors Katherine Barbieri is Assistant Professor of Political Science at Vanderbilt University. David H. Bearce is Assistant Professor of Political Science at the University of Pittsburgh. Jean-Marc F. Blanchard is a Research Associate with the Christopher H. Browne Center for International Politics and an Associate with the Center for East Asian Studies, both at the University of Pennsylvania. Janet M. Box-Steffensmeier is Associate Professor of Political Science at the Ohio State University. Eric O’N. Fisher is AGIP Professor of International Economics at Johns Hopkins University, Jean Monnet Fellow at the European University Institute, and Associate Professor of Economics at the Ohio State University. Erik Gartzke is Assistant Professor of Political Science at Columbia University. Christopher Gelpi is Associate Professor of Political Science at Duke University. Joseph M. Grieco is Professor of Political Science at Duke University. Gregory D. Hess is Russell S. Bock Professor of Public Economics and Taxation at Claremont McKenna College. Jack S. Levy is Board of Governors’ Professor of Political Science at Rutgers University. Edward D. Mansfield is Hum Rosen Professor of Political Science and Director of the Christopher H. Browne Center for International Politics at the University of Pennsylvania. Page 346 → Michael Mastanduno is Nelson A. Rockefeller Professor and Chair of the Department of Government at Dartmouth College. James D. Morrow is Professor of Political Science and Senior Research Scientist at the Center for Political Studies at the University of Michigan. John R. Oneal is Professor of Political Science at the University of Alabama. Jon C. Pevehouse is Assistant Professor of Political Science at the University of Wisconsin-Madison. Brian M. Pollins is Associate Professor of Political Science and Research Fellow at the Mershon Center at the Ohio State University. Dan Reiter is Associate and Winship Research Professor of Political Science at Emory University. Rafael Reuveny is Associate Professor of Political Economy in the School of Public and Environmental Affairs at Indiana University. Norrin M. Ripsman is Assistant Professor of Political Science at Concordia University. Bruce Russett is Dean Acheson Professor of International Relations at Yale University. Beth Simmons is Professor of Government at Harvard University. Etel Solingen is Professor of Political Science at the University of California, Irvine.

Arthur A. Stein is Professor of Political Science at UCLA. Richard J. Timpone is a marketing consultant in the Office of Administration’s Planning and Strategy Group at Nationwide Insurance. Christopher J. Zorn is Associate Professor of Political Science at Emory University.

Page 347 →

Index Abdelal, Rawi, 179 Absolute gain, 12, 16 Academics, 31, 207. See also Scholars Achen, Christopher, 164 Action-reaction conflict dynamics, 259 Actors, 5, 6, 60, 211, 261; action and, 255–56; choice of, 318–19; defined, 255; multiactor system, 140; nonstate, 260; private, in trade, 32, 33, 112, 113, 121. See also State leaders Africa, 224 Agent-based model, 326, 329, 339. See also Computer simulations Ali, Salvatore, 312 Alliances, 114, 132, 152, 153–54. See also Coalitions; Dyads Alvarez, R. Michael, 297 AMELIA (statistical routine), 163 Anderton, Charles H., 135, 217 Angell, Norman, 45 Arab Common Market, 71 Arab-Israeli arena, 71, 80, 162 Argentina, Brazil and, 71, 226 Armed conflict, 5–6, 7, 52. See also Conflict; Militarized interstate disputes (MIDs) Asian tigers, 182 Association of Southeast Asian Nations (ASEAN), 75, 76, 224 Asymmetric dependence, 161, 210, 331 Asymmetric information, 102, 117, 120 Asymmetric interdependence, 13, 34, 93, 320 Asymmetry, in trade, 12, 123–24n. 10 Autarky, 330, 332, 333 Authoritarian leaders, 47–48, 52, 56n. 10. See also Nondemocratic leaders Autocratic states, 49, 51–53, 56–57n.11

Autonomous state, 33–35, 38, 40, 41nn. 2–3 Azar, Edward E., 241, 255 Backlash coalitions, 61, 63, 67, 75, 76; clusters of, 70–71; entrepreneurs in, 61, 62, 76; international crises and, 72–73 Baldwin, David A., 177, 179 Barbieri, Katherine, 13, 44, 52, 190, 207, 241; quantitative studies by, 212, 217 Bargaining theory, 96, 102, 104, 109n. 9 Baseline hazards, 279 Bearce, David H., 224, 324, 327 Beck, Nathaniel, 196–97, 201 Behavioral Correlates of War (BCOW), 242, 256. See also Correlates of War (COW) project Bennett, D. Scott, 241 Bilateral dependence, 320n. 3 Bilateral relations, 120 Bilateral trade, 113 Bilateral trade/GDP ratio, 190, 191, 196, 199, 228–29 Biostatistics, 283 Blanchard, Jean-Marc F., 13, 20 Page 348 → Boehmer, Charles, 14, 104, 106 Bootstrapping (statistical technique), 297, 298–300, 307 Boundedness: of claims, 69–75; contingency and, 7–10 Box-Steffensmeier, Janet M., 273 Brazil, Argentina and, 71, 226 Brecke, Peter K., 20 Brinkmanship, 7 Britain. See England; Great Britain Bueno de Mesquita, Bruce, 47 Bush, George W., 232 Buzan, Barry, 3

