FIB Papers
W issenschaftszentrum Berlin für Sozialforschung
Veröffentlichungsreihe der Forschungsgruppe Internationale Beziehungen
P 93-301
Democracy and the Peaceful Settlement of
International Conflict
by
W illiam J. Dixon
Publication Series of the International Relations Research Group Reichpietschufer 50
D 1000 Berlin 30
ABSTRACT
The research reported here develops an explanation for the well-known democracy-war results based on universal democratic norms for reconciling competing values and interests. I show how these norms can be extended beyond national borders to generate novel propositions about international disputes between democracies. One key implication o f this explanatory logic is that democracies are likely to be more amenable than others to the peaceful settlement o f their disputes.
This is the general hypothesis examined in the present study. A sample o f strictly interstate disputes acquired from the Alker-Sherman disaggregated conflict set provide the basis for this inquiry. Democracy is assessed for each disputant party with the composite index from the Polity II data collection. Because the dependent variable in this analysis is a binary indicator, I employ probit regression to estimate the effects o f democracy while partialling out a number o f other factors serving as controls. The empirical results show that democracy does carry the systematic positive influence on the probability o f peaceful settlement expected o f it.
ZUSAMMENFASSUNG
Die Arbeit, die hier vorgestellt wird, entwickelt eine Erklärung für die bekannten Untersuchungsergebnisse zum Thema Demokratien und Krieg, die besagen, daß Gesellschaften, die auf universellen demokratischen Normen basieren, rivalisierende W erte und Interessen eher miteinander versöhnen können. Ich zeige auf, wie diese Normen jenseits nationaler Grenzen wirksam werden können.
Dadurch werden neue Sichtweisen auf internationale Konflikte zwischen Demokratien gewonnen. Ein Schlüsselargument besteht darin, daß Demokratien einer friedlichen Konfliktlösung eher zugänglich sind, als andere Systeme. Diese Hypothese steht in der vorliegenden Arbeit im Mittelpunkt. Eine Stichprobe von rein zwischenstaatlichen Konflikten aus dem "Alker-Sherman disaggregated conflict set" bildet die Basis für diese Untersuchung. Demokratie wurde für jede Konfliktpartei nach dem zusammengesetzten Index aus der Polity II Datensammlung bewertet. D a die abhängige Variable in dieser Untersuchung ein binärer Indikator ist, habe ich Probit Regressionsanalysen für die Schätzung der Effekte von Demokratie verwendet, während eine Anzahl anderer Faktoren als Kontrolle dienten. Die Ergebnisse zeigen, daß Demokratie in der Tat einen positiven Einfluß auf die W ahrscheinlichkeit einer friedlichen Konfliktlösung hat.
Introduction
Nearly 200 years ago Immanuel Kant envisioned a liberal democratic zone o f peace resulting from what he thought to be an inexorable evolutionary trend toward republican government. Kant’s evolutionary ideas seem outmoded today, but his vision o f democratically governed states coexisting without resort to war appears remarkably prescient. Empirical evidence gathered over several decades has consistently shown that while democracies are no less prone to foreign conflict and war than states organized under alternative arrangements, democratic states rarely engage in violent conflict against one another. Unfortunately, there is considerably less consensus about why democracies behave this way.
The research reported here treats the virtual absence o f war between democracies as a point o f departure. Thus, the question posed here is not whether democracies fight one another, but rather why don’t democracies fight one another? Based on theoretical reasoning advanced below, I offer an explanation based on shared norms for regulating competition and conflict that are integral to the democratic process. This explanatory logic is not only consistent with the published findings on democracy and war, it also entails additional implications that have thus far largely escaped empirical investigation. One implication is that disputes between democracies are likely to be more amenable than others to peaceful settlement. This is the hypothesis I examine in the present study.
The next section briefly reviews previous empirical work relating democratic governance to foreign conflict and then develops the main theoretical argument motivating the research.
This is followed by discussion o f the research design and analytic procedures. The empirical domain is a worldwide sample o f approximately 250 postwar interstate security disputes compiled by Alker and Sherman (1986) from a list originally developed by Butterworth and Scranton (1976) and others. The next section presents empirical estimates evaluating the peaceful settlement hypothesis. This is followed by a concluding section discussing the significance o f the study and future research directions.
Democracy, Conflict, and Peaceful Settlement
Extensive research on the relationship between democracy and international conflict has exposed two empirical patterns that when taken together pose a most interesting puzzle. First, democracies are no less likely to become involved in foreign wars, crises, or potentially violent disputes than states organized under alternative governing arrangements (W right 1942; Russett
and Monson 1975; Small and Singer 1976; Chan 1984; Weede 1984; Domke 1988; Maoz and Abdolali 1989). The weight o f evidence thus suggests that democracy per se does not entail any discernable pacifying effect on foreign policy.1 Yet a second and equally compelling pattern o f results shows that democratic states rarely, if ever, fight wars against one another (Babst 1972; Small and Singer 1976; Rummel 1983; Chan 1984; Weede 1984,1992; Doyle 1986; M aoz and Abdolali 1989; Ray 1993). In fact, democracies seldom even engage one another with threats o f military violence, and when disputes do arise they hardly ever result in military hostilities (Maoz and Abdolali 1989; M aoz and Russett 1992a; Brem er 1992).
These empirical results have come to be known as the "democratic peace".
There is little doubt about the existence o f the democratic peace; on this the historical record has been manifestly clear. W hat is not yet settled is why democratic states behave in this curious way. Efforts to explain this phenomenon have identified a variety o f causal mechanisms broadly based either on common democratic norms, values, and practices—such as conceptions o f legitimacy or conciliatory means o f dispute resolution—or on the institutionalized impediments to war and conflict imposed by democratic decisionmaking structures (e.g., Doyle 1986; Russett 1990; M erritt and Zinnes 1991; Bueno de M esquita and Lalman 1992; Ember, Ember, and Russett 1992).2 M uch recent work on the democratic peace has attempted to articulate competing claims o f each approach and to adjudicate them on empirical grounds; here, however, the historical record has proved much less accommodating, yielding ambiguous and generally inconclusive results (e.g., M organ and Campbell 1991;
M organ and Schwebach 1992; Maoz and Russett 1992b; W ang, Chang, and Ray 1992).
