Ethnofederalism,separatism,and conflict: what have we learned from the Soviet and Yugoslav experiences?

The dangers of ethnofederalism

Ethnofederal systems do not decrease the probability of conflict by addressing the mobilized groups’ grievances and fears, but actually increase that probability, the critics argue. They base their claim on three distinct causal logics. The first has to do with the mobilizational potential afforded by autonomous institutions. As Russell Hardin (1995: 30) has argued, the state’s authority is in part a function of its superior ability to coordinate its actions relative to that of its potential challengers. Autonomous institutions reduce or remove that asymmetry. Thus, using the term ‘segment-state’ for federal units or autonomous units, Roeder (2007: 111) writes:

Segment-states can solve … problems of coordination. The segment-state becomes a unique, stable focal point for expectations. Members of the elite, intelligentsia, and public are drawn to this focal point in part because they expect others to be drawn to its unique statehood. Members of the elite, intelligentsia, and public are also drawn to this nation-state project because the unique capabilities of the segment-state increase the likelihood that this nation-state project will succeed where projects without segment-states are likely to fail.

Bunce and Watts (2005: 254) similarly claim:

[Ethnofederal] arrangements can … give the minorities the institutional resources and leadership they need to press for independence – a position all the easier to embrace given the plausible impact of ethnofederalism on group isolation, intergroup distrust, and heightened competition among local elites unable to build careers outside their region and in search, therefore, of local issues they can use to mobilize and outflank their competitors.

Institutions of autonomy improve the minorities’ mobilizational capacity also by bestowing a certain legitimacy on the demands made through those institutions, which makes resistance against such demands much more complicated for the state. Svante Cornell (2002b: 254) elaborates:

Autonomous regions typically possess statelike institutions that can be crucial factors in promoting ethnic mobilization. Unlike nonautonomous minorities, minorities in autonomous regions typically have governments and parliaments that act as legitimate representatives of their ethnic constituencies and constitute legitimate decision-making bodies. Parliaments can pass language laws, refuse to accept legislation from the central government, and issue declarations of sovereignty and independence. A minority with autonomous status hence has institutions for challenging state authorities in general and its specific policies and actions in particular. A minority lacking such institutions, by contrast, would find mounting such a challenge more difficult.

The movement that the Karabagh Armenians launched in 1988 seems to be a perfect example of what these authors are describing. Soon after the upheaval in and around Karabagh began, the autonomous region’s Supreme Soviet passed a resolution demanding transfer of the region from the jurisdiction of the Azerbaijani Soviet Socialist Republic to that of the Armenian Soviet Socialist Republic. Up to that point Karabagh’s Supreme Soviet had been little more than a body for rubber-stamping decisions from above. In a changed political environment, however, it quickly became a powerful and effective instrument for pressing political demands (Cornell, 2000: 65–7; De Waal, 2004: 10–18). Moreover, it was not just the passing of this resolution that was important. The institutions of autonomy of Nagorno-Karabagh did become the focal point and the nerve system of the popular mobilization. The institutions of local autonomy in Abkhazia and South Ossetia assumed similar roles (Cornell, 2002a: 151–6, 2002b: 263–9).

The second logic, which combines a number of different processes, has to do with the identity-preserving function of autonomous institutions. One of these processes has to do with the privileges accorded to the members of titular groups of the federal units in the form of administrative appointments and economic opportunities, which provides individual members of these groups with incentives to cling to their parochial identities. In her study on how their ethnofederal designs contributed to the demise of the USSR, Yugoslavia, and Czechoslovakia, Bunce (1998: 35) argues:

The federal design in the USSR, Yugoslavia and Czechoslovakia gave a distinctly national and spatial content to the process of bargaining over power and privileges within these federations … The very process of deciding who gets what in the Soviet Union, Czechoslovakia, and Yugoslavia politicized national–territorial divisions, while giving them economic content in the process.

Taking aim at ethnic power-sharing arrangements in general, Rothchild and Roeder (2005: 35) write along similar lines:

Power-sharing institutions shape the agenda of politics and privilege issues of interethnic allocation of power and resources. Consequently, the issues that divide ethnic groups from one another come to occupy a central place in politics under power sharing, sustain interethnic conflict at high levels, and keep alive fundamental issues of renegotiating the rules of power sharing.

Their identity-preserving function is also served by the fact that autonomous institutions are typically in charge of, or have a large say in, matters of education, cultural policies, and local-language media (Cornell, 2002b: 254–5).

Finally, there is the most ambitious of the logics explaining why a state adopting an ethnofederal constitution is digging its own grave. It is rooted in the sociological theory of ‘new institutionalism,’ which looks at institutions not only as instruments for managing interaction, but also as forces that shape identities (see Hall and Taylor, 1996). As Brubaker (1996: 23) argues in his influential study of the demise of socialist federations, ‘[The Soviet Union] established nationhood and nationality as fundamental social categories sharply distinct from the overarching categories of statehood and citizenship. In so doing, it prepared the way for its own demise.’ A few paragraphs later, he discusses the ‘constitutive’, rather than the merely constraining, role of institutions, and writes:

The Soviet institutions of territorial nationhood and personal nationality comprised a pervasive system of social classification, an organizing ‘principle of vision and division’ of the social world, a standardized scheme of social accounting, an interpretive grid for public and private identities, and when political space opened up under Gorbachev, a ready-made template for claims to sovereignty. (Brubaker, 1996: 24)