Canada, 138 Capital, monetary policy and, 75, 106 Carter, John, 135, 217 Case studies, 20, 314, 319; analyses, 128, 179–80; research, 310, 313. See also Qualitative research; Scholars Castro, Fidel, 179 Central American Common Market, 224 China, 113, 175; United States and, 10, 159, 168–71, 180, 184–85 Clinton, Bill, 232 Coalitions: conflict/cooperation and, 36, 48, 55n. 3, 60–82, 181; domestic, 60, 61, 63, 64, 69, 79; dynamics of, 70, 72; entrepreneurs in, 61–62, 76; hybrid, 61, 62, 63; liberal tradition and, 61, 77, 78; nationalist, 181; strategies of, 62, 63, 64, 68, 70, 72; strength of, 64, 77, 80nn. 23, 30. See also Backlash coalitions; Internationalizing coalitions Coercion, trade and, 111–12, 115–16, 117 Cold War era, 74, 217, 241, 316; economic engagement in, 175, 177–78, 179, 183; liberal peace in, 190, 192; trade in, 113, 114, 115; trade sanctions during, 162, 169 Collective security arrangement, 153 COMECON (Council for Mutual Economic Cooperation), 190 Commercial liberalism, 131–34, 312, 314–17, 319. See also Liberalism Commercial liberalization, 226 Commercial regionalism, growth of, 223–24 Commercial relations, 201. See also Trade Commercial rivalries, 8 Commodities traded, 217 Competition, coalitional, 69 Computer program, 256 Computer simulations, 20, 324–41; agent-based model, 324, 326, 329; economic geography and, 324–25, 327–28, 330–32, 335–40; evolution of model, 330–33; how they work, 333–35; model-based, 325–26; results in, 335–39; in social sciences, 325–26; static elements of, 327–30; subsistence elements of, 330, 333; trading networks and, 324–25, 326, 327–28, 330–32, 333, 336 Conciliatory state, 329 Conflict: change in causes of, 274–77; computer model of, 326, 328, 329; contained, 68; and contests, 104–5; diversionary, 151, 152; inducements and suppressors of, 165–68; international, 16–17, 150, 242–48; likelihood vs. level of, 259; measure of, 242, 266n. 16; observing, 17–18; repeated, 277–79; selfreinforcing, 277; self-

restraining, 278; theory of, 112, 116; time and, 274, 340; trade and (see Trade-conflict dynamics). See also Armed conflict; Militarized interstate disputes (MIDs) Conflict and Peace Data Bank. See COPDAB Conflict intensity, 258 Conflicts of interest, 116, 117 Conflictual process, 273 Contests, theory of, 99–104 Context sensitivity estimator (CSE), 313 Contingent nature of war/trade relationship, 2 Continuous endogenous regressors, 299, 304 Continuous variables, 295, 296, 298–99 Cooperation, 63, 66, 251; measure of, 242, 248–49; trade and, 112–13, 240 COPDAB (Conflict and Peace Data Page 349 → Bank), 17–18, 164, 248, 251n. 2, 254, 257; conflict/cooperation measure and, 243–45; empirical investigation and, 241, 242; GEDS comparison with, 264; WEIS data compared with, 243–45, 251nn. 10, 11, 256, 260, 262 Copeland, Dale C., 131, 142n. 11, 180, 312, 314 Correlates of War (COW) project, 90, 191, 195; Behavioral (BCOW), 242, 256 Cost/benefit decisions, 46, 100. See also Opportunity-costs model Costly signaling, 35, 96, 107, 117–18; informational focus of, 105; problems with, 120; theory of contests and, 102, 104; trade as source of, 89–94. See also Opportunity-costs model; Signaling Cost of vulnerability. See Vulnerability, economic Cost of war, 89, 112, 138 Costs, in trade, 327–28, 329, 336 Counterfactual analysis, 14, 325 Covariant structure models, 300–301 Covariate effects, 275, 276, 277, 278; heterogeneity and, 281, 282 COW. See Correlates of War (COW) project Cox models, 274 Crisis decision making, 316, 318 Cuba, 176, 179 Cure (“split-population”) models, 280, 281