It may well be that the empirical evidence is ambiguous because the theoretical arguments are themselves ambiguous. As Morgan and Schwebach (1992,318) aptly point out, "A nation imbued with a democratic culture [i.e., norms] will likely establish a correspondent political system and a state structured to constrain will likely foster a democratic culture." In other words, norms and institutions go hand in hand and any effort to untangle their causal implications faces serious obstacles. The problem is exacerbated by the fact that most o f these studies employ a testing ground based on established patterns o f violent conflict between democracies, that is, on the democratic peace itself.3 But if this phenomenon is already
"overdetermined" by overlapping theories, as Ember, Ember, and Russett (1992,577) suggest, then the search for empirically grounded explanation may be more productively conducted elsewhere. This is the strategy I adopt in the present study.
My search for a satisfactory explanation o f the democratic peace takes the following form. First, I develop an explanatory logic showing how democratic norms for regulating competition and conflict can account for both sides o f the democratic peace results. This logic is then further elaborated to encompass new empirical propositions about democracy, conflict management, and the settlement o f disputes. Finally I undertake an empirical test o f the key hypothesis that democratic states are more likely to achieve peaceful settlements than are other types o f disputants.4
Democracy and Norms of Bounded Competition
To understand what is distinctive about the international behavior o f democratic states it is necessary to begin with a clear conception o f democracy.5 Definitions abound, but most converge on the central role o f certain procedural characteristics, including, but not limited to, fairly contested elections, virtually universal suffrage, restrictions on government power, accountability o f political leaders, and public expression o f preferences (Lipset 1959; Dahl 1971; Powell 1982). Yet identification o f common governing procedures such as these in no way implies that all democracies must necessarily share common political institutions. "There are many types o f democracy," observe Schmitter and Karl (1991, 76), "and their diverse practices produce a similarly varied set o f effects." Nor do all democracies necessarily share a broad array o f core values. Certainly there are some values common to virtually all democracies, but there are deep differences as w ell-w itness, for example, the divergence between social-democratic and classical liberal conceptions o f democracy.
My purpose in emphasizing the diversity o f democratic norms, values, and institutions is to underscore the need to identify as precisely and narrowly as possible those features o f democratic systems that are both universally shared and causally relevant to foreign conflict behavior. It will not do to explain the democratic peace in terms o f a broadly conceived liberal ideology or a democratic political (or civic) culture. Concepts such as these are too strong to encompass all modern democratic forms and at the same time too weak to carry the explanatory load required o f them.
W hat specific features o f democracy meet these criteria? W ithout minimizing the significance o f other defining characteristics or essential procedures, I will argue that special emphasis should be given to the universal trait o f bounded competition. All modem democracies are openly competitive systems o f governance where conflicting material interests and basic political values routinely clash over the proper course o f public action.
Sometimes the competition is fierce and the dissention deep, sometimes the differences are less than profound, but always there are competing interests and values. Elections are one important arena o f democratic competition dealing with what is perhaps the m ost fundamental question o f government—who will rule? But it would be a mistake to conclude that the full measure o f democratic competition is embodied in intermittent elections (Schm itter and Karl 1991). Competitive processes may take a wide variety of forms and involve many types of actors, including the political elites who legislate and execute policy at all levels and the citizens who organize and express their interests through a bewildering array o f associative arrangements.
Just as competition is a constant o f democratic governance, so too is the presence of rules, procedures, or guidelines for setting its boundaries. It may be a commonplace to characterize democracy as less a battleground than a playing field, but this juxtaposition of images clearly conveys the distinctively circumscribed nature o f democratic competition.
Democracy is feasible only so long as politicians and citizens implicitly agree on the delineation o f these boundaries and constrain their actions accordingly. Some boundaries and rules may be universal while others vary from society to society, but all rest on a form o f exchange that Schmitter and Karl (1991) call "contingent consent." To take the most ubiquitous example, losers o f elections are expected to acknowledge the winners’ right to govern and to abide by their policy decisions, but only so long as the election was fairly conducted and the winners’
fulfill their obligation not to ban losers from future participation. W hile contingent consent does not deny to democracy a political culture based on deeply held values such as tolerance, compromise, and fair play; neither is democratic culture a necessary condition for contingent consent. Rather, the principle o f contingent consent highlights the emergence o f "operative guidelines" or "rules o f prudence" as the glue holding democracy together (Schmitter and Karl 1991).
I have already noted how contingent consent bounds democratic competition for political office. This is illustrative o f a more general pattern o f bounded competition common to all democracies—namely, that political actors, whether inside or outside o f government, agree not to employ physically coercive or violent means to secure a winning position on contentious public issues. This is not to say that modem democracies never experience politically motivated violence; such a position would be untenable from the start. My claim is only that the prevailing mode o f democratic competition, particularly among politicians and other political elites, is nonviolent. Nor is it the case that democratic states never resort to coercive means within their borders. All states retain exclusive rights to coercion, but democratic states do not ordinarily
use coercion for day-to-day management o f political competition, though they may do so under extraordinary circumstances. Furthermore, the contingent consent to eschew violence is by no means immutable, and under some conditions it may even be very fragile and tenuous.
Democracy is known to break down. Finally, the claim that democratic actors compete nonviolently in no way implies that they always compete fairly, or nobly, or even legally.
Democracy is as much an ongoing process as it is a static form o f government, and the process o f democracy is fundamentally competitive. W hen factions and interests collide in the public realm, as they necessarily must, the resulting conflicts are regulated by contingent consent to follow familiar procedures and practices. The particular form o f regulation varies widely from country to country and issue to issue—it may be political or juridical, formal or informal, routinized or ad hoc, authoritative or negotiated-but always it is nonviolent and noncoercive. Although the immediate outcomes o f such procedures are by no means certain, unregulated competition and potentially violent conflict risk even greater uncertainty.
Contingent consent implies that within democratic societies political actors will prefer to follow nonviolent regulatory procedures and will expect competing actors to do likewise.
W hen actors’ preferences and expectations for regularized behavior are widely generalized and reinforced they frequently become redefined as not merely preferred and expected but also as right and proper, that is, they become norms (Opp 1982; Axelrod 1986).
Admittedly, the preceding discussion paints a thin picture of democracy, but one sufficient to highlight an essential point. All m odem democratic societies foster the internalization o f norms for regulating and reconciling competing interests and values in public affairs in ways that are neither violent nor coercive. For convenience I will designate this the democratic norm o f bounded competition. I follow conventional usage by stipulating norms to be commonly held expectations of acceptable behavior within society. Norms thus have an evaluative component delineating what is regarded as acceptable and a behavioral component encompassing rules or procedures that are regularly followed. M y conception o f this universal democratic norm is both proscriptive and prescriptive. First, it narrowly proscribes as inadmissable any use o f violent, coercive, or forceful means to prevail in contentious public issues or to gain advantage over political opponents. Second, it promotes use o f institutionalized procedures for regulating political competition and resolving contested issues, though I leave necessarily vague the precise structures or mechanisms prescribed in order to accommodate the full range o f democratic forms.