Bunce (1998: 84) argues similarly that ‘national federalism constructed nations at the republican level,’ although she adds the caveat that in the case of nations ‘that were, for historical reasons, already relatively well defined’ these institutions merely reinforced their identities. Roeder (2007: 11) makes the same claim in the form of a bold counterfactual: ‘if the USSR had created economic regions (oblasts) based on economic efficiency rather than union republics based on purported nations, as the early economic planners advised, few if any nation-state projects would have been able to challenge the USSR.’ Although theoretically not quite as committed and clear-cut, and despite the fact that they present large amounts of evidence that contradicts this logic, Suny (1993) and Slezkine (1993) are also associated with this claim. Indeed, the most evocative description of the ‘new institutionalist’ logic belongs to Suny, according to whom, ‘rather than a melting pot, the Soviet Union became the incubator of new nations’ (1993: 87).

It would be unwise to deny that this literature has made important contributions to the study of nationalism under communist rule. It has been a very important corrective to the dominant sovietological paradigm, which considered the ‘nationalities problem’ as marginal at best. Its warning that ethnofederalism may not be an unproblematic solution to the problems of divided societies is also not without merit. It has serious flaws as well, however. I focus on five in particular: (1) it exaggerates the role of agency and fails to see ethnofederalism as a bargaining outcome, rather than an ideational choice; ¹ (2) it uses a wrong baseline of comparison for judging the performance of ethnofederal states; (3) it suffers from a bias induced by selecting on the dependent variable; (4) it has problems of equifinality; ² and (5) its unexamined bias against separatism is problematic both normatively and analytically. I deal with these in turn.

Was federalism a cause or an effect? The curious case of the communist embrace of nationalism in the USSR and Yugoslavia

Why did the USSR and Yugoslavia adopt ethnofederal constitutions? Any state doing so has to be concerned about losing a measure of sovereignty and control, as well as the possibility that such a concession may embolden and empower minorities to make new demands (Grigoryan, 2012; Treisman, 2007). States that subscribe to an ideology explicitly hostile to nationalism and to a philosophy of governance that emphasizes the virtues of concentration of authority should be that much more unlikely to find federal systems, particularly federal systems based on the ethnic principle, attractive. Yet ethnofederal systems are precisely what the USSR and Yugoslavia adopted despite being precisely such states. These cases, therefore, are especially interesting and puzzling.

Some of the contributors to the literature on Soviet and Yugoslav ethnofederalisms have ignored this puzzle, assuming probably that one need not understand the causes of something in order to understand its effects. Others, however, have dealt with it extensively, because they have anticipated the argument that ethnofederalism and its hypothesized effects may actually be endogenously related. Thus, according to Slezkine (1993: 414), some Bolsheviks, most importantly Lenin, sincerely believed that there was a difference between oppressor peoples’ nationalism and the nationalism of the oppressed, which they did not find inconsistent with Marxist doctrine. In addition, being ‘uncompromisingly hostile to individual rights, they eagerly, deliberately and quite consistently, promoted group rights that did not always coincide with those of the proletariat’ (Slezkine, 1993: 418). Slezkine (1993: 418) also argues that the Bolsheviks, like all missionaries, thought that their gospel could best be made accessible if communicated in non-Russian peoples’ native languages and by their native cadres. Roeder (2007: 59) offers a different solution to the puzzle. Endorsing an earlier argument made by Moshe Lewin, he argues that Lenin’s insistence on ethnofederalism was the manifestation of his desire to create a check on the bureaucracy and the party apparatus under Stalin.

The evidence, however, demonstrates that the Bolsheviks resorted to a campaign of ‘nation-building’ and decided to make the Soviet Union a federal state quite reluctantly and out of a justified fear that their entire enterprise would fail if they were to choose a different policy. They did not create or even unleash a monster, which is what much of the literature under examination here claims. They had to find a way to deal with an already unleashed monster– the mobilized nationalisms, or the simmering nationalist discontent, that would have led to new mobilizations if preemptive concessions were not made. The evidence for this is actually so overwhelming and so thoroughly studied by historians (see Carr, 1960; Connor, 1984; Martin, 2001; Pipes, 1997) that one hardly knows what to make of the claim that the Soviet nations and their nationalisms were Bolshevik creations.

Nationalists and nationalist movements in non-Russian provinces had been a major headache for the Russian empire for several decades, well before the Bolsheviks appeared in the political arena. Such movements existed in Georgia, Armenia, Poland, Finland, Estonia, Latvia, Lithuania, and Ukraine. Perhaps less advanced and articulate, but unmistakably nationalist movements had already sprung up among some of the Turkic groups of the Russian empire, particularly the Volga Tatars and the Azerbaijanis. The peoples of the North Caucasus had bitterly resisted Russian expansion in the 19th century and were no less keen on having their national rights and aspirations recognized.

