Sebastian Rosato (2003) has given us another spirited critique of the democratic peace project. His argument is similar to other realists' claims that the correlation between democratic–state interaction and peace is spurious, better understood as a function of power, threat, and national interests. His approach differs from others in that he attempts to scrutinize the many causal propositions contained in democratic peace theory, concluding in the end that all of them are contradicted by empirical evidence, and are consistently contradicted. But it fails on at least two counts. First, most of what Rosato cites as evidence against democratic peace theory does not in fact contradict the theory. Second, the evidence that does contradict the theory, in addition to being widely known among democratic peace researchers, is not particularly damaging to the theory, which continues to evolve at the core of a progressive research program.

The democratic peace is a dyadic empirical phenomenon. The empirical evidence that democracies rarely fight each other is robust, and most theoretical efforts have kept this finding front and center. Yet Rosato (2003, 589, 596), at various points in his critique, suggests that the dyadic claim is a retreat from some original monadic position in the face of arguments and examples to the contrary. Thus, dyadic propositions are cast as “restatements” or “new arguments” designed to “rescue” the theory's causal logic. This mischaracterizes the evolution of the democratic peace research program. Although some studies have offered evidence that democratic states generally conduct their foreign affairs more peacefully than nondemocratic states (Benoit 1996; Ray 1995; Rousseau et al. 1996; Rummel 1995), the early theoretical and empirical work on the democratic peace, and most of what has followed, recognizes that a core element of democratic peace theory must be located in the nature of democratic states' interaction. Doyle (1983a, 1983b), one of the founders of the democratic peace project, is very clear on this score: “liberalism is not inherently ‘peace-loving’; nor is it consistently restrained or peaceful in intent.” It has, however, “strengthened the prospects for a world peace established by the steady expansion of a separate peace among liberal societies” (Doyle 1983a, 206; see also Russett and Starr 1981, 439–44).

Rosato (2003) is well aware of the dyadic argument, but he does not seem to take it seriously. In dissecting the normative explanation, he identifies two links in the causal chain connecting domestic conduct in democracies to peaceful conduct in foreign affairs: elites externalize their norms of negotiation and nonviolent conflict resolution, which in turn encourages them to trust and respect their counterparts in other democracies. If this is the case, Rosato believes, then democracies should have a record of fighting wars only in self-defense or to prevent egregious violations of human rights. Clearly democracies have not limited themselves to such conflicts and Rosato produces a list of wars fought for other, imperial reasons; this is supposed to refute the claim that democracies “generally externalize their internal norms of conflict resolution” (589, 590, my emphasis). The list does refute the claim, of course, but it is not a claim made by the corpus of democratic peace theory.

According to most variants of the theory, democratic restraint is conditioned on expectations about the conduct of the other party in the interaction, expectations informed by the other's internal political processes.¹

We need to know something about those processes (or perceptions of those processes) if the cases are to be counted as anomalies. Rosato (2003) acknowledges the rebuttal, but again does not take it seriously, insisting that “[t]he key to this logic is that democracies must reliably externalize democratic norms” (590, my emphasis). Ultimately, however, his assertion is much stronger than this: “[l]iberal states have consistently violated liberal norms when deciding to go to war” (590, my emphasis). If this is not true by definition—isn't the decision to go to war, in the end, always a violation of liberal norms of conflict resolution?—then it is hard to imagine the type of evidence that would count against it. And even if democratic states did reliably externalize their norms, Rosato maintains that “[s]hared democratic values provide no guarantee that states will both trust and respect each other” (592). If it has come to making guarantees, then democratic peace theory surely must throw in the towel.

The dyadic logic of democratic peace theory is also set aside when Rosato (2003) turns to explanations focusing on the institutional constraints operating in democracies. He finds unconvincing the classical liberal argument that mass publics, because they bear the costs of war, have an interest in peace, and that mass publics in democracies, because their voices are heard, are a force for peace. Nor does he buy the variation on this argument, which states that certain groups within society, if not the masses, are advocates of peace, and their views are more likely to have an impact on the foreign policies of democracies than those of nondemocracies. That democratic publics and interest groups are not always pacific has long been established in public opinion research (Mueller 1973), and democratic leaders often look forward to a rally-'round-the-flag effect even when the balance of prewar opinion tilts against the use of force.