Although the norm o f bounded competition pervades all segments o f democratic society, it is likely to be more firmly rooted among politicians and political elites than elsewhere. The competition o f interests and values I have argued to be inherent in democracy is actually manifested as competition among the political elites who are responsible for articulating and advocating representative policy positions and political preferences. Political elites thus have more opportunities to act in accordance with the norm and to benefit from its operation, which together serve to strengthen its internalization (Opp 1982). Strongly internalized norms are powerful influences on behavior, but they are not ironclad laws. Deviations may be rare—else we would not speak o f norms—but they surely do occur, which is one reason why democracies so often institutionalize at least some areas o f bounded competition by establishing constitutional or other legal mechanisms along with measures for enforcement.
Democratic Norins and International Conflict
All governmental decisions and policies, including those relating to war and peace, are ultimately made by individuals empowered by institutional roles to act in the name o f the state. Recruitment o f individuals to these role positions is highly variable from state to state, democratic or otherwise. Nevertheless, the institutional structures o f democratic states usually entail a recruitment path that limits such positions to seasoned politicians, career government officials, or experienced elites who are accustomed to operating in the public realm within the standards o f bounded competition. Furthermore, regardless o f precipitating events or contested issues, interstate conflicts and wars arise for much the same reason as domestic political disputes: over competing or incompatible material interests and basic values. This is not to imply that foreign policy is nothing more than an extension o f domestic politics; the former is sharply differentiated by fundamental concerns for national sovereignty, security, identity, and the like (though such issues can also penetrate domestic affairs). All I mean to suggest is that the competition o f interests and values in foreign affairs is not wholly unlike the competition inherent in the democratic process.
The international disputes o f democratic states are thus in the hands o f individuals who not only have experienced the politics o f competing values and interests, but who have consistently responded within the normative guidelines o f bounded competition. Even so, democratic political leaders do not automatically externalize norms governing domestic politics to international affairs. So facile an assumption would ignore the fact that norms are rooted in a social setting. I have argued at some length that the democratic norm o f bounded
competition operates in a setting characterized by contingent consent and the mutual expectations o f competing political actors. More specifically, the norm ative guideline proscribing violent or coercive action to secure political advantage is not some general predisposition, it applies only when there also exists strong expectations that one’s opponent will also eschew violence and coercion.
Democratic states confronted by the actual or threatened use o f military force from foreign powers will thus show little reticence to respond in a similar fashion. In fact, such instances are probably governed by norms o f reciprocity that demand a m ilitary response (Axelrod 1986). Nor are democracies unconditionally constrained from initiating military hostilities against others. In short, the norm o f bounded competition does not imply that democracies are, on the whole, any more peaceful or any less war prone than nondemocratic states.
But what o f disputes in which both parties are democracies? This is a special case not only because both sides subscribe to norms o f bounded competition, but also because leaders o f each party are fully cognizant that bounded competition is the norm, both for themselves and their opponents. Democracy, after all, can not be conducted in secret. Thus, when democracies confront other democracies, both sides are secure in the knowledge that its opponent is also normatively proscribed from violent and coercive means for reconciling divergent values and interests, at least in the domestic political arena.6 These norms are externalized to foreign affairs only if the practices underlying them are sufficiently robust to foster mutual expectations o f congruent behavior beyond borders. But it is not enough that democratic leaders merely expect their foreign opponents to forbear violent pursuit o f their interests or values. The expectation o f nonviolence must be accompanied by a belief that the opponent defines forceful means as illegitimate behavior under the circumstances and therefore omits it from serious consideration. This critical condition prevails only when democracies confront other democracies.
Disputes between democratic states seldom escalate to violence, but not because norms o f domestic conduct generate peaceful predispositions or are externalized willy-nilly; it is because only in this special case is there a basis for contingent consent. A normbased explanatory logic thus accommodates both parts o f the democratic peace result because it establishes an expectation o f the absence o f militarized conflict only in cases where both parties share the democratic norm o f bounded competition.7 But the absence o f violence is only half the story. To claim that norms effectively eliminate force as a policy option is to say
nothing about the options that remain, about how disputes are handled. The distinctive advantage o f a norms-based account is that it can also explain what democracies do in disputes with other democracies.
Democratic Norms and Peaceful Settlement
Thus far I have emphasized only the proscriptive side o f bounded competition. There is also a prescriptive element that encourages reconciliation o f competing political interests and values, though the available means are too varied across both countries and issues to permit identification o f any single procedure or formula as the democratic norm. For example, democratic leaders are sometimes portrayed as being willing to compromise in foreign affairs because they are accustomed to doing so on domestic matters. No doubt political bargaining and compromise are common regulatory mechanisms in democracies, but it would be a mistake to assume all issues are resolved this way. Some types o f competition m ay require legal resolution while others may be decided by raw political influence with little give and take.
Furthermore, to regard democratic elites as inherently open-minded or willing to compromise is to overlook the fact that even committed democrats can be extraordinarily intolerant and intransigent from time to time.
W hile democracies can never be assured o f quick, easy, or successful resolution o f their disputes with other democracies, the mutual norm o f bounded competition does at least provide a basis for contingent consent. This suggests that leaders o f democratic states are likely to explore a wider range o f alternative modes o f dispute resolution with other democracies than with nondemocratic opponents. Frequently the preferred approach o f disputing democracies will coincide—for example, with both sides defining the contested issue as a legal matter calling for a ruling by an international tribunal, or with each party realizing the other is willing to concede enough demands to permit a compromise solution. Even when democratic states do initially bring diverging interpretations and approaches to bear on their disputes, their common adherence to bounded competition is likely to increase the chances o f some procedural accommodation.
If democratic norms o f bounded competition operate as I have described, then disputes between democracies should unfold somewhat differently than those between states not sharing these norms (i.e., between democratic and nondemocratic states or between two nondemocratic states). Two such differences-one procedural, one substantive-are of sufficient importance to warrant treating them as distinctive and novel empirical propositions.
The procedural proposition contends that democratic states will be more likely than others to attempt resolution or amelioration o f their disputes by arrangements secured with the assistance o f third party intermediaries. The substantive proposition carries the logic o f bounded competition a step further, claiming that democratic states will be more likely than others to actually succeed in achieving mutually agreeable settlements. Until now only the former hypothesis has been subjected to systematic empirical investigation; the latter is examined here for the first time.