Powerful nationalist mobilizations swept practically all the non-Russian regions of the empire as the twin revolutions of 1917 undermined the government’s ability to exercise authority. Poland, Finland, Estonia, Latvia, Lithuania, Armenia, Azerbaijan, and Georgia became independent states governed by nationalist parties. Nationalists in Ukraine also established control over parts of the country and declared independence. The emergence of these states was accompanied by several ethnoterritorial conflicts, including those between Poles and Ukrainians, Poles and Lithuanians, Armenians and Azerbaijanis, Armenians and Georgians, Georgians and the Abkhaz, Georgians and the Ossetians, and so on. As this evidence makes plainly obvious, national identities and nationalisms of many of the non-Russian groups predated the USSR and its ethnofederal system, which means they cannot possibly be the consequences of Soviet ethnofederalism. ³ What this evidence also suggests is that the Bolsheviks had to confront a large number of mobilized nationalisms as they embarked on the project of reincorporating the lands of the former empire into the one they were trying to build. Martin (2001: 2) writes:

Although Lenin always took the nationalities question seriously, the unexpected strength of nationalism as a mobilizing force during the revolution and civil war nevertheless greatly surprised and disturbed him. The Bolsheviks expected nationalism in Poland and Finland, but the numerous nationalist movements that sprang up across most of the former Russian empire were not expected. The strong nationalist movement in Ukraine was particularly unnerving. The direct confrontation with nationalism compelled the Bolsheviks to formulate a new nationalities policy.

What was that nationalities policy, and how did it evolve? As good Marxists, the Bolsheviks were initially opposed to any form of political organization that would rely on the ethnic principle. Nationalism, after all, was nothing more than a clever invention of the bourgeoisie designed to deceive and divide the working classes of the world, pitting them against each other, rather than against their true oppressors. Most ‘nonethnic’ Marxists (and even some ethnic ones) considered any concession to such a principle not just wrong, but heretical (Pipes, 1997: 33; Suny, 1993: 86). As Suny (1993: 86–7) informs us, Lenin was no exception in this regard, claims about his soft spot for the oppressed peoples’ nationalisms notwithstanding. Their views, however, did begin to soften over time and a search for compromises (both ideological and political) began. Temporally that shift does coincide with the mobilizations of 1917, but do we have direct evidence showing that the change in attitude was the result of these mobilizations? Actually, Russian Marxists had already been forced to reconsider their views on nationalism a little earlier when, much to their surprise and frustration, the working classes of the belligerents of WWI sided with their ‘bourgeois oppressors’ rather than with the working classes of their adversaries. They had come to realize that even if nationalism was an illusion, it was an illusion with staying power. The nationalist mobilizations in the aftermath of 1917 solidified this realization, and confronted the Bolsheviks with the need to make a political choice if they wanted to extend their rule outside the areas populated by Russians. Lenin decided that ethnoterritorial autonomies were the Bolsheviks’ only reasonable option if Bolshevik rule was going to be minimally palatable to the non-Russians. He understood very well that it was a major ideological concession and never tried to pretend otherwise. Consider the wording of the following question Lenin posed to his party comrades:

Why set up ethno-territorial autonomies under socialism if most socialists agreed that federalism was a ‘philistine ideal,’ that ‘national culture’ was a bourgeois fiction and that assimilation was a progressive process that substituted a ‘mobile proletarian’ for the ‘obtuse,’ ‘savage,’ ‘somnolent’ peasant ‘glued to his pile of manure’ and beloved for that very reason by conniving connoisseurs of national culture? (Slezkine, 1993: 417)

He answered it at the Bolsheviks’ 8th Congress in 1919, where the ‘nationalities question’ was one of the central items on the agenda, and where he had to respond to criticisms from such prominent Bolsheviks as Bukharin and Pyatakov, who considered ethnofederalism a mockery of Marxism:

Bashkirs do not trust the Great Russians, because the Great Russians are more cultured and used to take advantage of their culture to rob the Bashkirs. So in those remote places the name ‘Great Russian’ stands for ‘oppressor’ and ‘cheat.’ We should take this into account. We should fight against this. But it is a long term thing. It cannot be abolished by decree. We should be very careful here. And a nation like the Great Russians should be particularly careful, because they have provoked such bitter hatred in all other nations. (Slezkine, 1993: 421)

Lenin stated in the same speech that ‘the working masses of other nations are full of distrust towards Great Russia, as a kulak and oppressor nation,’ which could only be overcome by respecting their right to self-determination, which, of course, was defined as the right to form national homelands within the USSR. Most delegates still had misgivings about Lenin’s approach, but they voted in its favor, because, as one of those delegates pointed out, national self-determination was a ‘necessary evil,’ and it had to be tolerated even if participants of the congress did not consider it ‘normal’ or ‘desirable’ (Slezkine, 1993: 421). The name of the game, in short, was political necessity and pragmatism.