Rosato (2003) cites several examples of supportive (or quiescent) democratic publics during wars fought for reasons other than self-defense—but all of them involved nondemocratic opponents. Noting the character of opponents is the sort of “restatement” he dismisses as an attempt to save the theory from contradictory evidence—a charge that sticks only if one paints dyadic democratic peace theory as a retreat from the monadic argument, which it is not. Moreover, the examples adduced to falsify the claim that “democratic citizens are only averse to costs in their relations with other democracies” include colonial conflicts between Britain and France during the first half of the 19th century, when France was not democratic, and between Ecuador and Peru during the 1990s, when Peru was not democratic (596, note 16). During the 1830–32, 1838–41, and 1844 confrontations with Britain, the Polity Project locates France at −1 on their democracy–autocracy scale ranging from +10 to −10; whereas in the 1990s, Peru is scored as +1 (and −3 in 1992). Even if Rosato has some reason to believe that the regimes ought to be considered democratic, he gives us no indication of prowar public sentiments in these or any of the other democratic societies involved in the crises.²

After all, the stated purpose of his analysis is not to challenge the “powerful empirical generalization” that democracies rarely fight each other, which “remain[s] robust” (585), but to dispute the causal mechanisms that purportedly steer democracies away from war with each other.

Few would deny that hawkish interest groups often prevail in domestic debates or that “pacific interest groups may not generally influence the foreign policies of democratic states” (596). In the case of the recent Iraq War, there was indeed surprisingly little debate in the United States—until after the war. Rosato (2003) goes further, hypothesizing that, when contemplating going to war, autocratic leaders are more constrained by domestic constituents than are democratic leaders. He believes this may be true because wartime taxation without representation threatens to mobilize domestic opposition to nonrepresentative political institutions, sweeping away the autocracy in the process. This is an interesting argument, perhaps, as long as it applies to the avoidance of very costly wars. Autocrats do not typically shy away from taxation in pursuit of personal enrichment—presidential palaces and Swiss bank accounts—for fear of domestic disapproval, so they are unlikely to avoid foreign conflicts that they expect will not be terribly costly. In the end, the persuasiveness of Rosato's own causal logic will turn on the evidence. Curiously, although Rosato cites them to support his statement that autocracies “often represent groups that have a vested interest in avoiding foreign wars” (597), Peceny, Beer, and Sanchez-Terry (2002, 25) find “no unambiguous evidence of a dictatorial peace”; “only joint democracy was consistently related to a lower frequency of militarized disputes.”³

The possibility that autocrats exercise more restraint in international crises is also raised in the discussion of political accountability. The argument found in democratic peace theory is that democratic leaders risk removal from office after unsuccessful and/or costly wars, a risk that is much diminished for autocratic leaders (Bueno de Mesquita et al. 1999, 2003, chap. 6; Gelpi and Griesdorf 2001; Reiter and Stam 2002). Rosato (2003, 594) disputes this logic, reasoning instead that a democratic leader is no more accountable than an autocratic leader “who is unlikely to lose office but can expect to be punished severely in the unlikely event that he is in fact removed.” “Fear” is perhaps a better word for what the autocratic leader is feeling here—certainly the leader is not “answerable” in the sense understood by political theorists (e.g., Pitkin 1972, 55–9)—but Rosato's point is worth considering. To support the contention, he reports that after participation in costly wars, a larger percentage of autocrats than democrats are removed from office, and a larger percentage are punished (594, Table 4). He finds that after losing wars, democrats, not autocrats, are more likely to be removed from office (though not punished), but he dismisses this contrary result. “This evidence is not strong,” he says, because there are so few instances of democratic losers. Rosato is right, but his evidence that autocrats are more likely to be removed from office as a consequence of involvement in costly wars is also weak. The relative infrequency of democratic involvement in both lost and costly wars argues against making much of these differences.⁴