Third party intermediaries may consist of other states or coalitions o f states, international organizations, transnational or subnational organizations, or even individuals with international standing; together we refer to them as conflict management agents. The intercession o f third parties has a long history in international relations and consequently has evolved a wide variety o f forms. Sometimes intermediaries play a decisive role in resolving conflicts, for example by formally adjudicating or arbitrating contested issues, while at other times they serve more as supporting players, perhaps proposing compromise solutions or merely extending good offices support. In most cases, though by no means all, the involvement o f conflict management agents hinges on the willingness o f both disputant parties to entrust in them at least some measure o f responsibility for shaping valued outcomes. W hile democratic leaders may not always acquiesce to the intercession o f third parties, they are probably inclined to do so under a broader range o f circumstances than leaders lacking norms o f bounded competition. For this reason, purely democratic disputes are more likely than others to involve some form of third party management.
Recent evidence indicates that the democracy-management hypothesis is not simply airy conjecture. A systematic study o f management practices over some 250 postwar interstate disputes has shown that the more democratic the disputant parties, the more likely they were to accede to the involvement o f conflict managing agents (Dixon 1993). Moreover, this finding proved exceedingly robust even when controlling for potentially confounding factors such as the intensity o f the conflict and the disputants’ power and level o f economic development. As a corollary to the basic hypothesis, Raymond (1992) proposed that democracies would be more amenable than others to uses o f binding arbitration as a means for settling disputes.
Although his analysis covered a much longer time span, from 1816 to 1965, it was limited to major power disputes that were referred to third parties for settlement. Using both bivariate and multivariate procedures, Raymond found strong empirical support for the proposition.
W hile definitive conclusions are as yet unwarranted, these studies do offer clear and consistent evidence o f the hypothesized relationship between democracy and conflict management.
The logic o f bounded competition further suggests that democracies are not merely more acceptant o f management than others, they are also better positioned to actually settle their disputes. There are two ways to justify this assertion, each resting on a som ewhat different causal mechanism. One is evident in the empirical findings o f most conflict management studies (Butterworth 1978; Haas 1983; 1986; Bercovitch, Anagnoson, and W ille 1992; Miall 1992). This body o f research generally shows that third party management efforts in the postwar era have the greatest probability o f success when conflicts are neither ideologically divisive nor intensely violent. Disputes between democracies meet both criteria o f successful management: they are obviously not ideologically motivated and the empirical record demonstrates that they are seldom violent. By this account, then, democratic norms have only an indirect effect on peaceful settlement that operates through the intervening presence of conflict management. Because democratic norms are one step rem oved from directly influencing dispute resolution, I refer to this rationale as the indirect democracy-settlement hypothesis.
Democratic norms may also carry a direct impact on the probability o f settlement quite apart from any management effect. This follows from a straightforward extension o f preceding arguments. W hen democracies confront other democracies each side is willing to abstain from violent pursuit o f its interests in favor o f peaceful solutions because each fully expects that its opponent will do likewise. Again, mutual norms o f bounded competition provide no assurance that opposing democracies will always reach a resolution or even concur on how their disputes should be handled (other than nonviolently). Nevertheless, because leaders with democratic norms will be inclined to seek some form o f peaceful resolution so long as they can trust their opponent to abide by the provisions o f a settlement, they will be most likely to succeed when confronting other states whose leaders operate under similar normative guidelines. Although this logic does not deny the potential contribution o f third party managers, it does suggest that democratic norms directly enhance the prospects o f peaceful settlement even in the absence o f management efforts. Accordingly, I term this proposition the direct democracy-settlement hypothesis.
These two hypotheses comprise the core o f the present study by building on assumptions o f democratic norms to make empirical claims about the peaceful settlements o f disputes. The direct democracy hypothesis posits the most causally proximate effect by assuming that democratic leaders locked in disputes expect a peaceful resolution and are prepared to work directly with one another toward mutually agreeable settlements. The indirect democracy hypothesis depicts a rather different mechanism whereby democratic norms encourage the
acceptance o f third party efforts to manage conflict and it is these efforts that bear the fruit of peaceful settlement. In addition, it is at least plausible that the causal mechanisms underlying these hypotheses combine nonadditively to reveal mutually amplifying effects. That is, conflict management may be more successful between democracies and, similarly, the direct democracy effect may be more potent when accompanied by management.8
Also note that the two hypotheses make empirically independent claims—one, both, or neither may represent an accurate reflection o f the historical record. The strongest support for the norm-based perspective outlined above would result from findings consistent with both hypotheses. In particular, such an outcome would require evidence o f a direct democracy effect improving the probability o f peaceful settlement even while controlling for the expected impact o f conflict management. Alternatively, it is entirely possible that both claims are inconsistent with the historical record, an outcome that would severely undermine a norm-based explanatory logic.9 Finally, evidence for only one hypothesis would offer partial support, though in the absence o f a direct democracy effect even a partially supportive outcome would raise doubts about the efficacy o f democratic norms. Having laid out a general theoretical rationale and my principal research hypotheses, I now turn to an empirical evaluation focusing on dispute settlement in the post W orld W ar II era.
Research Design
This examination o f the democracy-settlement hypotheses requires a sample of international disputes over some reasonably long time span along with information about the governing arrangements o f the parties to these disputes. For this initial look at the hypotheses it is necessary only to determine whether or not some form o f settlement was achieved by the disputant parties. Since it is unlikely that democracy and conflict management alone can fully explain the occurrence o f peaceful settlements, some suitable control variables m ust also be identified to avoid potential confounding effects. Each o f these components o f the research design is discussed below.
Conflicts and Conflict Phases
The conflict sample is drawn from an evolving list o f post-W orld W ar II interstate political and security disputes that for two decades has served as a principal source for systematic studies o f international conflict management. Conflict situations represented in the
data base were originally identified by Haas, Butterworth and Nye (1972) for their pioneering study o f international organization management practices. Coded attributes o f the disputes and a collection o f synopses for a modified set o f 247 conflicts updated though 1974 were published by Butterworth and Scranton (1976), with a subsequent update through 1979 by Butterworth adding another sixty conflicts. Further updates have been undertaken by Haas (1983,1986), though here I examine only the 1945 to 1979 period from the files compiled by Alker and Sherman (1986).10
The Alker-Sherman set has several specific advantages for the present study. First, as Alker and M idlarsky (1985) note, the empirical domain is defined so as to correspond to international diplomatic and legal practice. In following Butterworth (and Scranton 1976,3), the encoding procedures " . . . included only postwar conflicts that centrally involved specific power-political aims and demands having direct impacts on national behavior, and that were perceived internationally as being focused on political and security affairs. Excluded are domestic issues that did not directly lead to interstate conflict, together with more diffuse antagonisms or instances o f general interstate competition." Second, this particular version of the data incorporates a dynamic element missing from other collections by disaggregating conflict situations into distinct phases based on a schema developed by Bloomfield and Leiss (1970; Alker and Sherman 1982). And third, the actors on the contesting sides o f each dispute are itemized individually during each separate phase o f the conflict. These latter two features are so distinctive that the Alker-Sherman data are sometimes designated the "phase/actor disaggregated" conflict set. They are also o f critical importance to the present effort.
