Annu. Rev. Polit. Sci. 1998. 1:27-46.

DOES DEMOCRACY CAUSE PEACE?

James Lee Ray
Department of Political Science, Vanderbilt University, Nashville, Tennessee 37235; e-mail: rayjl@ctrvax.vanderbilt.edu

ABSTRACT

The idea that democratic states have not fought and are not likely to fight interstate wars against each other runs counter to the realist and neorealist theoretical traditions that have dominated the field of international politics. Since the mid-1970s, the generation of new data and the development of superior analytical techniques have enabled evaluators of the idea to generate impressive empirical evidence in favor of the democratic peace proposition, which is reinforced by substantial theoretical elaboration. Some critics argue that common interests during the Cold War have been primarily responsible for peace among democracies, but both statistical evidence and intuitive arguments cast doubt on that contention. It has also been argued that transitions to democracy can make states war-prone, but that criticism too has been responded to persuasively. The diverse empirical evidence and developing theoretical bases that support the democratic peace proposition warrant confidence in its validity.

INTRODUCTION

The proposition that democratic states do not fight interstate wars against each other is one of the most influential ideas to appear in the academic subfield of international politics in recent years. Nevertheless, the basic idea is an old one. This review briefly traces the history of the idea that democracy is an important cause of peace and evaluates the evidence in its favor, necessarily including a discussion of its theoretical underpinnings. This essay focuses on the proposition that democratic states are peaceful in their relationships with each other, and not on the related but distinct notions that democratic states are less war-prone in general, or that the greater the number of democratic states in the international system, the lower the incidence of war in that system. Evidence provided by Morgan & Schwebach (1992), Bueno de Mesquita & Lalman (1992), Rummel (1995, 1997), Siverson (1995), Benoit (1996), Rousseau et al (1996), Huth (1996), Gleditsch & Hegre (1997), among others, casts some doubt on the validity of the widespread assumption or assertion that democratic states are just as conflict- or war-prone as undemocratic states, in general [though not in their relationships with each other (see Ray 1998)]. Theoretical discussions and interpretations of relationships between the proportion of democratic states in the system and the incidence of war in the system (Maoz & Abdolali 1989, Maoz 1996, McLaughlin 1996, Gleditsch & Hegre 1997, Senese 1997) have tended to overlook or obscure problems involved in making inferences across different levels of analysis (see Ray 1997b).

THE HISTORY OF THE IDEA

The most often cited classical source of the idea that democracy is an important force for peace is Immanuel Kant's 1795 essay, "Perpetual Peace." Kant was, however, no admirer of democracy. According to him, perpetual peace would occur only when states had civil constitutions establishing republics. According to Doyle (1983a, p. 226), for Kant a republic was a regime that respected private property and established a legal equality among citizens as subjects "on the basis of a representative government with a separation of powers."

More recent influences have come into play, of course, among them Woodrow Wilson. "President Wilson became the world's most influential statesmen in the aftermath of the First World War. His arguments dominated the new utopian discipline of International Relations" (Knutsen 1994, pp. 196–97).

Although in the 1920s this new American discipline was dominated by Wilsonian ideas and ideals, this was a brief preliminary phase that was beginning to end by the 1930s and was buried completely by the Second World War. "Realism" or "neorealism" came to dominate the field during that time—or such is the oft-repeated consensus.

But the consensus is not universal. One of the best-known and most influential realists in the field, Henry Kissinger, has an expansive view of the extent to which "Wilsonianism" dominates American thought on international politics. According to Kissinger's recent volume on diplomacy,

Woodrow Wilson was the embodiment of the tradition of American exceptionalism, and originated what would become the dominant intellectual school of American foreign policy.... The idea that peace depends above all on promoting democratic institutions has remained a staple of American thought to the present day. Conventional American wisdom has consistently maintained that democracies do not make war against each other. (Kissinger 1994, pp. 33, 44)

As important as Kissinger's opinion may be, it is more commonly asserted that realism and neorealism predominate among theoretical ideas in the North American academic scene. However, the research agenda fostered by the democratic peace proposition and the "neoliberal" theoretical notions supporting it, have posed what may turn out to be a significant challenge to that predominance (Kegley 1995).

CONTEMPORARY ORIGINS OF THE DEMOCRATIC PEACE PROPOSITION

Even during the Cold War, when realism clearly predominated, the idea that regime type has an important impact on foreign policies and international politics received some attention. One ultimately important research effort by Babst (1972), however, was virtually invisible to most specialists in international politics at the time. Utilizing data from Wright (1942), as well as a reasonably clear, operational definition of democracy, Babst (1972 p. 55) concluded that "no wars have been fought between independent nations with elective governments between 1789 and 1941."