Data, limitations of, 216–17 Data sets, 263–65; availability of, 243. See also Events data; and specific data sets Davis, David R., 230, 241 Davis, Patricia, 179 De-escalation process, 248 Democracy, 47, 91, 163, 170, 341n. 10; peace and, 189, 200, 201 Democracy, trade, and peace, 44–57; autocracy and, 49, 52, 56–57n .11; current debate on, 45–46; Kantian view of, 44, 45–50, 55n. 1 Democratic governments, 36 Democratic leaders, 36, 49, 50, 150 Democratic peace, 127, 152 Democratic states, 163, 189, 214 Democratization, 77, 170 Dependence: asymmetric, 161, 210, 331; bilateral, 320n. 3; economic, 5, 213; theory of, 104; on trade, 50, 314 Dependent variables, 259, 289, 294, 315, 319–20; conflict risk in, 192; continuous, 295, 296, 298–99; dichotomous, 296–98; lagged, 164, 165; trade conflict and, 163–65 Deterrence, 89, 99, 108, 123, 124; failure of, 37; literature of, 114 Developing countries, 9, 214, 215. See also specific countries Dichotomous dependent variables, 296–98 Differential costs, 100, 109n. 6. See also Costs Diplomacy, 97, 117 Diplomatic alignments, 140 Diplomatic climate and relations, 18 Diplomatic disputes, 166, 167, 168 Direction of Trade (DoT) statistics, 15, 190, 250 Disease/violence analogy, 166 Disputes: diplomatic, 166, 167, 168; fatal, 163, 198–200, 202; interstate, 20, 96, 106, 114; low-level, 245, 248; militarized, 94, 116, 168, 279 (see also Militarized interstate disputes [MIDs]); multiple, 278; political, 116, 122; selection of, 93–94; theory of, 100 Distributed-lags models, 201, 202 Domestic citizens, 151

Domestic coalitions, 60, 61, 64, 69; political economy and, 63, 79 Domestic disequilibria, 211 Domestic factors, 6, 7 Domestic incentives, 44 Domestic innovations, 34 Domestic politics, 35, 181, 182 Domestic regime, 54 Domestic sensitivity to costs, 128, 133 Domke, William, 12 Page 350 → Dyadic form, 257 Dyadic level of analysis, 5, 12, 21, 127, 128, 139 Dyadic measure, 314 Dyadic models, 139 Dyadic outcomes, 140 Dyadic peace, 194 Dyadic time series, 189 Dyadic trade, 134, 136, 143n. 19, 330; dependence, 52, 214; events data and, 262; flows of, 212–13; value of, 160 Dyads (pairs of states), 19, 90; autocratic, 51–54; choice of, 317; disputes between, 258; exchanges between, 79, 162; history of, 201, 276; homogeneous, 75; interdependent, 164, 214; politically relevant, 10, 162, 163, 227, 280; power, 192; studies of, 10, 228; symmetric, 93 East African Common Market, 224 Economic Community of West African States (ECOWAS), 224 Economic cooperation, 66 Economic engagement, 175–85; conditions for success of, 176; current state of knowledge of, 177–80; domestic politics and, 180–84; United States and, 175–76, 179, 180, 182, 183–85 Economic equilibrium, 328 Economic gain, 12, 16, 211 Economic geography, 326, 335–40; trade networks and, 324–25, 327–28, 330–32, 333, 336 Economic growth, 2, 49, 55n. 5 Economic interdependence, 31; asymmetrical, 180; bilateral, 196; democratic state and liberal peace, 44–49;

empirical investigations in, 241–42; histories of, 276; impact on conflict of interests and contests, 104; internationally oriented, 79; measurement themes, 11–12; not just trade, 105; objective or subjective, 314; observing, 11–16; operational indicators of, 15; similar interests and, 109n. 8; substantiation of, 275; uncertainty and, 101; war probability and, 139 Economic opportunity costs. See Opportunity-costs model Economic regionalism, 232 Economic relationships, long-term, 275 Economic restraints, 136 Economic sanctions, 115–16, 117, 123, 177; failure of, 118–19, 124nn. 5, 20, 181; negative, 178, 179, 181; tariffs and, 148, 152; trade sanctions, 90, 92, 94 Economic signals, 34, 134. See also Costly signaling Economic size, 329 Economic technology, 339 Economic variables, 163 Economic vulnerability, 4 Economy dependence, 5, 213. See also Economic interdependence Efficiency gains, 3, 13, 128 Elasticity measures, 217 Electoral risks, 35, 36, 55n. 2 Elites, revolutionary, 161. See also State leaders Elizabeth I, queen of England, 137–38 Empirical analysis, 9, 111–12, 260–63 Endogeneity, 50, 112, 138, 201 Endogeneity, in statistical analysis, 289–308; bootstrapping and, 297, 298–300, 307; caveats on correction of, 301–4; covariance structure models and, 300–301; OLS model, 290, 295, 297, 302, 305, 306; reduced-form equation and, 299, 305, 307n. 1; regression in, 296, 297, 298–300, 303, 305, 306, 307; statistical estimators and, 293, 302, 304, 307; testing for corrections in, 304–6. See also Endogenous variables Endogeneous growth theory, 49, 56n. 8 Endogenous variables, 294–99; continuous, 294–96, 305; dichotomous, 296–99; explanatory, 291, 298, 301, 304, 307; instrumental, 295, 301, 302, 307; trade, 306 England, and war with Spain, 137–38, 144n. 25. See also Great Britain Entente Cordiale (1904), 197–98