The argument is further corroborated by the next major shift in the nationalities policy. Before they had consolidated their power, the Bolsheviks had to make concessions to the non-Russian nationalities, ⁴ because they were not powerful enough to refuse these concessions and risk a large number of nationalist revolts, as well as the alienation of many ‘ethnic’ Bolsheviks. ⁵ As soon as they had consolidated their control and removed all conceivable threats to their rule, however, the Bolsheviks, or rather the communists as they had started calling themselves by then, radically changed their nationalities policy in the 1930s. Much has been said and written about the terror Stalin unleashed against the different segments of Soviet society – the peasantry, the intelligentsia, the officer corps, and so on. Historians have only recently begun to draw attention to the fact that certain groups, and with them the entire edifice of the nationalities policy, had become Stalin’s targets as well. The rhetoric about Great Russian chauvinism and the need for empowering the non-Russians was replaced by the rhetoric of Russia’s and the Russians’ special role in the union. Prominent non-Russian intellectuals started becoming targets of accusations of nationalism. Many of them wound up dead or in labor camps in Siberia. Even policies that until recently have not been regarded as associated with this shift turn out to have been closely linked to it. Thus, the dekulakization campaign was not just directed against the peasantry, and did not just aim to establish control over agricultural production. It was simultaneously a campaign of mass murder against the Ukrainians, with the aim of terrorizing them and shifting the republic’s demographic balance. The terror campaigns of the 1930s were aimed at removing all potential opposition and scaring the population into total obedience, as well as populating and absorbing the wastelands in Siberia and Kazakhstan, but they were also serving the cause of the Soviet melting pot. The policy continued during and after the war with the mass deportations of the Chechens, Ingush, Karachai, Crimean Tatars, and Volga Germans, as well as the forcible removal to Siberia of large numbers of Moldovians and citizens of the Baltic republics. In most of these cases, the murdered or deported populations were replaced by Russian and non-Russian settlers from other parts of the country (Martin, 2001; Snyder, 2010). The terror against the Soviet nationalities stopped after Stalin, but Moscow remained ever vigilant toward expressions of ‘anti-Soviet nationalism.’ It also vigorously pursued the melting-pot ideal. Now it was not done by forceful deportations of people to other areas of the country, but by creating incentives for the Russification of education in the non-Russian republics, as well as through grandiose projects such as the development of Kazakhstan or the Baykal–Amur railway, which would provide incentives for people from all over the USSR to leave their home republics and move to the sites of those projects.

In sum, there is little in this story to suggest that the Bolsheviks opted for an ethnofederal system because of some puzzling belief in the inherent goodness of that system, Lenin’s soft spot for minority nationalisms, or because they failed to anticipate the consequences of institutionalizing ethnicity. They did so because they had to placate the already mobilized nationalities of the Russian empire in order to make Bolshevik rule minimally palatable to them. Claiming that they made the wrong decision, and that the Soviet Union would not have faced the problems it did in 1988–91 had its founders rejected the idea of ethnofederalism, therefore, is a non sequitur. ⁶ It implies that they had a choice, when, in fact, they did not.

It is probably true that not every detail of the policy can be explained by necessity. For example, the particular zeal and enthusiasm with which the Bolsheviks embarked on the policy of indigenization was probably a function of their deeply inculcated radicalism, which characterized every aspect of their behavior, rather than sheer necessity. Nor can we explain why Kazakhstan’s and Kyrgyzstan’s borders were drawn where they were. But the available evidence leaves little doubt that political institutionalization of ethnicity in some form was fundamentally inescapable and that the Bolsheviks adopted the policy they did because they understood that. Scholars pointing to the USSR’s ethnofederal design as the cause of its demise do not show sufficient appreciation for this fact, which is why they end up drawing the causal arrow from federalism to separatism, and not in the other direction, or at least not also in the other direction.

The argument about the pernicious effects of ethnofederalism does no better in the Yugoslav case. From its creation in 1919, the Yugoslav Communist Party had to develop a strategy for dealing with the problem of mobilized nationalities, facing constraints and pressures very similar to those faced by the Bolsheviks. Like the Bolsheviks, Yugoslav communists eventually succumbed to these pressures and adopted an ethnofederal constitution after a long period of debate and soul-searching. Like the Bolsheviks, they considered ethnofederalism a ‘necessary evil.’ And even more so than in the case of the USSR, the real puzzle in Yugoslavia is not that a party calling itself communist opted for an ethnofederal constitution in 1946, but the fact that it somehow managed to hold Yugoslavia together.

Being a socialist and revolutionary force, Yugoslav communists were focused primarily on class struggle and revolution during the initial period of their existence. But, as Paul Shoup writes in what remains the best study of the evolution of the Yugoslav communists’ views on ethnofederalism, ‘there was never complete agreement on how this objective was to be attained … [T]he task of the Communists was made infinitely more difficult by the atmosphere of continual tension and controversy prevailing in the country as the result of the national problem’ (1968: 14). There were debates among the Yugoslav communists about how socialism and the national question related to each other throughout the interwar period, that were mirror images of the debates raging among the Bolsheviks. For example, Sima Markovic, the most prominent Serbian communist at the time, argued that nationalism was fundamentally incompatible with class struggle and was a purely bourgeois phenomenon. He was against federalism as a result (Shoup, 1968: 25). Markovic was widely criticized, however, particularly by Croatian socialists, who were always anxious to balance Serbian dominance, and who, therefore, favored mixing their national aspirations with socialism. The latter approach became more dominant in the party discourse, because too vigorous an opposition to it would have driven the Croats and other non-Serbian groups away.

The commitment to decentralization remained strong and in the 1930s national communist parties were created in Slovenia and Croatia, with Macedonian communists considering such a move as well. They all had to appear sufficiently nationalistic in order to compete with their own ‘bourgeois nationalists’, which was identical to the behavior of the non-Russian communists in the USSR. Meanwhile, the relationship between the different Yugoslav nationalities (and not just their communists) remained tense, with the war and the Nazi occupation of Yugoslavia pushing things from bad to disastrous. Large-scale violence took place during the war that pitted Croats against Serbs and Serbs against Muslims. At the end of the war, the situation with the Albanian minority also deteriorated to the point of an open rebellion, which was put down by force.