International disputes are not static situations that remain fixed at the same degree of intensity and antagonism. M ost disputes unfold through a series o f stages or phases that are distinguished by the seriousness o f the conflict and its potential for military hostilities. For this reason Bloom field and Leiss (1970) proposed a classification scheme marking six distinct phases through w hich a conflict may proceed until it is settled or disappears. In recognition that conflicts do not evolve along a single trajectory, the scheme is explicitly nonlinear and iterative by permitting phases to be skipped or repeated as necessary. Phase types include initial disputes (I) or disagreements; conflicts (II) in which at least one party exhibits some commitment to military force; hostilities (III) during which military force is actually used;
post-hostility conflicts (IV) where fighting has ceased but threats o f force remain; post-hostility disputes (V) in which military resolution is no longer a serious possibility; and termination (VI) in which the dispute is resolved or no longer o f concern to the contesting parties.11
Alker and Sherman’s key insight was that transitions from one conflict phase to another may also entail changes in certain other attributes o f the conflict situation. For example, specific actions taken by the parties to the dispute may vary dramatically from phase to phase, and even the parties themselves m ay change, though this is relatively rare and usually involves an additional actor joining one side or the other. Indeed, even the presence or absence o f conflict management activity, a factor o f central importance to the present study, seldom remains constant across separate phases o f a single conflict. O f the 307 postwar conflicts in the Alker-Sherman set, only about 20% exhibit either the presence or absence o f management in every observed phase, and even this is inflated somewhat for when we consider only those 265 conflicts with multiple phases the proportion declines to only 14%. Disaggregation by phases thus affords a unique and valuable perspective for examining the shifting conditions that determine the course o f international conflict situations.
Disaggregation o f dispute phases also has the benefit o f clearing away some o f the conceptual and operational ambiguity that has thus far remained implicit in my use o f the phrase peaceful settlement. I am now in a position to stipulate a settlement to be any written or unwritten mutually agreeable arrangements between disputant parties that at least temporarily resolve or remove from contention one or more, but not necessarily all, o f the issues underlying the dispute. This definition has several advantages. In the first place, it does not carry any implication o f finality and thereby avoids the problem o f specifying a necessarily arbitrary time horizon for distinguishing final settlements from temporary ones. It also explicitly recognizes partial settlements as diplomatic achievements in which some issues, often the most contentious ones, may be deliberately and skillfully omitted from consideration in order to secure agreements. Furthermore, it effectively disentangles the act o f successfully reaching agreement from the conceptually distinct task o f implementing its terms.
Finally, this definition is operationally well-suited to the disaggregated phase structure introduced above. Settlements represent mileposts o f stability and reduced tensions along the unfolding course o f a conflict situation. Any subsequent change in the situation—whether it be toward still further reductions in tension or toward renewed hostilities—would trigger the start o f a new conflict phase. O f course some settlements do succeed in bringing about a full and final resolution that terminates the conflict; others merely mark the boundaries o f phase transitions. For this reason the conflict phase serves as an especially useful observational device for the study o f peaceful settlements.
M y unit o f analysis is thus the individual conflict phase with peaceful settlement represented by a binary indicator variable that assumes a value o f 1 if some agreem ent was concluded during a given phase and 0 otherwise. I define peaceful settlement as above with no distinction between comprehensive and partial agreements. The latter category may includes cases o f agreements that are deliberately crafted around some unresolved issues as well as those in which transpiring events make certain issues obsolete or subordinate them to new issues. Because I stipulate peaceful settlements to be mutually agreeable arrangements, I do not count those instances in which settlements were imposed with the defeat or collapse o f one disputant party. In a few very unusual cases partial settlements occur even as the conflict escalates or spreads; these, too, are omitted. By this definition peaceful settlements are neither exceptional achievements nor politics as usual; they occur in about 31% o f the 927 conflict phases contained in the Alker-Sherman disaggregated set.
Conflict management, which also plays a key role in this inquiry, is defined broadly to encompass a wide range o f activities undertaken by third parties with the aim o f resolving or ameliorating interstate disputes. Skjelsbaek’s (1986) seven-fold typology o f conflict management practices, modified slightly to extend beyond intergovernmental bodies, offers a useful organizing device. Accordingly, I count as management any one o f the following actions by some third party: public appeals or resolutions directed at the disputants; facilitation o f communication between disputants; mediation or arbitration; on-site observation or field investigation o f the situation; peacekeeping or intervention; humanitarian assistance to one or both parties; and formal adjudication o f the disputant’s claims. All o f these activities occur repeatedly during the period o f observation, though they vary substantially in frequency of use with communication, mediation and observation the most common forms o f management.
For this investigation it is not necessary to distinguish among specific management activities but merely to record the presence o f any such activity during each phase. Conflict management is somewhat more prevalent that peaceful settlement, with one or more management activities present in 43% o f the Alker-Sherman conflict phases.
Democracy
A second key feature o f the Alker-Sherman disaggregated conflict set is its identification o f specific actors comprising the primary parties during each dispute phase. The two sides in
the dispute are arbitrarily designated parties A and B as an organizing device. W ith the identification o f disputant actors it becomes possible to take explicit account o f their governing arrangements.
W hile I cannot directly measure democratic norms o f bounded competition, I can gauge the competitiveness o f national political institutions using the eleven point democracy scale predefined in the Polity II data collection (Gurr, Jaggers, and M oore 1989).12 This instrument is constructed as a weighted summative index combining judgm ental codings o f four national level authority traits recorded annually for 134 nations through 1986. Although space does not permit a full presentation o f this scale, I can briefly note its four components and their weights. These include competitiveness o f political participation (with competitive worth 3 scale points, transitional worth 2, and factional 1); competitiveness o f executive recruitment (election worth 2 and transitional 1); openness o f executive recruitment (open o rpartial election worth 1); and constraints on the chief executive (parity or subordinate worth 4, substantial limitations 2, with two intermediate categories awarding 3 and 1). Because the notion o f democracy underlying this scale emphasizes political institutions, I follow Gurr, Jaggers, and M oore (1991) in referring to it as "institutionalized democracy."