Epidemiological analogy, 166 Page 351 → Error terms, 121, 293, 294, 301, 304, 305 Escalatory and de-escalatory process, 17 Ethnic and religious diversity, 71 Ethno-confessional cleavage, 69 Europe, 138. See also specific countries European economic community, 1, 224 European Union (EU), 79, 209 Events data, 18, 91, 93, 255–63, 345; actors in, 255–56, 260, 261; analysis of, 249; coding of, 242–45, 246, 250, 255; dyad-years in, 243–45, 248, 249; frailty models and, 281–82; limitations of, 250, 259; MIDs data and, 240, 255–57, 260, 262–63; theory evaluation with, 263. See also COPDAB; WEIS data set Exchangeable observations assumption, 279 Exogenous technological progress, 333 Exogenous variables, 165, 291, 292, 293, 298. See also Endogenous variables Exploitation, 39 Externalities of trade, 33, 34 Falklands/Malvinas war, 71 Fatal disputes, 163, 198–200, 202 Fisher, Eric O’N., 324, 327 Five-tier intensity scale, 243 Fixed-point mapping, 335 Foreign investment, 2, 70, 72 Foreign markets, 3 Foreign policy, 21, 182 Foreign security policy executive, 318, 319 Formal model, 325–26. See also Computer simulations Frailty models, 281–82 France, 139, 144n. 28, 198 Free capital mobility, 75. See also Capital

Free trade, electorate and, 36 Galtung, Johan, 181 Game theory, 133, 141 Gartzke, Erik, 7, 14, 164, 241, 328; on peace and interdependence, 96, 104, 106 Gasiorowski, Mark, 241 Gauss Programming Language, 333 GDP. See Gross Domestic Product GEDS (Global Event Data System), 256, 264 Gelpi, Christopher, 6, 36, 44, 170 General Agreement on Tariffs and Trade (GATT), 230, 231, 233n. 6 Generalized estimating equations (GEE), 128, 172n. 2, 192–93, 196, 197 Geographic proximity, 161, 167–68, 329. See also Economic geography Germany, 139, 144n. 28, 176, 214; Soviet Union and, 177, 178 Glasgow, Garrett, 297 Gleditsch, Kristian, 163, 190, 197, 198 Global access, 65, 66 Global economic integration, 102 Global Event Data System (GEDS), 256, 264 Global institutions, 61, 231 Globalization, 20, 97, 106–7 Global markets, 60, 61, 171 Global politics, 105 Global trade, 6 GNP. See Gross National Product Gorbachev, Mikhail, 178 Granger causality, 201, 242, 300 Gravity models, 161 Great Britain, 137–38, 143n. 23, 171, 198; Germany and, 56–57n. 11 Grieco, Joseph M., 6, 36, 44, 170 Gross Domestic Product (GDP), 37, 53–54, 72, 168, 228; as trade-conflict measure, 213, 216, 250; utility function

and, 327. See also Trade/GDP ratio Gross National Product (GNP), 212, 216, 329 Hamilton, Alexander, 4 Hawaii, kingdom of, 180 Hazard models, 179, 277, 279; proportional, 274–75, 276 Heagerty, Patrick, 197, 198 Hegemony, theories of, 132–33, 142nn.15–16 Hegre, Håvard, 9 Hess, Gregory D., 8, 148 Page 352 → Heterogeneity, temporal dynamics and, 19, 274, 279–82 Heterogeneous dyads, 75 “Highest action in dispute” variable, 258, 262 Hirsch, Leonard Paul, 212 Hirschman, Albert O., 3, 14, 34, 161; on economic engagement, 177, 180, 185 Hirst, Monica, 226 Historical contingency, 324 Hostility level variable, 248, 258, 262 How Scientists Explain Disease (Thagard), 166 Hughes, Barry B., 212 Hussein, Saddam, 179 Hybrid coalitions, 61, 62, 63 Iceberg costs, 328. See also Costs Identification, problems of, 307 Ideological tone, 208 IGO (intergovernmental organization), 167, 170, 230 IMF. See International Monetary Fund Independent variables, 312–15, 317, 318; dependent variables and, 310, 319, 320 Industrial concentration, 38, 41

Inertia, 259 Informational asymmetries, 102, 117, 120 Informational constraints, 219 Informational perspective, 101, 104, 118 Information environment, 35. See also Events data Institutional context, 222 Institutions: alliances of, 153–54; domestic, 6; international, 73, 142n. 10, 148–49; regional, 79–80 Instrumental variables, 295, 301, 302, 307 Interdependence, asymmetric, 13, 34, 93, 320 Interdependence, financial, 141n. 1. See also Economic interdependence Interdependence and conflict: conceptualization and measurement of, 10–18; conceptual links between, 2–4; methodological issues and advances on, 18–20; stability of relationship between, 21; theoretical issues, 5–10 Interdependent and dependent relations,211 Interdependent dyads, 164, 214 Interest groups, 37 Interests, commonality of, 109n. 8, 132 Intergovernmental organization (IGO), 167, 170, 230 International arena, 63 International conflict, 6, 150; armed, 5, 7, 17, 52; conceptualizing, 16–17; cooperation and, 242–48 International institutions, 73, 142n. 10, 148–49 Internationalizing coalitions, 73, 76–77, 78, 181; backlash regimes and, 68, 81n. 7; clusters, 70, 75; cooperation and, 65, 66; domestic coalitions and, 61, 63; interdependence and, 79; strategies of, 62–63 International markets, 2, 50. See also Global markets International Monetary Fund (IMF), 162, 172n. 2; trade statistics of, 15, 190, 217, 250 International political economy (IPE), 76, 107 International politics, 2, 22 International relations theory, 182 International trade and security, 3 Interstate cooperation, 112, 114 Interstate disputes, 20, 96, 106, 114. See also Militarized interstate disputes (MIDs)