Even this truncated account shows that Tito and the Yugoslav communists were not starting with a clean slate after they became the dominant political force in Yugoslavia. They had taken over a country ravaged by ethnic antagonisms and a recent history of large-scale interethnic violence. The communists were fortunate to have as much support as they did, which was due to their leadership of the partisan movement and well-deserved credit for helping to defeat the Nazis. That support allowed them to hold the country together, but they understood very well that had they insisted on a unitary constitution, they would have had not just the Albanian rebellion on their hands, but several others, and that would have almost certainly provoked extensive violence and destroyed the prospects for a unified Yugoslavia. As in the case of the Soviet Union, therefore, the critics of Yugoslav ethnofederalism confuse the effect with the cause. As in the case of the Soviet Union, they make a claim about a hypothetical non-ethnofederal Yugoslavia’s different, and better, fate, which violates the cotenability rule.

Beyond counterfactuals

The analysis in the preceding section is at least in part a criticism of a counterfactual with another counterfactual: had the Bolsheviks and Yugoslav communists insisted on unitary constitutions, they would have provoked a fierce nationalist resistance and made the very creation of these states impossible. This, I believe, is a much sounder counterfactual, but it is a counterfactual, nonetheless. The question is whether we can do better than that. We clearly cannot test that specific claim empirically. We can, however, subject the generalized version of that claim to empirical examination. More specifically, we can identify instances when states reject mobilized minorities’ demands for group rights, and examine the outcomes. If rejection turns out to produce intensified conflict, then we should be even more skeptical of the argument that ethnofederalism was fatal for the USSR and Yugoslavia and more skeptical of prescriptions that advocate resisting institutionalization of group identity.

Drawing attention to cases in which demands for group rights were rejected is important for yet another reason. The critics of ethnofederalism treat the propensity for violence of all nonfederal states as their basis of comparison for determining how prone to separatism and violence ethnofederal states are. But that is surely not the correct baseline, because comparing the stability of monoethnic Japan with that of multiethnic Yugoslavia does not tell us much. Indeed, even multiethnic states that are not confronted with demands from mobilized minorities should be considered irrelevant. The focus should be exclusively on cases in which there were demands for group rights that were rejected. There are plenty of such cases, but I will limit my discussion to those from the former USSR and Yugoslavia due to space limitations and because some of these cases have provided fodder for the critics of ethnofederalism.

The wars in Croatia and Bosnia-Herzegovina resulted from nothing more than rejecting the demands of these republics’ Serbian minorities. After it became clear that neither holding Yugoslavia together nor separating the Serbian-populated areas of Croatia and Bosnia-Herzegovina were going to be possible, the Serbian community in Krajina demanded autonomy in Croatia and the Bosnian Serbs demanded an ethnofederal constitution. Following very similar paths, Croats and Bosnian Muslims refused to grant these demands and unilaterally declared independence after receiving signals of international support. The Serbs then decided to achieve through violence what they had failed to achieve at the negotiating table (Burg and Shoup, 1999: 108–20; Johnstone, 2002: 44; Woodward, 1995: 105–11, 189–98). But what about the post-Soviet wars? Surely rejecting demands for autonomy could not have been the problem in those cases, because with the exception of the war in Transnistria, all the other wars there (Abkhazia, South Ossetia, Nagorno-Karabagh, and Chechnya) involved groups that already possessed autonomy. Moreover, as I have already pointed out, the autonomous institutions played important mobilizational roles in those cases. We should also not forget that at least in some of these cases the minorities were not just demanding status upgrades, but outright secession. Finally, Cornell (2002b) draws attention to the striking fact that while the disputes with two groups endowed with autonomy in Georgia (the Abkhaz and the Ossetians) escalated to war, the disputes with two other large, compactly settled, but nonautonomous minorities (the Armenians of Javakheti and the Azerbaijanis of Kvemo-Kartli) did not, which seems to make the case against autonomy even stronger.

A closer look at these cases, however, reveals a number of problems. Some of them (such as the problems with case selection and equifinality) I will deal with later. What I will focus on here is the problematic interpretation of the evidence, whereby the initial demands by the autonomous minorities are presented as sufficient causes of war, while the behavior of their adversaries (that is, the union republics) is simply assumed to be irrelevant. It should not be, for the violent outcomes in those cases were not just the consequences of the minorities’ actions, but of the interaction of the minorities and their adversaries. Thus, although both Karabagh Armenians and the Abkhaz started out with demands to be taken out of Azerbaijan’s and Georgia’s jurisdictions respectively, their demands changed over time as their respective conflicts with Azerbaijan and Georgia unfolded. As Melander (2001: 48–50) reports, the position of the Karabagh Armenians was moderated considerably by the spring of 1991, when some Armenian leaders began arguing that Karabagh should negotiate with the authorities in Baku instead of hoping for a favorable solution to be imposed from the Kremlin. De Waal (2004: 118) informs us, similarly, that on 19 June 1991 officials in the suspended regional Soviet passed a resolution stating their intention to change ‘the course from the policy of confrontation to a policy of dialogue and negotiation.’ The Azerbaijani response was to harden its position, and the subsequent war was the inevitable consequence. ⁷