The appropriate institutionalized democracy score is recorded for the primary parties in each dispute phase for the year in which the phase begins. Several specific points require elaboration. First, because conflict phases may span several years, there is a possibility that governing arrangements may change within a given phase period. Visual inspection o f the data suggests that this is not likely to be a serious problem, though it does occur on occasion.
Second, because my interest is in assessing the effect o f institutionalized democracy on peaceful settlements, I confine my attention to cases with formally recognized national political institutions. These are cases o f strictly interstate disputes, that is, phases in which each party includes at least one nation state represented in the Polity II collection. This filter eliminates exclusively nonstate actors such as colonial era independence movements, revolutionary groups, political factions, and the like, as well as a few states failing to meet the Polity II requirement o f at least one million population by the mid 1980s. The resulting sample encompasses 264 postwar disputes disaggregated into 718 phases o f which 33% exhibit peaceful settlement.13
But strictly interstate disputes are not necessarily bilateral and this fact raises a third point for elaboration. Nearly 40% of the remaining dispute phases encompass at least one party comprised o f multiple state actors. In assigning a democracy score to coalitions o f states
I could rely on some summary value such as the mean o f coalition members. The theoretical implications o f such a procedure are not very appealing given that a coalition comprised o f equal numbers o f highly democratic and nondemocratic states, and one o f all moderately democratic states might well receive identical mean scores implying that their behavior would be governed by equally democratic norms. A more theoretically satisfying approach to this problem rests on what I shall term the weakest link principle. I began with the assumption that democratic norms will induce states to seek out or at least acquiesce to peaceful settlements.
This implies that more democratic actors are more favorable to settlement than less democratic actors, other things being equal. In addition, it is already stipulated that settlements are mutually agreeable arrangements and thus require at least tacit approval o f the active disputants. Taken together these assumptions imply that for coalitions o f states the norms most likely to promote or discourage the potential for peaceful settlement are those of the least democratic actor--the weakest link. Accordingly, coalitions o f states are assigned a value o f democracy equal to that o f the coalition’s least democratic member.
For each dispute phase I now have measures o f democracy for each o f the two contending parties. Once again it is necessary to consider the basic theoretical process linking democratic norms to the prospects for peaceful settlement. If it were reasonable to believe that party A ’s level o f democracy affects the probability o f settlement independent o f the governing arrangements o f B, and vice versa, then it would be appropriate to estimate these effects in a model including democracy scores for each party while controlling for the other. The assumption o f independent effects seems implausible, however, given my earlier stipulation that peaceful settlements require the disputants’ acquiescence. If settlement is indeed facilitated by democratic norms in the manner hypothesized, it seems likely that both parties’
level o f democracy must come into play jointly and simultaneously.
There are several forms capable o f representing this type o f jo in t effect, but perhaps the most theoretically parsimonious procedure again relies on the weakest link principle, though here applied to the contending parties rather than within coalitions. By the same reasoning given above, the weakest link principle implies that, other things being equal, peaceful settlements are most directly fostered or impeded by the norms o f the less democratic party to the dispute. This is surely a simplification, though it does not seem unreasonable as an initial approximation. In addition, other things are seldom equal, a m atter I take up shortly.
The procedures outlined thus far yield a sample o f strictly interstate dispute phases classified according to the presence or absence o f peaceful settlements and conflict management along with two institutionalized democracy scores, one for each o f the two opposing sides. In cases where a party to a dispute is a coalition o f states I assign the democracy value o f the least democratic member. I have also assumed that it is the norms and expectations o f the less democratic disputant that affect the likelihood o f settlement during a given conflict phase. M y measure o f democracy thus simplifies to the minimum value observed for any actor directly involved in the dispute.
Determinants of Peaceful Settlement
Peaceful settlements surely depend on more than democratic norms held by the disputants and management attempts by third parties. Following a vast literature on dispute resolution, I assume that the probability o f reaching a settlement is affected by certain general attributes and relations o f the disputant parties and by historical factors specific to each conflict situation.
In particular, I take account o f the relative power o f disputants and their existing political relationship, the duration and evolution o f the conflict situation, and the short-term costs of conflict.
Just as power shapes the general contours o f an evolving conflict, so m ay it condition the prospects for peaceful settlement. Some analysts have found settlements to be more likely between adversaries o f roughly equivalent power (Bercovitch, Anagnoson, and W ille 1991).
From a realist standpoint the chances o f a state prevailing in a dispute are generally proportional to its power, other things being equal, and thus the greater the power disparity, the less likely it is that the stronger actor will acquiesce to any form o f negotiated settlement. M iall (1992, 120), on the other hand, reports "no significant association" between pow er disparity and peaceful settlement. Because power is merely a control serving to isolate the independent effects o f democracy I leave this issue to be resolved empirically. Power is m easured with the continuous composite indicator o f national capabilities compiled by the Correlates o f W ar Project (Singer 1988).14 Cases in which multiple actors comprise a primary party were assigned the highest value o f the coalition participants. Finally, I arrange the individual pow er scores so as to differentiate the stronger and weaker parties and then take the ratio o f the stronger to the w eaker as my summary indicator o f power disparity.
The prospects for peaceful settlement will also depend on the general atmosphere of political relations between the parties. Certainly states with historically friendly interactions or those that m aintain some areas o f amicable relations even while engaged in disputes will find it easier to reach settlements than actors with longstanding or broadly defined animosities (Bercovitch, Anagnoson, and W ille 1991; Miall 1992). A full assessment o f political relations is not possible here, though as an initial approximation I rely on common memberships in a formal alliance to signify the presence o f a cooperative, interdependent relationship in the important area o f national security. This is not to say that allied states necessarily agree on all issues, but surely we would expect that states choosing to ally with one another and to maintain that alliance even during periods o f conflict are better positioned to reach peaceful settlements than are nonallied states. Controlling for alliance bonds also sharpens my empirical assessment o f the direct democracy hypothesis by limiting the possibility that effects attributed to democratic norms are actually due to common alliance memberships (Ray 1989) o r to a more general level o f policy conformity (Siverson and Emmons 1991). Alliance ties (Small and Singer 1969) are recorded as a dummy variable with 1 indicating mutual membership in any type o f alliance by any state on one side o f the dispute with any state on the other.