Iran, 154, 176 Iraq, 154, 179 Israel, 71, 80, 162 Jackson-Vanik Amendment, 184 Japan, 176, 214 Kang, Heejoon, 242, 262 Kansas Event Data System (KEDS), 256, 264 Kant, Immanuel, 16, 44, 55n. 1, 129 Page 353 → Kantian equilibrium/peace, 150, 189 Kantian hypothesis, 45–46, 129–30, 149, 166; modern microfoundations for, 46–50 Katz, Jonathan N., 196, 199 Keynes, John Maynard, 223 Kim, Soo Yeon, 201 Kim Il Sung, 70 Kirshner, Jonathan, 179 Kissinger, Henry, 1, 184 Korean peninsula, 80, 175 Large-N statistical studies, 163, 259, 310, 340n. 1; analyses and, 209, 317, 324; database for, 178, 184; imperfections of, 8, 10, 20; qualitative research and, 128, 313, 314 Latin America, 80. See also specific nations Leaders. See Nondemocratic leaders; Political leaders; State leaders League of Nations, 190, 224 Least squares. See Ordinary least squares (OLS) model; Two-stage least squares (2SLS) method Levy, Jack S., 8, 127, 151, 312 Li, Quan, 96, 104, 106 Liberal argument, 3, 4, 32, 45–46, 209–11, 224; openness and absolute gain and, 12, 16; simultaneous model, 19; testing claims of, 313 Liberal democracies, 56n. 11. See also Democracy Liberalism: classical, 61, 77, 97, 98, 111, 113; commercial, 131–34, 312, 314–17, 319; critics of, 208, 210

Liberalization, 36, 50, 63 Liberal peace, 45, 54, 105, 196–97, 217; case for, 198–200; limits of, 202 Liberals, 2, 38, 45, 189, 240; assumptions of, 143n. 19; explanations of, 37; hopes of, 153; institutions of, 200; theory of, 19, 97, 129, 190; tradition of, 140, 211 Liberman, Peter, 9, 311, 330 Limited information approach, 294, 304 Linear probability model, 298 Logit models, 57n. 12, 290, 296 Logrolling, 61 Low-level disputes, 245, 248 Macroeconomics, 36, 63, 335 Madison, James, 138 Major powers, 9, 191, 192. See also specific countries Manchester liberals, 132, 240 Mansfield, Edward D., 1, 45, 142n. 10, 159, 241; on actors in conflict, 60; on data sets, 239; on trade effects, 131, 241 Maoz, Zeev, 191, 257 Marginal-cost argument, 123n. 2 Market conditions, 120 Markets, 3, 106, 115, 120; global/international, 2, 50, 60, 61, 171; power of, 119, 122 Martin, Lisa, 170 Marx, Karl, 33 Marxist theories, 38–39, 42n. 7, 210, 263 Mastanduno, Michael, 10, 170 McClelland, Charles A., 242, 255 Measurement: of trade-conflict dynamics (see Trade-conflict dynamics, measurement of); and zeros for trade, 162–63 Median economic interests, 9 Median-voter theories, 35–36, 37, 40, 55n. 2 Mercado Común del Sur (MERCOSUR), 71, 224, 226, 227 Mercantile era, 10

Mercantilists, 3, 8, 128. See also Realists MERCOSUR. See Mercado Común del Sur Microeconomics, 335 Microfoundations, 5–7, 45 Microtheories, 164 MILEX (military expenditure), 72, 73 Militarized conflict, 46, 50, 51–52, 332. See also Armed conflict Militarized disputes, 91, 94, 116, 168, 279 Militarized interstate disputes (MIDs), 127, 130, 191, 240; conditions for, 166–67; models of, 193, 199; traderelated, 90, 164, 227–28 Page 354 → Militarized interstate disputes (MIDs) data sets, 19–20, 114–15, 191, 242–51, 254; causal linkages and, 130; coding scheme, 242–45, 246, 248, 250; conflict intensity and, 258; episodic nature of, 245; events data compared to, 17–18, 240, 255–57, 260, 262–63; limitations of, 163, 258–61, 263; missing disputes in, 246, 248, 251n. 8; models of, 193, 199; time variables in, 261, 265 Military: alliance, 115; buildups, 65; conquest, 2; expansion, 4; reachability, 283 Minor/zero trade distinctions, 216–17 Mixed dyads, 70 Mixed regional clusters, 67 Mixture models, 279 Modeled war, 326. See also Computer simulations Model estimation, 19 Monetary policy, 106 Monotonic relationship, 98 Montesquieu, Baron de, 3 Moral hazard, 179 Morgenthau, Hans, 183 Morrow, James D., 34, 37, 89, 104, 151, 164 Morse, Edward, 45 Most-favored nation (MFN), 113 Multiactor system, 140 Multilevel inferences, 282