Because it was published in Industrial Research, this article escaped the attention of almost all specialists in international politics. One who did cite it was RJ Rummel, in the fourth book of his five-volume Understanding Conflict and War (1975–1981). The first three volumes of this work develop the theoretical bases for 54 propositions, 33 of which focus specifically on the causes and conditions of conflict. The eleventh of the 33 propositions about conflict is that "Libertarian systems mutually preclude violence" (Rummel 1979, p. 279).

Although Rummel cites Babst in his discussion of the democratic peace proposition, probably few would be aware of Babst's work had it not been cited by Small & Singer (1976), who attempted to discredit Babst's assertion that democratic states are peaceful in their relationships with each other. Their attempt suffered from several shortcomings. The most serious was their inability to compare the rates of war proneness for democratic and autocratic states. Their data analyses were limited to the question of whether wars involving democratic states have historically been significantly different in length or in degree of violence than wars involving only autocratic states. "Even if democratic states had been, over the time period analyzed, 99 percent less likely to become involved in wars than autocratic states and 100 percent less likely to become involved in wars with each other, the wars in which they did become involved might still have been equivalent in length and severity to those involving only autocratic states" (Ray 1995, pp. 12–13).

Nevertheless, the paper by Small & Singer helped evoke the interest in the democratic peace proposition manifested some two decades later. It saved from obscurity Babst's claim about the peaceful nature of relationships among democratic states. More importantly, it distinguished the national-level relationship between regime type and international war proneness from the dyadic-level relationship between regime type and conflict.

This distinction was important, because the national-level idea that democracy has a pacifying effect had been around for a long time and was discounted by most analysts. Even 55 years ago, for example, Wright (1942) observed that "statistics can hardly be invoked to show that democracies have been less often involved in war than autocracies...probably there are tendencies toward both peace and war in democracies as there are in autocracies—tendencies which approximately neutralize each other and...render the probabilities of war for states under either form of government about equal." An attempt in the 1980s or 1990s to revive the idea that democracies are more peaceful, period, might well have confronted insurmountable skepticism. During that period, however, the focus on relationships among democratic states [a focus brought to mind, perhaps, by Small & Singer's (1976) distinction between the national- and dyadic-level relationship] was sufficiently distinct from the idea in its traditional form that it received a more responsive hearing.

Another important source of the current interest in the democratic peace proposition was a pair of articles by Doyle (1983a, b). He highlighted the Kantian basis for the democratic peace proposition and articulated its dyadic form. "Constitutionally secure liberal states have yet to engage in war with one another" (Doyle 1983a, p. 213). These two papers had an important long-range effect. Their essential contents were ultimately published in a more visible outlet (i.e. Doyle 1986), gaining the attention of the many potentially interested analysts not familiar with the work of Rummel, Babst, or Singer & Small. In addition Doyle, like Rummel, performed a systematic analysis of data regarding the validity of the democratic peace proposition. That is, both authors analyzed authoritative data on interstate wars (from the Correlates of War project at the University of Michigan; see Singer & Small 1972, Small & Singer 1982) and systematically attempted to classify by regime type all the states involved in those wars. (Doyle made a particularly energetic effort to categorize political regimes.) This allowed Doyle to reiterate with more authority the claim made by Babst (1972), acknowledged somewhat begrudgingly by Small & Singer (1976) and repeated by Rummel (1975–81), that no two democratic states have ever fought an interstate war against each other.

This absence of wars between democratic states may be the single most important and psychologically persuasive piece of evidence supporting the democratic peace proposition and probably accounts for its current prominence more than any other. The absence of wars between democracies seems to underlie Levy's (1988, p. 662) oft-quoted assertion that the democratic peace proposition is "as close as anything we have to an empirical law in international relations," or the equally sweeping statement by Gleditsch (1992, p. 372) that the "perfect" correlation between democracy and nonwar in dyads implies that "most behavioral research on conditions for war and peace in the modern world can now be thrown on the scrap-heap of history, and researchers can start all over again on a new basis."

Statements like these, as well as the current prominence of the democratic peace proposition, warrant at least a brief discussion of the validity of the assertion that, as a rule, democratic states do not fight interstate wars against each other. A lengthy list of possible exceptions to the rule has been discussed (Ray 1993, 1995), including the War of 1812 between the United States and Great Britain, the American Civil War, the Boer War, the Spanish-American War, the First World War, and the "wars" between Finland and democratic allies of the Soviet Union that occurred within the Second World War. A resolution of the debate over these putative exceptions requires the solution of at least three basic conceptual problems.