The Abkhaz conflict had a very similar trajectory of escalation with one interesting difference. Despite its hypernationalistic rhetoric and preferences, Georgia’s first post-communist government under Zviad Gamsakhurdia had negotiated an important compromise with the Abkhaz in the summer of 1991: a quota system for the elections to Abkhazia’s Supreme Soviet. The Abkhaz obtained the majority of seats in the Soviet, although they constituted only 17 percent of the population of Abkhazia at the time. This compromise was far from having solved all the problems, but the tensions in Abkhazia had subsided substantially. It had also created the potential groundwork for further negotiations and compromises on Abkhazia’s status. In December 1991, however, the Soviet Union collapsed and then Gamsakhurdia’s regime was overthrown in January 1992. The new regime decided to do something which effectively nullified the earlier bargains and presented the autonomous regions, including Abkhazia, with a new challenge: it restored pre-Soviet Georgia’s constitution, which defined Georgia as a unitary state. The Abkhaz let it be known that this was unacceptable, but they did not retaliate with a declaration of independence. What they did was propose a model of a ‘confederal’ constitution for Georgia, which meant in essence that the Abkhaz were willing to forgo secession if an acceptable degree of autonomy for the region could be negotiated (International Crisis Group, 2006: 5). The new authorities in Georgia did not even respond to the Abkhaz offer, which led to the Abkhaz declaration of independence and the subsequent Georgian attack on Abkhazia in August 1992. As for South Ossetia, even its initial demand was not separation from Georgia. What it requested in 1989 was an upgrade of its status from an autonomous region to an autonomous republic within Georgia. The request was sent not to the Kremlin, but to the Georgian Supreme Soviet (Zürcher, 2007: 124). The Georgian response was a raid of 30,000 people on the Ossetian capital under the leadership of Gamsakhurdia, who was at the time the leader of the opposition, but who was already calling the shots. The result was Ossetian radicalization and the low-intensity war that ensued.

What about the nonviolent outcomes in Javakheti and Kvemo-Kartli? Do the peaceful outcomes in those cases not strengthen the case against autonomy? The answer has to be no, because it was not for lack of the capability to mobilize that the disputes in these regions did not escalate. The peaceful outcomes in these cases were due to a different factor, which Cornell himself acknowledges and which I will discuss later in more detail in the context of the problems of equifinality. Let us recall also that the Serbs in Croatia and Bosnia had no autonomous structures, but that did not hinder their mobilization, nor did the lack of autonomous status hinder the mobilization in Transnistria. In fact, it is to point out the obvious that the vast majority of nationalist movements and mobilizations in the past 150 years have taken place without any autonomous institutions.

In sum, even if we grant that autonomous institutions sometimes facilitate separatism and conflict in the long run, we see that refusing to satisfy such demands for such institutions or for upgrades may generate exactly the same result – something the critics of ethnofederalism ignore. Doing so, by the way, amounts to a bias induced by selecting on the single value of the independent variable, the implications of which are much more poorly understood and much less studied than those of the bias induced by selecting on single values of the dependent variable – the problem I turn to next.

Dogs that refuse to bark: ethnic federations that work and autonomies that do not produce separatist wars

The critics of ethnofederalism have not taken sufficient care to avoid one of the most common biases in social scientific research – that of selecting cases on single values of the dependent variable. ⁸ What they have done is to select cases of failure and conflict and try to trace the causes of those failures and conflicts back to their ethnofederal roots. Had they allowed variation in the dependent variable, they would have been able to identify cases of success as well. Perhaps the best example of such a success is post-Soviet Russia, which has not only retained its ethnofederal structure inherited from the USSR, but expanded it. The result is best described in an article in The Economist (2010: 63) about Tatarstan, but its argument can be generalized to the other autonomous regions of Russia:

After the Soviet collapse in 1991, [the President of the Federal Republic of Tatarstan in Russia] Mr Shaimiev made full but cautious use of Boris Yeltsin’s offer to ‘take as much sovereignty as you can swallow.’ Mr Shaimiev’s skill and Mr Yeltsin’s belief in federalism as the only plausible model for post-imperial Russia prevented the further disintegration of the country and kept Tatarstan on an even keel.

Chechnya was the only autonomous region of Russia where the policy did not work, but that was because the Chechens’ demands were indeed unappeasable. Had the Russians chosen the same policy Croatia or Georgia chose, they would have had wars not just with the Chechens, but with several other minorities as well, which brings us back to the question of how we evaluate the conflict-proneness of autonomous regions in the post-Soviet space. Cornell’s method is to point out that four out of the five wars in the former Soviet Union were fought between an autonomous region and a union republic, which at first glance seems like a very high proportion. The association of autonomy with conflict seems even more robust when Cornell contrasts the violent cases with the nonviolent outcomes in Javakheti and Kvemo-Kartli, because he is able to demonstrate variation in both the dependent and independent variables. But Cornell’s method is actually not a good one. If we want to measure how war-prone autonomous regions were, it is not the proportion of the wars involving those regions to all wars that we should look at, but the proportion of wars involving autonomous regions to all autonomous regions. That number is 8 percent. ⁹ Measured thus, the correct way, autonomy seems far less dangerous than Cornell’s method implies, even if we do not question the claim that autonomous status was the cause in the violent cases, which we should, in fact, question.

What about the success stories outside of the former USSR? Stepan (1999: 20) writes:

the six longstanding democracies that score the highest on the index of linguistic and ethnic diversity – India, Canada, Belgium, Spain, Switzerland, and the United States – are all federal states. The fact that these nations chose to adopt a federal system does not prove anything; it does suggest, however, that federalism may help these countries manage the problems that come with ethnic and linguistic diversity. In fact, in my judgment, if countries such as Indonesia, Russia, Nigeria, China, and Burma are ever to become stable democracies, they will have to craft workable federal systems that allow cultural diversity, a robust capacity for socioeconomic development, and a general standard of equality among their citizens.

