We have argued in Electing to Fight and other writings that an incomplete democratic transition increases the risk of international and civil war in countries that lack the institutional capacity to sustain democratic politics. The combination of increasing mass political participation and weak political institutions creates the motive and the opportunity for both rising and declining elites to play the nationalist card in an attempt to rally popular support against domestic and foreign rivals.
Vipin Narang and Rebecca Nelson, in their critique of Electing to Fight, agree that incompletely democratizing countries with weak institutions may be at greater risk of civil war, but they are skeptical that this extends to international war except when opportunistic neighbors invade failing states.Footnote 1 Whereas we argue that nationalism is a key causal mechanism linking incomplete democratization to both civil and international war, they conjecture that weak institutions and state failure are probably sufficient to explain why such countries may be at greater risk of armed conflict. In contrast, we have found that weak political institutions generally have little effect on a state's risk of involvement in external war when considered separately from incomplete democratization.Footnote 2 We welcome the opportunity to advance this important debate by highlighting relevant portions of our previous research and summarizing some new findings on international and civil wars.
Support for our argument rests on statistical tests and extensive case studies that trace causal processes in detail. We have presented statistical results showing the greater likelihood of war involvement for incompletely democratizing states with weak political institutions between 1816 and 1992, the greater propensity of democratizing states to engage in militarized interstate disputes, and the increased risk of civil war in incompletely democratizing states. We have also published case studies of all of the democratizing great powers since the French Revolution, all the democratizing initiators of interstate war in our statistical study, all the post-Communist states, paired comparisons of postcolonial states, and several wars involving democratizing states in the 1990s.Footnote 3 Since we published Electing to Fight in 2005, elections have heightened identity politics and fueled cross-border violence in weakly institutionalized regimes in Georgia, Iraq, Lebanon, and the Palestinian territories.
To try to advance the debate, we will address the main points on which Narang and Nelson have criticized our evidence and methods, and then we will discuss issues for further research.
Can the Theory Explain Recent Cases?
Narang and Nelson argue that recent history offers few, if any, examples that support our theory. To address this issue, we consider the set of wars that have broken out since 1992, the last year of our statistical analysis in Electing to Fight. This analysis indicates that a sizable portion of these conflicts has involved a democratizing country.
In Electing to Fight, we analyzed data on war onset compiled by the Correlates of War (COW) Project. These data extend only until 1997. However, the UCDP/PRIO Armed Conflict Dataset covers the period through 2007 and lists three interstate wars since 1992: (1) the 1998 war between Ethiopia and Eritrea, (2) the 1996 conflict between India and Pakistan that culminated in the 1999 Kargil War, and (3) the 2003 Iraq War.Footnote 4
The first two episodes were wars of democratization. Based on the Polity IV data, Ethiopia experienced an incomplete democratic transition in the decade prior to war.Footnote 5 Eritrea might also be considered incompletely democratizing, having just ratified but not implemented a democratic constitution on the eve of war. As we argued in Electing to Fight,Footnote 6 this case exemplifies our theory's causal mechanisms of nationalism, factionalism, and foreign scapegoating.Footnote 7 Pakistan's attack on Indian Kashmir in the 1999 Kargil War followed constitutional changes strengthening the elected civilian leader's authority over the military, coded by Polity as a democratic transition. The military schemed to attack Kashmir in an attempt to recapture its nationalist luster. While Polity data on institutional strength and coherence are not available after 1994, contemporary observers agree that Ethiopia, Eritrea, and Pakistan suffered significant institutional deficits in the period leading up to the outbreak of these wars.Footnote 8
In addition to these purely interstate wars, the period since 1992 has witnessed thirteen internationalized civil wars; that is, wars between a country's government and a group or set of groups within the same country, with other countries assisting either or both combatants. Of these internationalized civil wars, Polity IV provides the data needed to code the regime type of the country where the conflict occurred in eight cases. Of these eight cases, six took place in a country that experienced an incomplete democratic transition over the five- to ten-year period leading up to conflict. In three of these eight cases, both the country where the war was waged and the intervening country had recently experienced such a transition.
In Electing to Fight, we presented brief case studies of a number of these episodes, including Bosnia in 1993 and Kosovo in 1999. We also examined the 1993 war between Armenia and Azerbaijan, an event listed by UCDP/PRIO for which Polity does not provide data on the combatants' regime types. Our argument helps to explain each of these cases.