The first problem is the definition of the concept "interstate war." With the exception of Babst's (1972), most efforts to address this issue relied on the definition devised some time ago by the Correlates of War project, which stipulates that only military conflict between independent states leading to at least 1000 battle deaths (that is, the deaths of soldiers) will count as an interstate war. This definition is not free of controversy. It excludes, for example, the American Civil War (on the grounds that the South was not an independent state), as well as the various wars between US troops and American Indian tribes.

But most of the controversy has focused on the second conceptual problem, the definition of "democracy." There is a widespread view that democracy is an essentially contested and time-dependent concept (Oren 1995). If this view is accepted, the validity of the democratic peace proposition is impossible to evaluate in any systematic or convincing fashion, because political regimes cannot be categorized in a consistent manner acceptable to a sufficiently large proportion of analysts to serve as a basis for consensus.

A democratic government is typically thought of as one based on the consent of, and responding to the wishes of, its constituents. Surely a fundamental reason for the difficulty of devising a universal definition of democracy is that governments of widely varying, even diametrically opposed, structure and attributes can be perceived as responsive to the needs and desires of their constituents. "Bourgeois" republics, fascist dictatorships, and dictatorships of the proletariat can be and have been perceived by numerous and passionate advocates to meet this essential criterion for "democracy."

Perhaps, though, "the debate about the democratic peace proposition need not resolve the philosophical debate about which regimes are `truly' democratic" (Ray 1997a, p. 52), or which regimes most faithfully reflect and respond to (or best serve, for that matter) the interests and preferences of their constituents. Advocates of the democratic peace proposition have an admittedly procedural concept of democracy, focusing on competitive elections, widespread suffrage, civil rights, freedom of the press, etc. Many of these attributes and structures are relatively easy to identify. Whether they lead to governments that truly serve the interests of their constituents is an interesting question, but its resolution is not crucial to the debate about the democratic peace proposition. That debate can focus instead on whether political regimes possessing such relatively easy-to-identify characteristics as competitive elections, widespread suffrage, and civil rights, for example, behave differently—especially toward each other—than do other sorts of regimes.

Admittedly, even identifying political regimes possessing these structures and attributes is no trivial task. Most of the current efforts to categorize regimes for the purpose of evaluating the democratic peace proposition have been based on data generated by Gurr and his colleagues (see Gurr 1974, 1978;, Gurr et al, 1989, 1990;, Jaggers & Gurr 1995). The most current form of the dataset is referred to as Polity III. Although valuable, these data are not without their limitations as a basis for resolving the debate about whether, or how often, there have been wars between democratic states (Gleditsch & Ward 1997). Gurr et al apply an 11-point ordinal scale of democracy to almost every state in the world for every year from 1800 to the 1990s. Those democracy scores are themselves sums of scores on various dimensions reflecting, for example, the selection of government executives by election, the openness of executive recruitment, and the parity between the executive and legislative branches of government. These separate dimensions are themselves complex, and when their scores are added, the resulting overall democracy score consists of compounded layers of complexity. Arguments that a particular state at a given time cannot be categorized as democratic because it does not rate a score of 7 on the Polity III democracy index are not likely to be persuasive to skeptics, or even disinterested observers, because the threshold of 7, or of any other score on the Polity III index, is unclear.

Nevertheless, the selection of some kind of threshold is logically necessary in order to address the question of whether there has ever been a war between democratic states. In one sense this creates an insoluble problem (the third of three conceptual problems alluded to above). Democracy is a continuous concept; states are democratic to lesser or greater degrees, and therefore it is impossible to sort states into two categories, democratic and not democratic. This makes it necessary for "those who attempt to defend [or evaluate] the assertion that democracies never fight wars against one another to acknowledge that in reality the assertion that they are defending, in more precise terms is `States that have achieved a certain level of democracy...have never fought wars against each other", (Ray 1995, p. 90).

Ideally, that "certain level" would be a threshold that is simple, intuitively appealing, operational, and theoretically defensible. One could, for example, stipulate that states are sufficiently democratic to avoid wars against each other if their executive and legislative leaders are selected in a process based on fair, competitive elections. For the purpose of identifying this threshold, competitive elections could be defined as those in which at least two formally independent political parties (or other groups) present candidates. Elections might be designated "fair" if at least 50% of the adult population is allowed to vote, and if the political system in question has produced at least one peaceful, constitutional transfer of executive power from one independent political party to another by means of an election. This last criterion, focusing on constitutional transfers of power, is both easily discernible in most empirically relevant cases and crucial to the identification of states sufficiently democratic as to behave in ways recognized by important theoretical approaches to "democracy." If this threshold is applied to states involved in various controversial cases, such as the War of 1812 between the United States and Great Britain, the American Civil War, the Boer War, or the Spanish-American War, it is possible to argue that no dispute between states has ever escalated to interstate war unless at least one of the states involved was not (sufficiently) democratic (Ray 1993, 1995).