As indicated by my reliance on phase disaggregation, conflict relationships tend to evolve over time, though not along any linear path or fixed trajectory. Even so, many analysts believe that conditions at some stages in the conflict process are more favorable to peaceful settlement than at others. Zartm an (1991) calls these "ripe moments for conflict resolution." W hile it is difficult to predict exactly when ripe moments occur, we can focus more broadly on the duration o f the conflict situation and the timing o f settlement attempts. Studying third party mediation efforts, Bercovitch, Anagnoson, and W ille (1991, 13) report that "the longer a dispute goes on, the less amenable it is to mediation; but there does seem to be a minimum amount o f time necessary before mediation is successful." W hether settlements are mediated or not seems immaterial to the basic point that peaceful settlements are least likely at both the earliest stages o f a dispute and in cases of prolonged conflict (Miall 1992). Although I am unable to measure duration directly, I can approximate this relationship by taking account o f the number o f prior phases through which each conflict has evolved. I use a simple phase counter with a 1 recorded for the initial phase o f each conflict, 2 for the second phase, and so on through the maximum o f 9 phases. To model the inverted U effect observed by Bercovitch and his associates I follow the usual practice o f including the square o f this phase counter.
Although my phase counter carries important information about the historical evolution o f the conflict situation, it cannot highlight specific turning points or phases o f particular historical significance. I therefore supplement the phase counter with an additional indicator to record the presence o f military hostilities at any previous point in the history o f a given conflict. It does seem that the obstacles to peaceful settlement become m uch more formidable once a conflict situation has passed the threshold o f major violence (Miall 1992). Inclusion o f a control for prior military hostilities also has a side benefit bearing on the direct democracy hypothesis. A chief rival to the norm-based explanation asserts that certain decisional structures o f democratic states serve as impediments to war involvement. The structuralist logic does not entail the prediction o f a direct democracy effect on peaceful settlements, although it could explain such an effect by appealing to the fact that democracies seldom resort to military hostilities, and it is this absence of any prior hostilities that favors settlements. By controlling for prior hostilities I eliminate this structuralist interpretation.
Finally, I assume that peaceful settlements become more likely as conflicts become more costly to both rival parties. This is not only because the parties themselves become more amenable to settlement efforts as their own costs o f conflict rise, but also because high costs in terms o f fatalities or lost territory will elevate the regional or global salience o f the dispute producing a climate o f world opinion that could provide further inducements to settlement.
Rising costs thus stimulate m ore initiatives from primary or secondary parties to the conflict, as well as increase the likelihood o f acquiescence to any such initiatives. Costs are appraised individually for each party during each phase o f a conflict using a 3-point rating scale distinguishing minor losses (less than 1% o f total resources), moderate losses (l% -9% ), and heavy losses (10% or more). Admittedly this is a crude instrument, though it is undoubtedly better than omitting costs altogether. Because my reasoning specifically emphasizes the costs to both parties simultaneously, I further posit a mutually amplifying effect by which the positive impact o f costs to one party increases as losses to the other mount, and vice versa. This conditional effect can be modeled as a statistical interaction with a positively signed estimate for the cross-product term.
The empirical elements for evaluation o f the democracy-settlement hypotheses are now in place. This section has pieced together a tentative model from a series o f ad hoc arguments and hypotheses about conditions influencing the probability o f peaceful settlements. My purpose has not been to develop a theory o f the settlement process but only to establish sufficient controls to expose the effects o f democratic norms. These include a power disparity
measure, an indicator o f common alliance memberships, a phase counter and its square to accommodate the curvilinear effect o f duration, the prior history o f military hostilities, and measures o f short-term costs to each party.
Analysis and Results
I assume that conflict phases vary in their latent propensity for peaceful settlement and that this propensity can be modeled as a linear function o f democracy and conflict management along with the set o f control variables introduced in the previous section. Letting S* symbolize this unobserved propensity for settlement in phase i, it becomes possible to represent the two democracy hypotheses as estimable parameters in the following specification. The summation sign merely encompasses the additive effects o f the eight control variables and is not o f
S* = ß0 + £ ß / j , + ß9 min(DxA , ) + ßioH + ß „ [ ^ , x m in (A A , ) l + «, (1)
i 7 = 1
immediate interest. Democracy and conflict management are symbolized by D and M, respectively, with the notation min(DAl , DflI) explicitly denoting the minimum value for parties A and B. The estimated signs and values for parameters ß, and ß10 are o f special interest because they carry effects associated with the two democracy-settlement hypotheses. A direct democracy effect would be revealed by a positive and significant estimate for minimum democracy (ß9) whereas an indirect effect through conflict management requires a positive param eter on the management indicator (ß10). Mutual amplification o f these effects would be carried by the estimate for the democracy-management product term (ßn).
Let S stand for the observed realization o f the latent propensity with values signifying the presence or absence of peaceful settlements defined by some threshold on S*. As is the usual practice this threshold is set at 0, establishing the relationship depicted in equation (2).
This leads directly to a binomial probability model such that the probability o f settlement (i.e., Prob[S,=l]) can be expressed as some function o f the right hand side o f equation (1) with the exact functional
if S * > 0
i
if S * < 0
i
(2)
form determined by the expected distribution o f the error term, u. For the present analysis I assume ut to be distributed normally, which defines the usual probit model.
Democracy and Peaceful Settlement
I begin by evaluating the control model alone before introducing terms for institutionalized democracy and conflict management. Probit regression estimates o f the parameters are displayed in column 1.1 o f Table 1. It is evident that most o f the control variables perform as expected with alliance ties and the phase indicators carrying particularly robust effects. The phase counter terms are properly signed to reveal a positive but declining impact on settlements in the early phases of conflict until switching to a negative effect by about the sixth phase. Somewhat surprisingly the two cost indicators are negatively signed, though their product is positive and large enough in relation to its standard error to infer that cost effects are not independent o f one another. Thus the cost estimate for the stronger party applies only when costs to the weaker are 0. But as costs to the weaker party rise, this estimate is first diminished in size and then switches signs to become slightly positive, though probably not significantly so. Similarly, the estimate for the weaker party switches signs and becomes positive as costs to the stronger party increase.
[Table 1 about here]
Another interesting result o f this first estimation is the lack o f any effect for the power disparity measure. Power is a most elusive concept and it is certainly possible that my measurement procedure based on the ratio o f material capabilities is inadequate for the task at hand. Yet the estimate reported in column 1.1 is actually the most robust o f numerous attempts to detect power effects with several variations on both the summary ratio indicator and individual disputant capability scores. Even a series o f simple dummy variables defined at a variety o f points to indicate balance and disparity failed to produce noticeable effects.