Along similar lines, Kohli (1997: 325) says the following about India:

two dimensions of the political context seem to be especially relevant, namely, how well central authority is institutionalized within the multicultural democracy and the willingness of the ruling groups to share some power and resources with mobilized groups.

Bermeo (2002) has presented statistical evidence, meanwhile, that armed rebellions are fewer in federal compared with unitary states. Interestingly, Bermeo (2002: 99) also reports that federal states are four times less likely to experience armed rebellion than unitary states when the sample is restricted only to dictatorships. Hoddie and Hartzell (2005) demonstrate that power-sharing arrangements in general increase the likelihood of a peaceful and durable settlement to civil wars. Burgess and Pinder (2007) present a set of case studies of successful ethnofederations such as Canada, Belgium, and Switzerland. Finally, I would argue that we should be skeptical even of the claim that the Yugoslav and Soviet ethnofederations were failures. The literature under examination implicitly insists that their eventual failures should be the operational criterion for the definition of failure. But how uncontroversial is that criterion really? Is it unreasonable to argue that given the monumentally difficult tasks at hand, the ethnofederal designs of these countries performed quite well? Is it not possible that the USSR and Yugoslavia lasted as long as they did not despite, but thanks to their ethnofederal designs? Such claims, in fact, are not unknown in the literature. For example, Yugoslavia’s ethnofederal makeup is one of the variables Woodward (1995: 29–41) credits, correctly in my opinion, for the stability of the country in the period between WWII and the crisis in the 1980s.

Dogs that are not even dogs: other variables doing the work of autonomy and ethnofederalism

One important objection I can anticipate to the analysis in the preceding pages is that I have mischaracterized the claims of the critics of ethnofederalism as monocausal, whereas many of them explicitly rely on additional variables to explain both the disintegrations of communist federations and the conflicts that ensued in their wake. Thus, Cornell (2002b: 269–73) invokes the role of third parties in addition to the presence or absence of an autonomous status to explain why the conflicts in Abkhazia and South Ossetia escalated, while those in Javakheti and Kvemo-Kartli did not. He argues, in particular, that the non-escalation in the latter two regions was in part the result of the restraining role of the Armenian–Azerbaijani conflict, which created incentives for Armenia and Azerbaijan, as well as their co-ethnics in Georgia, to behave with moderation. That, he maintains, was in stark contrast with Russia’s behavior in Abkhazia and South Ossetia.

This indeed is not a monocausal argument. But it is one with a problem of equifinality. The added variable is not interacting with the original one in any meaningful way, but rather seems to replace it. An analogy would make the problem more transparent. If somebody with a heart disease died of a heart attack, it would be perfectly legitimate to argue that it was not just the heart disease, but also the person’s high-fat diet that killed him. This would explain why some people with the same heart disease are able to live longer than others. Now contrast this with a situation in which our hypothetical person with a heart disease has been fatally shot. Clearly, the heart disease in this case could not be a part of the explanation for the person’s death. If we were to insist on including it, we would be assigning a causal role to a variable that had no effect on the outcome. Alternatively, if we were not entirely sure which of the two causes killed the person, including both would overdetermine the outcome. In Cornell’s argument, the ‘third-party variable’ is not explaining some additional variance. It is doing all the work, because its effect is completely independent of whether the particular minority has an autonomous status. Insisting that both the autonomous structures and the third parties affected the outcomes in Georgia’s conflicts with minorities, therefore, is tantamount to explaining explaining more than 100 percent of the outcome.

At first glance the employment of additional variables by Roeder, Bunce, and Suny in their attempts to explain the success of nation-state projects (Roeder) or the failures of communist federations (Bunce and Suny) seems less problematic, because these authors have attempted to explain how the additional variables interact with ethnofederalism. Thus, Roeder (2007: 9–10) writes that the success of nation-state projects is the result of the perfect confluence of ‘identity, grievance, greed, mobilization, and international recognition.’ He then elaborates:

For the proponents of a nation-state project to advance to sovereign independence, all of these elements must align so that they are mutually reinforcing. Misalignment in any one element can create an insurmountable obstacle to success. Misalignment is the reason why so few projects succeed. The argument in this book turns our attention to the question, what could possibly lead all of these elements to align favorably? Perhaps this alignment can result from simple luck or coincidence, but that is unlikely. Rather, I argue that there is a common overarching constraint that has increased the likelihood of such an alignment: almost every successful nation-state project has been associated with an existing institution I refer to as a ‘segment state.’ Independence represented the administrative upgrade of this existing jurisdiction. For example, after the demise of the USSR, the successful nation-state projects were the projects associated with the first order jurisdictions called union republics, such as Kazakhstan and Ukraine. The nation-state projects not associated with these segment states, such as the projects for Turkestan, Idel-Ural, the Mountain Republic, or Novorossia, failed in the 1990s. I will call this argument the segmental institutions thesis. (Roeder, 2007: 9–10)

Bunce (1998: 34–5), meanwhile, focuses on the effects of economic decline on the fortunes of the USSR and Yugoslavia, and argues:

beginning in the 1970s, virtually all the socialist regimes in Europe were in a difficult bind. An expanding surplus was vital to the survival of the regime, because it was a major component of both regime legitimacy and political stability – the former because of regime ideology and the latter because of the need to keep intraelite conflict within bounds while pacifying a restive and expectant public.