ANALYZING DATA REGARDING THE DEMOCRATIC PEACE PROPOSITION

Even were it universally agreed that there has never been an interstate war between democratic states, that fact might be devoid of theoretical or practical significance. Something on the order of 99% of all the pairs of states in the world have peaceful relationships in an average year. Until recently, the proportion of democratic states in the world has been small. So it is possible in principle that the only reason there has never been a war between democratic states is that such an event is statistically unlikely.

Substantial progress has been made in examining this possibility since Small & Singer (1976) published their groundbreaking work. First, because they did not have data on regime type over the long period on which they focused (1816–1965), Small & Singer were able to generate data only on the regimes of states actually involved in wars. Thus they were unable to compare systematically the rates of warfare of democratic pairs of states with those of others.

Small & Singer did report an absence of wars between democratic states with "marginal exceptions," but they discounted the significance of this pattern. What they did not do, again at least partly because they had no comprehensive data on regime types, was to evaluate the statistical likelihood of this absence of war between democratic states.

Rummel (1983) made a significant step toward dealing with this problem by using data on regime type generated by Gastil (1981), as well as Correlates of War data on interstate war (Small & Singer 1982). He analyzed virtually all pairs of states in the world in the years from 1976 to 1980, summing up the number of pairs each year for a total of 62,040 observations. From these data he concluded that the absence of war between democratic states, compared with the rate of warfare among other states, is statistically significant. But this finding had limited impact, because it only applied to the period from 1976 to 1980. He acknowledged that "unfortunately, I do not have data to determine the number of democracies in the world for the 1816–1965 period" (Rummel 1983, pp. 47–48).

Maoz & Abdolali (1989) made a significant contribution toward solving this problem, using the data generated by Gurr (1974, 1978). They analyzed virtually every pair of states in the world in the years from 1816 to 1976, for a total of 271,904 observations. They reported not only that there have been no interstate wars between democratic states over this lengthy period, but also that this number of wars (i.e. zero) is significantly less than would be statistically expected, despite the rarity of both wars and democratic states.

Maoz & Abdolali's work also reflects another important advance, that is, the development of data on militarized disputes that fall short of escalating all the way to interstate war. These are disputes in which at least one state overtly threatens, displays, or actually uses military force against another (Gochman & Maoz 1984). Analyzing such data, Maoz & Abdolali (1989, p. 21) reported that in the years from 1816 to 1976, "democratic states never fight one another. They are also less likely to engage in lower-level conflicts with each other...."

Subsequently, Gurr and his associates developed annual regime-type scores for most states in the international system from 1800 onward [Maoz & Abdolali (1989) had relied heavily on extrapolation], which contributed significantly toward subsequent comparisons of the rates of warfare between democratic states with the rates of warfare between states not jointly democratic. Bremer (1992) analyzed data on virtually every pair of states in the international system from 1816 to 1965, for a total of 202,778 observations. His central conclusion was that from 1816 to 1965 the rates of warfare between democratic states versus those between other pairs are significantly different.

Statistically speaking, Bremer's claim may be valid, but his observed proportion of democratic pairs of states involved in interstate wars between 1816 and 1965 was essentially zero, and his proportion of other pairs of states at war during this same period was 0.0005. Can that difference really be substantively significant?

One reason Bremer observed such a microscopic difference between the rate of warfare between jointly democratic states and that of other pairs of states is that his analysis included many irrelevant pairs. There were so few democratic states in the pre–World War II era (1816–1939) that one might question the utility of analyzing those years at all. In addition, a huge number of the pairs of states Bremer included are so geographically isolated from each other that there is no realistic chance of war between them. Such pairs as Burma and Bolivia or Chad and Chile will never, it is safe to predict, fight wars against each other.

Maoz & Russett (1992) focused only on the period from 1946 to 1986, thus eliminating from consideration all those less-relevant years during which democracies were relatively rare. Still, because the number of states in the system increased so dramatically between 1965 (the terminal year in Bremer's analysis) and 1986, Maoz & Russett, who included almost every pair of states every year, ended up with 264,819 observations—slightly more than Bremer (1992) analyzed. They reduced that number substantially by analyzing only "politically relevant" pairs of states, which they defined as pairs that are geographically contiguous or that include at least one major power. (The definition of a "major power" is an issue excluded to conserve space. During the Cold War a widely accepted list of major powers included China, France, Great Britain, the Soviet Union, and the United States.) This restriction reduced their number of observations to 29,081. Their data showed no democratic states at war with each other. The proportion of observations of nondemocratic pairs involved in an interstate war in the years from 1946 to 1986 is reported to be 0.001. Again, contrasting the absence of wars between democratic states to the occurrence of wars 0.1% of the time between states not jointly democratic might not seem substantively significant (even though it is statistically significant.) But, as Bueno de Mesquita (1984, p. 354) pointed out in a different context, the proportion of cigarette smokers who develop cancer is only 0.001482 higher than the proportion of nonsmokers who develop it. Perhaps it is not unreasonable to argue that such small absolute differences are not intrinsically trivial.