Multimethod strategy, 207, 212 Multiple alliances, 154 Multiple disputes, 278 Multiple equilibria, 151 Multiple testing measures, 215 Nasser, Gamal A., 70 National interest, 122 National-level studies, 217 National Power and the Structure of Foreign Trade (Hirschman), 177 National Science Foundation (US), 264 National security, 1, 33–34 Nation-pairs. See Dyads Nazi Germany, 177 Negative sanctions, 178, 179, 181. See also Economic sanctions Negotiations, 90, 100, 226, 231. See also Diplomacy Neo-Marxists, 263. See also Marxist theories Neorealists, 33. See also Realists New England, 138 Newnham, Randall E., 179 Nixon, Richard M., 1, 184 Nondemocracies, 44, 46, 48 Nondemocratic leaders, 49, 50, 54, 150; authoritarian leaders, 47–48, 52, 56n. 10; expectations for, 55–56n. 5 North, Douglass C., 149 North Korea, 175 Nuclear weapons, 67, 71–72 Occam’s Razor, 290, 307 OLS. See Ordinary least squares (OLS) model Oneal, John R., 44, 50, 160, 163, 171n; on PTAs, 227–28; trade and conflict data of, 45, 52, 241, 243 Openness, 195, 312; economic, 12, 16, 50, 72, 79

Open regionalism, 66 Operational indicator, 130, 212 Opportunity-costs model, 105, 118, 128–38, 143n. 24, 150–51; analytic problems and, 131; classic liberalism and, 96, 100, 133; hypothesis, 128–31, 132, 312, 313; interdependence and, 98, 99, 101; logic of, 139, 140; other implications of, 134–38 Ordinary least squares (OLS) model, 290, 295, 297, 302, 305, 306 Organization for Economic Cooperation and Development (OECD), 224 Organization of Petroleum Exporting Countries (OPEC), 154, 260 Organski, Kenneth, 195 Orphanides, Athanasios, 149, 152 Paine, Thomas, 45, 129 Papayoanou, Paul, 9, 124n. 12, 140, 179, 312 Parameter estimates, 295, 298. See also Endogenous variables Page 355 → Pareto, Wilfred, 223 Pareto improvement, 331 Particularist vs. public interest, 40 Partner dependence, 213. See also Dyads Pax mercatoria, 6, 37 Peace: democratic, 127, 152; dyadic, 194; opportunity costs of, 98–99; preference for, 227–31; zones of, 65, 73. See also Liberal peace; Pax mercatoria Penn World Table (Summers et al.), 190 Penubarti, M., 263 Perpetual Peace (Kant), 45, 55, 55n .1 Pevehouse, Jon C., 18, 142n. 10, 241, 332; on PTAs, 131, 224 Philip II, king of Spain, 137–38 Pluralist theories, 37–38, 40–41 Polachek, Solomon W., 12, 217, 241 Polanyi, Karl, 38 Policy accommodation, 180 Policymakers, 31, 33

Political agents, 66 Political context, 137 Political crisis, 112 Political disputes, 116, 122 Political-economic cleavage, 69 Political economy, 63, 79 Political entrepreneurs, 61–62, 76 Political leaders, 44, 128, 131, 136 Politically relevant dyads, 10, 162, 163, 192, 227; definition, 280 Political roots of trade, 114 Political security, 2 Political shocks, 105 Politics, domestic, 35, 181, 182 Politics, international, 2, 22 Polity III data, 191 Pollins, Brian M., 1, 45, 60, 159, 192, 339; on great power bias, 317 Power dyads, 191, 192 Power relations, 3, 4, 172n. 4 Power transition, 168, 171 Preferential trade arrangements (PTAs), 9, 15, 131, 142n. 10, 222–33; conflict and, 224–27; increased investment and, 225;MIDs and, 227, 228, 230; negotiation forums within, 226, 231; time inconsistency and, 233n. 6; trade barriers, 225; trade flow and, 113, 222, 224, 229, 231 Pre–World War II era, 213, 216, 316 Price elasticities, 12, 14. See also Markets Private actors, in trade, 32, 33, 112, 113, 121 Probit models, 290, 298, 306. See also Two-stage probit regression Protectionism, 225. See also Tariffs Proximate pairs of states, 193, 339. See also Economic geography PTAs. See Preferential trade arrangements Public interest, 40