In our writing on current affairs, we have noted that incompletely democratizing Middle Eastern countries that lack strong liberal institutions have recently elected ethnic militants, sectarian extremists, terrorists, Holocaust-deniers, and nuclear proliferation advocates, which has contributed to both civil and international conflict.Footnote 9 Of course, numerous countries have democratized peacefully over the past two decades in Eastern Europe, Latin America, East Asia, and South Africa. Consistent with our theory, these countries have generally enjoyed favorable domestic conditions, including a reasonably advanced economy or relatively strong institutions at the transition's outset. Taken together, the cases discussed in this section provide substantial evidence that democratization occurring in the face of weak institutions continues to be a potent source of foreign policy belligerence and war.
Different Types of Democratizers, Different Paths to War
Narang and Nelson claim that the initiation of large-scale attacks by the national army of an incomplete democratizer should be the main path to war based on our theory, so if the democratizer is the target of attack, the theory is refuted. In fact, we argued that nationalist politics in transitional states may lead to external war through various pathways, including cross-border spillover of ethnic conflict or belligerent nationalist diplomacy that triggers foreign intervention. The evidence supporting these arguments hinges not only on statistical findings but also on extensive case studies.
Our theory distinguishes between two kinds of incompletely democratizing states: (1) those that have generally weak political institutions and (2) those that have strong administrative institutions but weak institutions for representative government.Footnote 10 Incomplete democratizers with weak institutions are prone to belligerent ethnic nationalism or sectarianism that induces neighboring states to attack. In such settings, moves to expand popular political participation often spur the nationalism of ethnic minorities, which see a chance to escape domination by culturally alien groups that control the state. At the same time, ethnic or statist nationalism of dominant groups is also likely to intensify in an effort to regain control.Footnote 11 International violence may arise because politically mobilized ethnic populations straddle international borders and because nationalism makes the diplomacy of the democratizing state rigid or belligerent. Recent cases showing this pattern include the triggering of Russian attack by democratizing Georgia's rival nationalisms and Hezbollah's provocation of Israel's attack on democratizing Lebanon.
A different pathway to war is more common among incomplete democratizers with strong administrative institutions but weak or biased representative ones, such as Germany under Chancellor Otto von Bismarck and Kaiser Wilhelm. Such states are more likely to have fairly advanced economies, differentiated social class structures, and strong militaries. They are also more likely to develop what we called counterrevolutionary nationalism. In this pattern, ruling elites are struggling to retain power in the face of rapid social change. Nationalist ideology offers them an attractive alternative to class-based appeals, and the exaggeration of foreign threats and rivalries can help them rally popular support. Strong military institutions provide a tempting tool to reinforce this strategy of rule. This kind of democratizer is less likely to be the target of attacks than those with generally weak domestic institutions and more likely to initiate wars, in the narrow sense of crossing the border first with their regular army.Footnote 12 Examples discussed in detail in Electing to Fight include nineteenth-century France as well as Prussia/Germany. More recent examples include states that alternate between military and civilian regimes such as post–World War II Turkey and Argentina.
We offered statistical evidence that incompletely democratizing states with generally weak institutions face a greater risk of war than other states. However, we were not able to statistically test our arguments about the role of nationalism in causing belligerence in incompletely democratizing states with strong administrative and weak representative institutions. Since these states score higher on our measure of institutional strength, their wars are not reflected in the statistical test of weakly institutionalized democratizers on which Narang and Nelson focus. Instead, we base our arguments about administratively stronger democratizers on causal process-tracing in case studies of all of the democratizing initiators of interstate war in our database.
In tracing causal mechanisms in the cases of war-initiating democratizers, we began with wars drawn from our statistical results when we measured regime change over ten-year periods. However, we noticed that these initiators sometimes fought a series of subsequent wars, some of which appeared in our statistical results as wars of democratization and others that did not because they occurred slightly more than a decade after a regime change. Our narratives included all the wars in such a series, because we were interested in studying how incomplete democratization could instigate and institutionalize nationalism or militarism, locking in an enduring pattern of belligerent politics. We mentioned Nazi Germany's aggression as the culmination of a series of wars fueled in this way by a deeply embedded German nationalism. Narang and Nelson see this as a mismatch between our case studies and our statistical tests. Rather, it was part of our multimethod research design, in which the statistical analyses and case studies were connected, but played a distinct role in testing different parts of the argument.