Small & Singer (1976, p. 67) discounted their finding of the absence of wars between democratic states on the grounds that most wars occur between geographically contiguous states, and "bourgeois democracies do not border upon one another very frequently over much of the period since 1816." They must have based that conclusion on an unsystematic review of the available information, because at that time there were no available data categorizing pairs of states in the international system as geographically contiguous or not. This is another front on which research has made substantial progress since 1976. The Correlates of War project has generated data on the geographical relationships of virtually all pairs of states in the international system since 1816 (Gochman 1991). These data allowed Bremer (1992), as well as Maoz & Russett (1992) and by now many others, to control for the impact of contiguity on rates of warfare between pairs of states while evaluating the impact of regime type. This analysis established definitively that the speculation by Small & Singer (1976) regarding the impact of contiguity on the relationship between regime type and conflict proneness was erroneous. Using data on distance between states, Gleditsch (1995) demonstrated persuasively that democratic states have not historically been geographically separated substantially more than non–jointly democratic pairs of states.

Since Small & Singer (1976) were unable to generate data on the relative rates of warfare for democratic pairs versus other pairs of states, they could not evaluate the possibility that the difference between those rates (even if statistically significant) might not result from regime type. In other words, correlation does not prove causation, and even if there is a correlation between regime type and conflict or war proneness, the pattern might be produced by some third factor that has an impact on both war proneness and regime type. The most recent research on the democratic peace proposition has considered this possibility thoroughly.

Bremer (1992, 1993), for example, examined the relationship between regime type and international conflict while controlling for contiguity, power status, alliance ties, militarization, economic development, and power ratios. Maoz & Russett (1992, 1993), Russett (1993) focused, unlike Bremer, only on the period from 1946 to 1986, and primarily on "politically relevant" pairs of states. They controlled for contiguity, alliance ties, economic wealth and growth, political stability, and power ratios. These analyses represent significant progress over the effort made by Small & Singer (1976) not only because they control for additional factors, but also because they take advantage of such techniques as logistic regression, poisson regression, and negative binomial analyses, which are better suited to the type of data being analyzed than the traditional ordinary least squares method relied on in earlier decades (King 1989). In general, the findings of these analyses are supportive of the democratic peace proposition. Bremer (1993, p. 246) concludes, for example, that "even after controlling for a large number of factors...democracy's conflict-reducing effect remains strong."

ATTACKS ON THE EVIDENCE IN SUPPORT OF DEMOCRATIC PEACE

Impressive though it may be, research on the democratic peace proposition has generated vigorous criticism attempting to discredit the evidence marshaled on its behalf. One of the counterarguments focuses on the lack of independence of the observations of, for example, all pairs of states in the international system on an annual basis. The observations of Great Britain and the United States in successive years, for example, are not independent from each other. It is unfair or misleading, according to this argument, to count each annual observation of peace between Great Britain and the United States as a statistical victory of sorts for the democratic peace proposition (Spiro 1994).

One of the responses to this criticism has involved moving away from the dyad-year as the unit of observation for statistical calculation, and instead making a single observation for each pair of states over a given period, say from 1946 to 1986. Pairs are categorized by their regime types, i.e. both democratic, only one democratic, or neither democratic. Then the probability can be calculated that all pairs at war with each other during that period would be either mixed dyads (one democratic, one autocratic) or uniformly autocratic. Both Russett (1995), Rummel (1997) performed analyses along these lines and concluded that the absence of any democratic pairs among those states becoming involved in wars against each other from 1946 to 1986 is unlikely to have occurred by chance. Maoz (1997) has completed analogous analyses for the years 1816 to 1986 and came to the same conclusion.

Admittedly, such applications of significance tests do not conform to their classical purpose of generalizing from a sample to a population, but there are convincing rationales for relying on significance tests even if one has access to the population of cases being analyzed; a good recent rationale can be found in Rummel (1997, p. 46).

Another approach relies on statistical techniques designed to deal with at least some of the problems created by interdependence of observations, as well as by binary dependent variables, building on the work of Huber (1967). Beck & Tucker (1996) report that application of such techniques to dyad-years from 1946 to 1986 provides evidence in support of the democratic peace proposition.