A better interpretation of the results is that democrats tend to avoid wars they do not expect to win with modest cost. Rosato (2003, 594, note 14) rejects the plausibility of this “selection effect,” but his reasoning is suspect. He refers to Desch's (2002, 23) calculations that the marginal effect of democracy on the probability of victory is lower than the marginal effects of other predictors, like terrain and military capabilities. Even if these calculations are taken at face value, they are irrelevant. The selection effects argument is not that democratic governance per se increases the likelihood of winning, but that democracies have access to better information about the likelihood of winning—whatever the factors contributing to victory—and are more inclined to stay out of conflicts when this information suggests that war is a losing proposition.⁵

This means that militarized disputes between democracies, if they do occur, are more likely to become especially bloody affairs, and are avoided by leaders concerned with their political survival. The dyadic logic of democratic peace theory thus pertains to the probability of such nonevents, and the challenge for empirical investigation is well beyond the reach of Rosato's ex post evidence on office removal and punishment rates (Smith 1999). If fear of punishment is supposed to serve as a restraint on autocrats' propensity to resort to ill-conceived wars, what his evidence tells me is that a fair number of them have not gotten the message.

If there were a dictatorial or autocratic peace alongside the democratic peace, the causal logic explaining it almost certainly would be dyadic (e.g., Peceny, Beer, and Sanchez-Terry 2002). Rosato's autocratic constraints proposition is intriguing, to say the least, but to date the empirical evidence has not shown autocracies to be generally less disputatious than other regime types. Either way, his critique of democratic peace theory stumbles on just this point. Aware of the dyadic arguments found in the literature, he nevertheless does not take the dyadic logic of the theory seriously. If democratic dyads are more than the sum of the democratic monads, as virtually all proponents of the democratic peace maintain, then the theory does not collapse under the weight of evidence suggesting less-than-virtuous behavior by democratic states.

Among the starkest empirical anomalies for democratic peace theory are those instances of American military interventions against other, weaker democratic regimes, so Rosato is correct to once again draw our attention to such cases. However, his list of seven or eight anomalies (590, Table 2) is longer than most democratic peace researchers will concede. The U.S. intervention in a democratic Chile in 1973 is beyond dispute, and in Guyana—not formally independent in 1961—American subversion occurred during a time of limited democratic self-government. Brazil was democratic in the early 1960s, but Rosato says the U.S. role in Quadros's resignation is unclear. Guatemala might be called democratic under Arbenz, but the Polity Project locates the regime only at +2 on their democracy–autocracy scale. The other three targets of American intervention are even less democratic according to Polity: Nicaragua, Indonesia, and Iran (each with a scale value of −1). In the cases of Indonesia and Iran, Rosato's own source classifies these regimes as “bureaucracies”—that is, “institutionalized dictatorships” (Przeworski et al. 2000, 32, 65).

Regardless of how these cases are ultimately judged, most proponents of the democratic peace are probably not inclined to quarrel with Rosato's conclusion that at least some of the American interventions are at odds with the normative logic of the theory. The real difference of opinion concerns the implications of these and other anomalies for the theory-building enterprise. Throughout his critique, Rosato adopts a falsificationist stance, suggesting that in the face of historical cases that belie the causal logic he distills from the democratic peace literature, the theory should be thrown out. Actually, Rosato does not devote much effort to revealing flawed logic.⁶

Instead, he recites a list of empirical exceptions to the democratic peace—many of which are acknowledged as such by democratic peace proponents and some others that are not—while taking extra care to identify the causal mechanisms, postulated in democratic peace theory, that nevertheless seem to have gone missing in these cases. Thus, in regard to one such mechanism, he states that “whenever we find several examples of a democracy using military force against other democracies, the trust and respect mechanism, and therefore the normative logic, fails an important test” (591). Many will not agree that Rosato has refuted the dyadic hypotheses, but even accepting those particular refutations would not mean accepting that democratic peace theory itself has been falsified. The more fundamental problem is that the hypotheses Rosato derives from his rendition of democratic peace theory, and presumes to test, are too often monadic and do not square with the theory's prevailing dyadic logic.