The Ottoman Cases
Narang and Nelson's principal critique of our statistical analysis is that four cases of incomplete democratization and war involvement by the Ottoman Empire in 1877, 1911, 1912, and 1913 are statistical outliers that exert disproportionate leverage on our results. They assert that “these observations are inconsistent with the Mansfield and Snyder theory, since in each case the Ottoman Empire … was being repeatedly attacked and amputated by other powers.”Footnote 13
In fact, as we show below, the Ottoman cases conform extremely well to the causal pathways to war that we expect for incomplete democratizers with weak institutions. As such, there is no reason to discard these cases. Statisticians have distinguished between two types of outliers: (1) discordant observations, which are discrepant or surprising, and (2) contaminant observations, which are products of a different data-generating process than the remaining observations in a sample. Statisticians and methodologists are virtually unanimous in recommending against discarding an observation unless a researcher is absolutely sure that the observation is a contaminant. Moreover, they agree that this evaluation can only be conducted with considerable substantive knowledge about the matter at hand.Footnote 14 The historical record reveals that these cases are not contaminant observations.
The Ottoman cases exemplify the dangers of incomplete democratization when institutional weaknesses create the motive and the opportunity for ethnic minorities and state elites to mobilize popular support through nationalist appeals. The Ottomans' democratizing reforms gave minorities a chance to organize around ethnic identities and to link up with culturally similar groups abroad. At the same time, the reforms were designed to rationalize, strengthen, and centralize state power in the hands of Turkish nationalists, which threatened the minorities' autonomy and motivated them to rebel. Repeatedly, the Turkish nationalists responded by repressing the minorities and fought wars with their foreign supporters. In addition to four international wars, the Armenian genocide and the expulsion of the Greeks after World War I were part of this overall process.
Because this history has an eerily contemporary resonance, we will recount the highlights. The Ottoman Empire, competing with modernizing great powers, was compelled by its British ally to undertake state-rationalizing reforms on the Western model in 1839. These Tanzimat reforms, which yielded a state structure that was an awkward hybrid of proto-liberal and traditional forms, failed to revitalize the state and the economy, and had the unintended effect of unleashing a struggle among religious and ethnic minorities for privileges, tax relief, and autonomy.Footnote 15 Facing a debt crisis in 1876, the Ottomans gambled on additional reforms, including the introduction of a parliamentary system with press freedoms and civil rights.Footnote 16 The short-lived constitutional reform produced exactly the pattern of politics that our theory expects when institutions are weak: self-assertion by ethnic and religious groups, factionalism in the state and its representative institutions, inability to compromise or formulate a coherent policy, brutal state actions to subjugate restive minorities, and external military intervention on the pretext of restoring human rights.
The Ottomans enjoyed a thirty-year period of relative stability with the reestablishment of autocracy between 1878 and the Young Turk “revolution from above” of 1908. The Young Turk reformers reintroduced the 1876 constitution, held multiparty parliamentary elections, increased press freedom, and replaced the millet system of self-governing religious groups with uniform Ottoman citizenship. This reform created both threats and opportunities for neighboring states and internal nationalities. In the short run, the revolution weakened the Ottoman state, allowing the Austrians, the small Balkan states, and Italy to act forcefully to achieve their aims in Bosnia, Macedonia, and Tripoli. Likewise, religious and ethnic minorities inside the Ottoman Empire took advantage of the newly granted latitude to carve out spheres of autonomy or pursue their narrow interests.