Farber & Gowa (1995, 1997) offered additional important criticism of the idea that democracy is an important pacifying force, arguing that the bulk of the evidence in its favor comes from the Cold War era, during which democracies probably avoided serious conflicts with each other because of common interests generated by their confrontation with communism. The majority of democratic states that have ever existed have emerged during the Cold War era, and it is possible that that historical epoch may prove idiosyncratic with respect to relationships among democratic states. Only time will tell whether the large number of democratic states that have emerged in recent years will fight wars against each other in the absence of a serious threat from the Soviet Union or, perhaps, any other undemocratic state.

But Thompson & Tucker (1997) have cast some doubt on the arguments and evidence provided by Farber & Gowa (1995), and especially on the theoretical rationale they developed to account for their findings. Maoz (1997, p. 174–175) among others has pointed out that alliance ties would seem a likely indicator of the common interests among democratic states that Farber & Gowa claim really account for peaceful relationships among them. In their own analyses, Farber & Gowa discuss methodological problems in introducing alliance ties as a control variable. These problems indubitably exist, but Bremer (1992), Maoz & Russett (1992, 1993), Russett (1993), Oneal et al (1996), Barbieri (1996a), Oneal & Russett (1997), Oneal & Ray (1997), among others, all utilize alliance ties as a control variable in analyses of the relationship between regime type and international conflict. All report that joint democracy is significantly related to decreases in conflict for pairs of states, with alliance ties (reasonable proxies for common interests, as even Farber & Gowa acknowledge) controlled for.

On a more intuitive level, one might reasonably infer that if the opposition of the communists was sufficient to create common interests guaranteeing peace among the democratic ( or anticommunist) states of the world, then the opposition of the "Free World" (even more formidable, by most measures) should have been sufficient to guarantee peace among the communist states. Yet during the Cold War the Soviet Union invaded Hungary and Czechoslovakia, as well as Afghanistan, and experienced serious border clashes with communist China. Meanwhile, Vietnam attacked and occupied most of Cambodia, provoking a retaliatory attack by communist China.

The "opposition leads to common interests leads to peace" idea would also be hard-pressed to account for the fact that peace did not prevail uniformly on the anticommunist side of the Cold War divide. For example, El Salvador fought a war with Honduras in 1969, Turkey and Greece became embroiled in a war over the fate of Cyprus in 1974, and Great Britain clashed with Argentina over the Falkland/Malvinas islands in 1982. These cases are not anomalies for advocates of the democratic peace proposition; each of those wars involved at least one undemocratic state.

Probably the most visible attack on the credibility of the democratic peace proposition was that by Mansfield & Snyder (1995a, b), ably defended from initial criticism in Mansfield & Snyder (1996). Interestingly, however, theirs was not a direct attack on the proposition. They did not claim that joint democracy has no significant pacifying affect on relationships between states. Rather, they argued that democratization, that is, regime transitions from autocracy to democracy, make states significantly more war-prone.

This claim evoked a debate in print (initiated with direct replies from Wolf 1996, Weede 1996a, Enterline 1996) that has become exceedingly fine-grained. Evidence generated by Bremer (1996), Enterline (1996) and unpublished data), Oneal & Russett (1997), Thompson & Tucker (1997), Oneal & Ray (1997), among others, indicates that democratization does not increase the probability of war, except perhaps as a result of a dyadic-level pattern in which the democratization of states surrounded by autocratic states increases the "political distance," or difference in regime type, between themselves and their neighbors. In other words, while democratization in and of itself may not increase war proneness, it can have the effect of increasing the likelihood of war between a democratizing state and any of its neighbors if the neighbors are autocratic.

THEORETICAL BASES FOR THE DEMOCRATIC PEACE PROPOSITION

Additional criticisms of the evidence regarding the democratic peace proposition—such as those contending that it is misinterpreted as a result of cultural bias, or limited to a relatively narrow Western cultural region (Oren 1995, Cohen 1994), or that it is actually peace that produces democracy rather than democracy that leads to peace (Thompson 1996)—are acknowledged but not examined here in order to conserve space for consideration of the more fundamental and common argument that the absence of war between democratic states (even if one acknowledges it) is merely an empirical oddity for which no convincing theoretical explanation or rationale exists.

One reason for the common opinion that peace between democratic states is an empirical pattern devoid of sound theoretical bases is the lack of attention paid to Rummel's Understanding Conflict and War (1975–1981), which provided three volumes of epistemological, even metaphysical, not to mention theoretical elaboration before presenting the democratic peace proposition in the fourth volume (Ray 1998). Another reason may be that some of the more elaborate, systematic, empirical evidence in support of the proposition (e.g. Maoz & Abdolali 1989;, Bremer 1992;, Maoz & Russett 1992, 1993) was presented without much elaborate theoretical defense of the idea.