Finally I should add that the estimate in 1.1 fails even to exceed its standard error when democracy and conflict management are introduced to the estimating equation. W hile it may be premature to conclude that power has no impact whatsoever on peaceful settlements, these results do conform to M iall’s (1992) findings on a different sample o f disputes. Accordingly, I omitted the pow er disparity measure from further consideration and reestimated the control variables. The results are presented in column 1.2 and serve as a baseline model for subsequent evaluation o f the two democracy-settlement hypotheses.
Column 1.3 o f Table 1 tests the direct democracy hypothesis by introducing the minimum democracy measure to the previous specification. W ith an estimate o f .066 and a t-ratio of 3.4, it is immediately evident that democracy does have a systematic positive influence on peaceful settlement that is consistent with the theoretical arguments reviewed earlier. The contribution o f democracy is also revealed in the improved chi squared and log-likelihood values indicating a better overall fit. At the same time the effects o f the control variables rem ain much the same as before. It is especially noteworthy that direct democracy effects appear even while controlling for alliance ties and prior military hostility, since these two, more than the others, could be confounded with democracy. The inclusion o f democracy also induces a slight but noticeable decline in the estimate for prior hostility, not surprising given the general absence o f hostilities at the highest levels o f democracy.
It is possible that this preliminary finding o f democracy effects is merely a spurious result o f the propensity for democratic states to manage—and ultimately settle—their mutual conflicts with the assistance o f third parties. This is the mechanism posed by the indirect democracy hypothesis. I examine this possibility with the introduction o f conflict management in column 1.4. As we would expect, the presence o f conflict management activity does have a positive and highly stable influence on peaceful settlements with a concomitant improvement in overall fit. The democracy estimate is reduced by management, but only slightly and still remains substantially larger than its standard error. From these results it appears that democracy does positively contribute to peaceful settlements both directly and indirectly.
Finally, I consider the possibility that minimum democracy and conflict management combine nonadditively to amplify the impact of one another. I probe for mutual amplification effects in column 1.5 with the introduction of the multiplicative product o f democracy and management. The t-ratio for this new term fails even to exceed unity and thus reveals no evidence o f the statistical interaction necessary to support the mutual amplification hypothesis.
Overall, the pattern o f effects in Table 1, and particularly in column 1.4, suggests that democracy imparts a dual im pact on the likelihood o f settlement. There is, first, a direct effect that I have argued is best understood in terms o f mutual democratic norms and expectations governing the resolution o f conflicts; and second, an indirect effect that follows from democracies’ willingness to accede to conflict management, a willingness also induced by mutual democratic norms.
Stability Tests
The evidence o f a direct democracy effect on peaceful settlement is critical to a norm-based approach and therefore deserves more critical scrutiny. I undertake this task with a series o f stability tests across a wide range o f alternative specifications all o f which introduce some additional term posing a challenge to my procedures or the conclusions drawn from them. I focus on four general areas: assumptions regarding my measurement o f democracy, the composition o f disputant parties, the potential confounding effect o f economic development, and two relational attributes--geographic contiguity and alliance ties. In all, this exercise covers fifteen different probit regressions, though I do not present each estimation in full since my interest is only in those results that bear on the role o f dem ocracy in peaceful settlement. However, all o f these estimations also include the seven control variables found to be effective in previous analyses. The results are presented in Table 2. All entries follow a similar pattern, listing an estimated parameter and immediately below it a t-ratio in parenthesis, with minimum democracy in the first column, the newly introduced term in the second under the heading Test Parameter, and, where appropriate, a cross-product interaction in the last column. Because these alternative hypotheses challenge the democracy result it is prudent to minimize type II errors; accordingly I employ a significance level set at a conservative .10.
Because an interstate dispute has by definition at least two participating states, each with its own level o f democracy, I arrived at a single representation o f democracy by what was earlier termed the weakest link principle. The first test parameter entered in panel A o f Table 2 represents democracy as a mean score for all disputant parties. The estimate o f .052 carries about the same effect as seen for minimum democracy in Table 1. This is not too surprising since the measures are highly correlated with one another (r=.72). The next entry includes the minimum measure along with this mean democracy indicator. Again the estimates are nearly identical, though due to the collinearity between them neither reaches even marginal significance. The bottom portion o f panel A follows the same procedure using the maximum democracy score among all the parties involved on either side. Though there is no compelling theoretical justification for this measure, I include it as a further check on the stability and robustness o f the earlier findings. Here the alternative measure reaches significance when entered alone but fails to do so when estimated alongside minimum democracy. From these results I conclude that neither alternative measure is demonstrably superior to the minimum democracy indicator used here.
[Table 2 about here]
The set o f estimates in panel B also examines the w eakest link principle, here focusing on its application to coalitions o f states. Two test parameters are considered: first, a dummy variable indicating whether or not the minimum democracy score is drawn from a coalition o f states, and second, a dummy variable indicating whether or not both parties to the dispute are comprised o f coalitions o f actors. Because I am testing the stability o f m inim um democracy rather than alternatives to it, I include it along with the test parameter and then enter a product term to detect any interactive effects. The product term estimate is o f particular interest because if found to be significant it would mean that the democracy effect differs under the conditions indicated by the constituent dummy variable. In neither case is this so. Furthermore, the minimum democracy estimate is noticeably diminished in only one estimation o f panel B and this can be ignored because it includes a nonsignificant interaction term. I conclude from these estimates that the minimum democracy effect on peaceful settlement is unaffected by the existence o f coalitions on either side of the dispute.
Next I examine one o f the chief rival hypotheses to democracy as a factor discouraging state involvement in serious conflict or war. The argument is based on the observation that most democratic states are also economically developed, suggesting that any democracy effect is actually spurious and that the true causal agent is economic development. I gauge the economic development of the states involved in disputes by using total energy production per capita drawn from the COW capabilities data base (Singer 1988). Panel C includes estimates for two different representations o f development and their interactions with democracy. The first, which is loosely based on the weakest link principle, uses the m inimum level o f development among all the states on either side o f the dispute. The second enters the level of development o f that state identified to have the m inimum democracy score in the dispute (or, in case o f ties on minimum democracy, the minimum o f economic development). The results show that democracy retains its effect in all four estimations while neither development coefficient nor the interaction terms reach even my conservative level o f significance. These results compliment those o f Maoz and Russett (1992a) for conflict involvement by demonstrating that the effect o f democracy on peaceful settlement is not spuriously confounded with economic development.
Finally, in panel D, I consider geographic contiguity and alliance bonds, two factors that traditionally have received considerable attention in studies o f interstate conflict (Maoz and Russett 1992a). Because most wars and militarized disputes are known to occur between contiguous states (Bremer 1992) we might expect contiguity to lessen the chances o f reaching a settlement. Alternatively, we might expect contiguity to insure shared cultural norms and a