Rosato (2003) states clearly at the outset that the democratic peace project has discovered a “powerful empirical generalization.” He simply wants to replace their theory with an explanation centering on U.S. hegemony in the Americas and Western Europe, where most democracies happen to be located during the cold war period. Although elaborating his alternative “imperial peace” theory is not the main thrust of his critique, his brief presentation of the argument does suggest that, maybe, his is—to use the distinction drawn by Lakatos (1970)—a “sophisticated,” as opposed to “naive,” falsificationism. At various places in his essay, his complaints are directed at democratic peace theory as a degenerating research program.⁷

Owen (1997), for instance, is taken to task for his attempt to “repair” the theory by introducing perceptions: to wit, what matters to democratic elites, when they contemplate resorting to force, is whether they perceive their opponents as liberal, not whether they are liberal. Elsewhere, he refers to “ad hoc” adjustments and other attempts to “rescue” the theory's logic (589–90, 596).

Scrutinizing research programs for signs that they may be degenerating is essential for scientific progress, but Waltz (1997) makes a useful point about the difference between theory and the application of theory as the target of scrutiny. In response to Vasquez's (1997) critique of neorealism as a degenerating research program, Waltz argues that although the concept of “threat” is introduced by Walt (1987) for purposes of applying balance-of-power theory to some seemingly anomalous cases, it does not thereby become part of the theory. More generally, there does appear to be a strong temptation to call on perceptions—perceptions of intentions in the case of Walt, perceptions of liberalness in the case of Owen (1997)—when the application of theory confronts discordant diplomatic behavior. Rosato is right to say that we are “unlikely to be able to predict how democracies will classify other states' regime type with a high level of confidence” (592); the temptation to revise theory ought to be resisted. However, the attempt to explain anomalies by looking more closely at the perceptions of the actors involved is a worthy endeavor, as it improves our understanding of particular events. This sort of analysis may suggest that a revision of theory is in order if, for example, actors' perceptions are shown to be systematically biased under certain conditions, but it need not. And the undertaking of such studies is not perforce an indication that a research program is degenerating.⁸

There is a curious omission from Rosato's (2003) wide-ranging critique. Although he is aware of their analysis, the game-theoretic model of the democratic peace developed by Bueno de Mesquita et al. (1999) does not receive the attention it deserves in Rosato's discussion of political accountability (593–94). The omission is curious because Bueno de Mesquita and his colleagues offer a logically coherent theory that explains not only the propensity of democracies to remain at peace with each other but also many (I think most) of the empirical anomalies that Rosato finds problematic for democratic peace theory: namely, that democracies have often fought wars for reasons other than self-defense, including colonial wars; and that democracies have often attacked or destabilized weaker, nonthreatening states, including other democracies. Their model abandons the normative logic of democratic peace theory and retains just one basic element of the institutional logic—that a democratic government depends, for its political survival, on a larger constituency (winning coalition) than does a nondemocratic government. Beyond that, all the model assumes is that political leaders do in fact want to stay in power, and the policies they pursue, which yield a mix of public and private goods, are directed toward that end. It is thus in keeping with the democratic peace research program by virtue of the centrality of regime type in the theory.⁹

Whether Rosato's (2003) “imperial peace” theory represents a progressive problemshift—again, the term is Lakatos's (1970)—relative to this or other constructive efforts within the democratic peace project remains to be seen.¹⁰

Its focus on American hegemonic power as the key explanatory factor will displease most outside the realist tradition. Be that as it may, that Rosato prefaces the brief summary of his theory by restricting its temporal and spatial scope—that is, to the post-World War II period and to the Western Hemisphere and Western Europe—is not promising.¹¹ Neither is his blanket dismissal of every causal argument contained in an alternative theory that has nevertheless received extraordinarily robust empirical support by social science standards. Parsimony may be an admirable quality of realist international relations theory, but we should be wary of essentially monocausal explanations put forward with such conviction. A virtue of the democratic peace research program has been a willingness to represent competing arguments in their multivariate models—including realist hypotheses, like Rosato's, that regional hegemony has a pacifying effect on conflict propensity. Indeed, empirical researchers working in this tradition have done much to confirm the validity of certain realist propositions, even while demonstrating the limits of realist theory. Nevertheless, there seems to be no rest for the democratic peace.