But there are in fact several well-developed theoretical notions regarding relationships between democratic states, at least some of which are potentially complementary rather than competitive. Rummel (1997) has recast and updated his theoretical defense of the idea that democracy is an important pacifying force, emphasizing the impact of public opinion, cross-pressures in democratic societies, and a field theory focusing on the differences between political processes in democratic and autocratic societies. In the "first book since...volume 4 of Understanding Conflict and War (1979) to explicitly test whether democracies don't make war on each other" (Rummel 1997, p. 36), Bueno de Mesquita & Lalman (1992) develop a game-theoretical analysis of international interactions that emphasizes the impact of various domestic political incentives on bargaining among democratic states. Russett's (1993) structural and cultural theoretical frameworks in defense of the democratic peace proposition take the form of interrelated propositions about the impact of regime type on international conflict; the structural framework deals with institutional constraints that lead to peaceful resolutions of conflicts, and the cultural framework emphasizes the effect of cultural and normative restraints on bargaining between democratic states.

Relying mainly on the work of Maoz & Russett (1993), as well as Weart (1994 and unpublished data), one recent review of research on the democratic peace proposition concludes that "normative explanations of the democratic peace have been shown to be more persuasive than structural explanations," and that "normative explanations have fared better in research" (Chan 1997, p. 77–78). The normative approach, with an impressive forebear in Deutsch et al (1957), has spawned supportive research on regime type and related factors that foster peace between democratic states, and on propositions related to but distinct from the democratic peace proposition. Oneal et al (1996), Oneal & Russett (1997), Oneal & Ray (1997), for example, provide evidence that international trade among democratic states may have a pacifying impact complementary to that of regime type;1 Russett (1997) finds that international organizations may also contribute to peaceful relationships among democratic states (and among states in general). In addition, Dixon (1993, 1994), Raymond (1994) report that democratic states, once involved in disputes, are more likely than undemocratic states to rely on such conflict-resolution techniques as mediation and arbitration, and that such techniques are more likely to be successful if the states employing them are democratic. [Raymond (1996), however, also reports that although democratic pairs of states are more likely to choose a binding method of arbitration for disputes between them, they are not more likely, once having become involved in such a process, to successfully resolve the dispute in question through that method.]

But there is another important theoretical thrust regarding the impact of regime type that relies less on the idea that democracy evokes normative commitments to the peaceful resolution of conflicts, and more on the idea that "leaders in democracies might avoid wars against other democratic states...because they feel that fighting such wars might be harmful to their chances of staying in power" (Ray 1995, p. 40). Bueno de Mesquita et al (1992), Bueno de Mesquita & Siverson (1995) find that interstate wars do have important impacts on the fate of political regimes, and that the probability that a political leader will fall from power in the wake of a lost war is particularly high in democratic states. Interestingly, Lake (1992) reports that democratic states are more likely than autocratic states to win the wars in which they participate. He argues that this is because democracies, for internal political and economic reasons, have greater resources to devote to national security. This might mean that democratic leaders are unlikely to select other democratic states as targets because they perceive them to be particularly formidable opponents. Siverson (1995) argues, however, that democracies tend to select their opponents in wars more carefully than do autocratic states, being more likely to target opponents they are relatively certain they can defeat. (Why this would make them unlikely to pick democratic opponents is addressed below.)

Recently a formal model was developed (B Bueno de Mesquita, R Siverson, unpublished data) based on the axiom that political leaders in democracies, autocracies, military juntas, monarchies, and other forms of government share a common feature: "They desire to remain in office" (emphasis in the original). Bueno de Mesquita & Siverson claim that this model can account not only for the absence of wars among democratic states, but also for five additional, related, important empirical regularities: (a) the tendency for democratic states to fight other kinds of states with considerable regularity, (b) the tendency for democratic states to win the wars in which they participate, (c) the tendency for democratic states to experience fewer battle deaths in the wars they initiate, (d) the tendency for disputes between democracies to reach peaceful settlements, and (e) the tendency for major-power democracies to be more constrained to avoid war than less powerful democracies.

This list is potentially instructive because debates about competing explanations of the democratic peace are more likely to be resolved by successful subsumption of the absence of war between democratic states within related patterns and propositions than direct, increasingly detailed empirical examination of the relationship between joint democracy and the incidence of war, or analysis of individual cases (see Elman 1997). The list generated by Bueno de Mesquita & Siverson is also interesting because it omits one important pattern in the relationship between regime type and international conflict, the tendency for democratic pairs of states to be less likely than other pairs to become involved in serious, militarized disputes, convincingly demonstrated by Bremer (1993), Maoz & Russett (1993) among others.