But weakness was not the only goad to violent action. Also important was the potential threat posed by the Young Turks' state-building project. The democratic, centralizing revolution, if successful, could strengthen the empire vis-à-vis external foes and internal minorities. Shaw, a leading historian of this period, claims that “the Ottoman empire was attacked, not really because of its weakness, but rather in spite of its increased strength and perhaps in fear of it.”Footnote 17 Hall explains that “the celebration of Ottoman nationhood raised concerns in Balkan capitals that the Balkan populations in a reformed Turkey would be less susceptible to their nationalistic blandishments.”Footnote 18 For example, the Albanians were initially pleased by the Young Turk revolution, which they saw as giving them a license to run their own affairs. By 1910, however, the Albanians were in revolt against the centralizing efforts of the Young Turk government to collect taxes and disarm local fighters.Footnote 19 The combination of the Young Turks' short-run weakness and their long-term ambition for centralized control created the motive and the opportunity for internal minorities, including Macedonians and Montenegrins, to join forces with external foes in the Balkan League.Footnote 20
Rising Turkish nationalism and factionalism hamstrung Ottoman diplomatic efforts to head off wars with Italy and with the Balkan League. Helmreich, the authority on this topic, writes that “bitter political dissention in Turkey … checkmated any real policy, and many officials were foolish enough to believe that the problem might be solved by undertaking a war.” He quotes the German ambassador's report that a war party in the army “charged the government with cowardliness and treason. Cowed by street demonstrations, the grand vizier promised that he would not draw back from war with the Balkan states. The first consequence of the victory of the nationalistic movement was the fear of the cabinet that the conditions of peace with Italy, which they had agreed to, might be turned into a noose for themselves. While the excited mob in the streets called for war, the ministerial council revised the peace formula.”Footnote 21
Narang and Nelson suggest that the 1911, 1912, and 1913 wars should be collapsed into a single case. However, standard historical studies and databases all list them as distinct wars. The three wars had different initiators, alliances, and outcomes. In the First Balkan War, Serbia, Bulgaria, and Greece attacked and defeated the Ottomans. After a ten-week gap filled with unsuccessful negotiations over the spoils of victory, Bulgaria attacked Serbia to initiate the Second Balkan War. The Ottomans then attacked Bulgaria, expelled non-Muslim populations from various towns and cities, retook the city of Adrianople, even invaded the territory of Bulgaria proper, and retained many of these gains in the postwar settlement.Footnote 22
In short, the Young Turk period, as with its precursor in 1876, illustrates many of the causal mechanisms of our theory. Incomplete democratization occurring in the face of weak institutions deepened political factionalism, pushed state elites to use nationalist appeals to legitimate and strengthen their rule, touched off a competition between nationalisms propounded by new and old elites, created incentives for internal groups to link up with external foes, and triggered ethnic violence, genocide, and war. Since these are precisely the causal mechanisms that link incomplete democratization and cross-border violence in various contemporary cases—including Georgia, Pakistan, and the Middle East—the Ottoman cases are exemplary, not outliers to be discarded.
Assessing the Consequences of Democratization, Not the Causes of War
Finally, Narang and Nelson argue that adding regime change and institutional strength to our set of control variables does not improve the ability of researchers to predict when a war will break out.Footnote 23 This critique, however, seems to assume that our purpose in Electing to Fight was to develop a general theory of the causes of war or to search for the set of variables that provides the best fit to the data on war. That was not our purpose. Rather, we aimed to test the specific claim that democratization is a cause of peace, which was widely believed by scholars and policymakers. It was our interest in assessing that particular hypothesis that determined the specification of our models, not some more general effort to maximize “statistical fitness.”Footnote 24 Our findings indicated that the effects of incomplete democratic transitions are both statistically significant and substantively large.
State Failure or Democratizing Nationalism?
We stand by the arguments, methods, and evidence laid out in Electing to Fight and supported by our related publications. Nonetheless, Narang and Nelson have highlighted an important question of interpretation that might usefully motivate further research on both international and civil wars. Weak political institutions play a role in the state failure thesis of Narang and Nelson and in our argument about democratizing nationalisms. In light of this overlap, what evidence can discriminate between them?
We have relied on process-tracing in case studies to determine whether nationalism was a causal mechanism leading independently to war in democratizing states. In our research on international wars, we have shown that in the vast majority of cases, elite nationalist mobilization efforts preceded and sparked belligerent, ideologically self-aware mass nationalist movements. In our work on civil wars, we similarly examined whether the democratizing state experienced mass identity-based ideological politics, such as the ethnic political mobilization in Burundi during the elections of 1993 or the religious mobilization in Algeria around the elections of 1991.Footnote 25 While we found numerous cases that did exemplify these causal mechanisms, we also found cases that might fit the state failure thesis. For example, in a number of African countries during the 1990s, conflict was loosely organized around ethnic patronage ties, but the ideological program of ethnic nationalism was not well developed. Arguably, these cases reflect anarchical rivalry in failing states, which sometimes held elections at the behest of international donors, rather than nationalist persuasion campaigns in settings of rising mass political participation.
In short, varied and numerous case studies of institutionally flawed democratizing countries demonstrate that nationalist mobilization is a prevalent and independent causal mechanism leading to international war. For civil wars, more research is warranted to sort out causality in cases such as contemporary Iraq, where state failure coincided with internationally imposed elections and sharp identity conflicts. More research is also needed on our hypothesis that attempting to democratize in the absence of necessary institutions not only runs the risk of war, but also risks creating a pattern of illiberal politics that hinders the subsequent consolidation of democracy. From a policy standpoint, none of these causal pathways reassure us that holding elections in countries with weak political institutions is likely to provide a shortcut to political stability and peace.