Public officials, 3 Public policy, 48–49, 52, 54 Qualitative research, 310–20; choice of cases in, 315–18; independent variables, 310, 312–15, 317, 319–20; methodological challenges of, 319. See also Large-N statistical studies Quantitative research, 18, 282–83 Rare event logit analysis, 51, 52 Realists, 45, 140, 169, 248, 263; arguments of, 9, 33, 143n. 19; economic theory of peace and, 127, 128–29; liberals compared to, 2, 16, 32, 127, 140, 143n. 20, 168, 263 Reciprocal causation, 290–93, 297, 299, 306 Reciprocal influence, 132, 165, 242 Reciprocity, 259, 266n. 19 Reduced-form equation, 299, 305, 307n. 1 Regional coalitions, 64, 68, 69, 73, 75; domestic coalitions and, 60, 63. See also Coalitions Regional institutions, 79–80 Regression model, 289, 291. See also Two-stage probit regression Relative cost, 112. See also Costs Relative gains, 211 Relative resolve, 105 Republics, 46. See also Democracy Rescaling, 298–99 Resource allocation/extraction, 62, 331 Page 356 → Reuveny, Rafael, 18, 91, 242, 254, 262 Revisionist powers, 9, 183, 327 Revolutionary elites, 161 Ripsman, Norrin M., 13, 20 Risk tolerance of decision makers, 315 Rivalries, 277 Rosecrance, Richard, 311 Russett, Bruce M., 10, 130, 159, 189, 227–28; on interdependence and peace, 44, 50, 189; trade and conflict data of, 45, 52, 241, 243

Russia, 139, 144n. 28, 175. See also Soviet Union Sanctions. See Economic sanctions Schneider, Gerald, 212 Scholars, 180; biases of, 208–9; game theory, 133; liberal, 45, 53; Marxist and world systems, 14. See also Case studies; and specific scholars Science, normal, 97–98. See also Technology Security, 153; international trade and, 2, 3; national, 1, 33–34; vs. wealth, 122 Security policy, 6, 318, 319 Selection-of-disputes effect, 93 “Selectorate” and authoritarian states, 47 Sensitivity interdependence, 12–13 Sequential ordering, 279 Signaling, 34–35, 40, 124n. 12, 133–34, 140, 143n. 18; economic, 34, 134; theories of, 133, 141, 164. See also Costly signaling Simmons, Beth, 6, 31 Simulations. See Computer simulations Simultaneous equation models, 19, 258 Single-equation models, 290–93, 305, 306 Sino-American relations, 10, 168–71. See also United States, Chinese relations with Skålnes, Lars, 179 Small-N qualitative research, 318. See also Large-N statistical studies Smith, Adam, 45 Social science methods, 207, 325–26 Society-state relationship, 32, 35–39, 41, 69, 121, 124–25n. 20 Solingen, Etel, 60, 312 Southeast Asia, 74, 76, 227 Southern Cone, 71, 226, 227 Sovereignty costs, 75 Soviet bloc, 190, 217 Soviet Union, 113, 114, 169, 184, 241; Germany and, 177, 178

Spain, and war with England, 137–38, 144n. 25 Specialization, benefits of, 128 Stability, 21, 63, 132 State, theories of, 31–41; autonomous state, 33–35, 38, 40, 41nn. 2–3; need for, 31–32; society-centered, 35–39; vulnerability, independence and, 34, 36, 39–40 State actors, 6. See also Actors Statecraft, 111, 118, 175 State leaders, 45, 46–48, 149; nondemocratic, 47–48, 49, 50, 54, 55–56n. 5, 150; war-handling abilities of, 149–50, 152, 153. See also Political leaders State level of analysis, 5, 175 State pairs. See Dyads State policy, 122. See also Public policy States: conciliatory, 329; proximate, 193, 339; sensitivity of, 122; status quo and, 9, 183, 329 Statesman’s Yearbook, The (Epstein), 190 State-society relationship, 32, 35–39, 41, 69, 121, 124–25n. 20 State-specific composite economy, 327 Statistical analysis, 20, 257–58. See also Endogeneity, in statistical analysis Statistical models, 254. See also specific models Status quo states, 9, 183 Stein, Arthur A., 8, 164, 313 Stochastic error term, 291, 293, 329 Stocks, 105 Strategic calculus, 120 Strategic interactions, 5, 123, 313 Strategic trade goods, 280 Strategic variables, shifts in, 99 Strikes, 117 Structural-equation models (SEMs), 291, 301 Page 357 → Structuralist arguments, 96