It is possible that the leaders of democratic states, once they become involved in relatively serious disputes, avoid wars against other democratic states for the strategic, logical, or self-interested reasons, e.g. the impact of lost wars on their personal political fates, stressed by Bueno de Mesquita & Siverson (unpublished data). It is also possible that democratic states get into serious disputes with each other at a significantly lower rate than other kinds of states because of the normative or cultural constraints emphasized by a related, but nevertheless partially competitive, strand of research on the democratic peace proposition.

There is, however, at least one potential problem with the logical arguments provided by Bueno de Mesquita & Siverson. In their attempt to account for the propensity of democratic dyads to negotiate disputes peacefully, Bueno de Mesquita & Siverson (unpublished data) argue that when two democracies confront each other in a dispute, generally, one has a good chance of winning and the other does not.

Is it necessarily the case, in confrontations between democratic states or any states, that one has a good chance of winning and the other does not? Is it not the case instead that sometimes the probable winner of the dispute is relatively clear, and in others it is not? This partially rhetorical question brings to mind the argument by Vasquez (1993, p. 64–65) that interstate wars, for the purpose of analysis, need to be categorized. In short, he argues that wars between states of approximately equal military-industrial capabilities ought to be considered separately from wars between states that are clearly unequal in such capabilities. Perhaps such a categorization would also be helpful in the analysis of serious, militarized disputes. If so, then democratic leaders in confrontations with each other may be most likely to fear the kind of punishment emphasized by Bueno de Mesquita & Siverson (unpublished data) if their states' capabilities are nearly equal.

An additional factor, most likely to be relevant when democratic disputants are clearly unequal in tangible military capabilities, is the greater ability of democratic states to make credible commitments (Fearon 1994), because they are visibly faced with greater "audience costs" if they back down (see also Eyerman & Hart 1996). In other words, the processes leading to wars between unequal states are distinct from those leading to wars between equal states. For example, fear of failure (a lost war) may be more likely to have a constraining effect on the leaders of states roughly equivalent in capabilities (Bueno de Mesquita & Siverson, unpublished data). However, if the capabilities of the disputing democratic states are highly disparate, intangible elements may play a particularly important role in the decisionmaking processes on both sides of the dispute (Bueno de Mesquita et al 1997). The superior ability of democratic states to make credible commitments or to demonstrate resolve may help unequal, democratic states avoid the confusion or uncertainty that makes disputes involving unequal, undemocratic states more likely to escalate to war.

CONCLUSION

Does democracy cause peace? The empirical evidence in favor of the proposition that democratic states have not initiated and are not likely to initiate interstate wars against each other is substantial, especially when compared with that which could be brought to bear by specialists in the 1970s. Criticism of this evidence has so far met with reasonably persuasive counterarguments by the defenders of the proposition. Despite a common opinion to the contrary, the theoretical bases for the hypothesis regarding the absence of war between democratic states are highly developed and may to some extent be complementary as well as competitive. For example, some factors may make democratic states unlikely to become involved in serious, militarized disputes in the first place, while other factors enable them to resolve serious disputes without warfare when they do occur. No scientific evidence is entirely definitive, and the greater number of democratic states in the post–Cold War era may increase opportunities for conflicts that will cast grave doubts on the democratic peace proposition. But for the moment at least, well-developed theoretical bases reinforce a lengthy list of systematic empirical analyses in support of that proposition. Moreover, the multiple streams of arguments and evidence supporting the proposition are highly diverse in character: epistemological (Rummel 1975), philosophical (Doyle 1986), formal (Bueno de Mesquita & Lalman 1992; B Bueno de Mesquita, R Siverson, unpublished data), historical (Weart 1994, Ray 1995, Owen 1994), experimental (Mintz & Geva 1993), anthropological (Ember et al 1992, Crawford 1994), psychological (Kegley & Hermann 1995), economic (Brawley 1993, Weede 1996b), political (Gaubatz 1991), and statistical (Ray & Russett 1996, p. 458). Perhaps, then, the more defensible of the two possible definitive answers to the question "Does democracy cause peace?" is "Yes."

FOOTNOTES

1 Barbieri (1996a, b) provides contrasting evidence suggesting that trade can exacerbate conflict. Weede (1996a), on the other hand, incorporates a hypothesized pacifying impact of international trade into his sweeping comprehensive and integrative theoretical analysis of both international conflict and international political economy, arguing essentially that trade leads to wealth which in turn leads to democracy as well as peace.

Annu. Rev. Polit. Sci. 1998. 1:27-46
Copyright 1998 by Annual Reviews Inc. All rights reserved
0163-8998/98/0616-0027