Structural linkage, 180 Subnational level of analysis, 5, 6, 60 Subnational/supernational actors, 6 Subsistence level, in simulation, 331 Symmetric dyads, 93 Symmetric interdependence, 320. See also Asymmetric interdependence Symmetries, informational, 101, 117, 120 Systemic level of analysis, 6, 21, 127, 128, 139, 175 Taiwan, 165, 185 Tariff reduction agreements, 115. See also Trade-conflict dynamics, trade agreements Tariffs, 113, 123n. 5, 148, 152, 223 Technology, 49, 333, 339 Temporal dynamics, 10, 273–84; causes of conflict, 274–77; dependence on, 19, 165; heterogeneity and, 274, 279–82; of repeated conflict, 277–79 Territorial disputes, long-term, 94 Theoretical diversity, 209 Third-party states, 139, 218 Third World, 216. See also Developing countries Threats across data sets, 245–47 Time series, 164, 202. See also Temporal dynamics Timing of conflict, 340 Timpone, Richard J., 19, 289 Tough state, 329, 332 Trade: bilateral, 113, 196, 200; channeling of, 113–14, 143n. 23; coalitions and (see Coalitions); concentration of, 194; dependence on, 50, 116, 218, 314; diversion of, 119, 218; dyadic, 52, 134, 136, 212–13, 330; elasticities of, 39; private actors in, 32, 33, 112, 113, 121; and probability of war in democracy, 54; recovery of prewar, 136; right measure of, 160–61; as signal, 89–94, 117–18; strategic interactions/reactions and, 120–21 Trade asymmetry, 12, 123–24n. 10 Trade barriers, 113. See also Tariffs

Trade blocs, 223 Trade-conflict dynamics, 116–17, 122–23, 250; channeling of, 133–14, 143n. 23; coercion in, 111–12, 115–16, 117; comparing studies, 156–57; controversies in, 156–65; dependent variables in, 163–65; mediating or confounding, 161–62; reciprocity in, 132, 242; theoretical orientation, 208–11; trade agreements, 112–13 Trade-conflict dynamics, measurement of, 211–21, 239–51; data limitations in, 215–18; dyadic trade flow and, 212–14; theoretical differences in, 219. See also COPDAB; Militarized interstate disputes; WEIS data set Trade-conflict simulations. See Computer simulations Trade costs, 327–28, 329, 336 Trade data, 14, 216 Traded-goods sector, 38 Trade diversion, 119, 218 Trade flows, 4, 14, 239; conflict and, 78; dyadic, 212–13; minor, 216; PTAs, 113, 222, 224, 229, 231; ratio of, 13; signaling and, 90–91; symmetric, 93 Trade/GDP ratio, 36, 41, 160, 167, 202n. 1; bilateral, 190, 191, 196, 199, 228–29; of China and United States, 170; inadequacies of, 320; of less-dependent state, 161–62; liberal peace and, 194, 195–97, 198; total, 160, 215 Trade openness (TO), 72, 79 Trade organizations, 6 Trade-promotes-peace hypothesis, 129–30, 139. See also Kantian hypothesis Trade-related threats, 164 Trade-reliant groups, 19 Trade sanctions, 90, 92, 94. See also Economic sanctions Trade-to-output ratio, 12 Trade wars, 115, 129 Trading networks, 324, 325, 326, 330; defined, 327. See also Economic geography Trading relationship, 136 Page 358 → Trading with the enemy, 134, 135, 138, 139, 143n. 20. See also Trade-conflict dynamics Transaction costs, 65 Triangular/hierarchical model, 291 Two-stage conditional maximum likelihood (2SCML), 297 Two-stage least squares (2SLS) method, 294, 295, 296, 298, 303 Two-stage probit regression (2SPR), 296, 297, 298–300, 303, 307

Uchitel, Anne, 311 Uncertainty in international affairs, 101, 121 United Nations agency, 74 United States, 115, 116, 123n. 5, 148, 175–76, 183; British trade and, 138, 171; Chinese relations with, 10, 159, 168–71, 180, 184–85; economy, 325; government, 255; Russian relations with, 182; Soviet relations with, 1, 114, 179, 183, 185, 198, 241 Utility function, in computer model, 327 Variables, economic, 163. See also Endogenous variables; Exogenous variables Variance-corrected approaches, 278, 279 Vector autoregression (VAR), 201, 300 Voters, median, 35–36, 37, 40, 55n. 2 Vulnerability, economic, 12, 13, 34, 36, 39–40; and relative gain, 16 Waltz, Kenneth, 4, 240, 248 War, 99, 100; avoidable, 149; calculation of spoils of, 329–30; cause of, 137; costs of, 89, 138; deterrence of, 89, 99; domestic support for, 138; uncertainty reduction and, 102, 137. See also Military; and specific wars Ward, Michael D., 197, 198, 263 War of 1812, 138 War of the Spanish Armada (1585–1604), 137–38 War zones, 67 Wealth, 121, 122, 123n. 2 Weber, Max, 33 WEIS (World Event Interaction Survey) data set, 17, 18, 242–47, 248–50, 256; compatibility with other data sets, 260; conflict and cooperation in, 254–55; interstate actions coded in, 242–45, 246; use of force/threats in, 245–46, 247; weighting of events in, 257 West German/Soviet relations, 177, 178 William of Occam, 290, 307 Window Subseries Empirical Variance (WSEV), 192, 199 World Event Interaction Survey. See WEIS data set World Trade Organization (WTO), 170, 223 World trading system, 333. See also Global markets World War I, 4, 9, 190, 198, 223 World War II, 8, 56–57n. 11, 223, 316