Suny (1993: 127–8) argues that the Soviet Union was brought down by what he calls ‘Gorbachev’s dilemma.’ Its essence was Gorbachev’s inability to reform the USSR’s economically inefficient system without allowing or even encouraging some bottom-up mobilization in order to put pressure on that system. The problem was that it was impossible to control the content, the direction, or the intensity of that mobilization.

There is much to agree with in those arguments. I think Suny is exactly right about ‘Gorbachev’s dilemma’ being the proximate cause of the disintegration of the Soviet Union. Bunce’s focus on how stunted economic growth affected the fortunes of communist federations is also correct and complementary to Suny’s argument. At least that part of Roeder’s argument in which he emphasizes the importance of external support for the success of nation-state projects is also difficult to disagree with. The problem is that the real interaction in all of their arguments is not with ethnofederalism, but with nationalism, which seems to have been lost in the fog of adjusting and qualifying the original claim regarding the perils of ethnofederal designs. ‘Gorbachev’s dilemma,’ or rather the inability to find a way out of it, was indeed what destroyed the USSR, but that is precisely the same dilemma the Austro-Hungarian and Russian empires faced in the second half of the 19th century and the early decades of the 20th century. They were no better (or worse) at finding a solution, and neither one had an ethnofederal system.

I am not sure international support is a necessary condition for successful nation-state projects, nor am I sure how much of a role it played in the Soviet case, but it certainly was very important in the disintegration of Yugoslavia, and in many other cases. Again, however, we can easily recall successful nation-state projects which succeeded in large measure due to international support although these projects were not associated with ‘segment states.’ The Greeks were not associated with anything like a ‘segment state’ in the Ottoman empire, but they had considerable external support when they launched their nation-state project. The same is true for the Serbs and Bulgarians. In fact, practically all the Eastern European nations that are independent today launched their nation-state projects without association with a ‘segment state,’ and practically all of them had external support at different stages of their struggles. So, even if it is true that most such successful projects in the more recent past have been upgrades of existing ‘segment states,’ such association does not seem to be an ‘overarching constraint’, nor does it seem to be a necessary condition for external support. As for Turkestan, Idel-Ural, the Mountain Republic, and Novorossia, these projects failed not because they were not associated with ‘segment states,’ but because they were anything but nation-state projects.

Where do we go from here?

The central conclusion to be drawn from the preceding analysis is that many students of ethnofederalism, including both its critics and its proponents, may have been asking the wrong question. The question they have been asking is whether ethnofederalism ameliorates or exacerbates separatism and conflict, which in part is the result of thinking of ethnofederalism as a variable exogenous to that very separatism and conflict. Once we think of ethnofederalism as a bargaining outcome, however, the question becomes somewhat meaningless. ‘What makes ethnofederal bargains stable or unstable?’ becomes the meaningful and interesting question.

Some attempts to answer this question have already been made. Hale (2004) and Hale (2005), for example, have proposed a demography-based theory of the stability of ethnofederal arrangements, arguing that federations with core ethnic groups that are concentrated in one of the federal units are much less stable, whereas those with core groups dispersed among the other units are more stable. Wibbels and Bakke (2006) have argued that whether federations succeed or fail depends on regional inequalities, whether minorities are excluded from ‘national governing parties,’ whether the minorities are regionally concentrated, and on the interaction between such concentration and increased fiscal transfers.

These studies are a good beginning, but they are just that. Many unanswered questions remain. For example, it is not clear what exactly the relationship is (or should be) between fiscal transfers and a propensity for separatism. Even if we have statistical evidence that such transfers reduce incentives for separatism, we have quite a few important anomalies that need to be explained. Kosovo was the most heavily subsidized region of Yugoslavia, and at the same time one of the most determined to secede. Bosnia-Herzegovina and several Soviet republics that favored secession were heavily subsidized as well. Given the intuitively appealing logic that subsidies from the center should create disincentives for subsidized regions to secede, this seems to be an interesting puzzle.

It would also be interesting systematically to examine the effects not only of static demographic variables on the fortunes of ethnofederations, but of demographic changes. Even if they did not cause the conflicts per se, anxieties about demographic changes affected the intensity of the conflicts in Kosovo, Nagorno-Karabagh, and Abkhazia. The separatist ideologies of some of the republics in the USSR (particularly Georgia and the Baltic republics) were infused with such anxieties. Another important and largely neglected question has to do with the effects of external threats against members of a federation. The logic here is basically the same as that of seeking improved security by joining an alliance. The stability of federations, just like the stability of alliances, should in part be a function of such threats. ¹⁰

The stability of federal bargains is not the only issue that should inform the future research agenda of the study of ethnofederations. Future research should also question two unqualified and related assumptions central to this literature: that separatism is necessarily a bad thing and that separatism equals violence. I would argue that separatism is only bad if on balance it increases oppression, which is why some liberals claim to have such distaste for it. But, surely, separatism can increase freedom. ¹¹ Indeed, unqualified opposition to separatism is an indirect endorsement of imperialism, which is normatively far more problematic than endorsing separatism.

As for the identification of separatism with violence, many authors seem to forget that separatist conflicts are interactions with two actors. Clearly, states bear at least part of the responsibility for separatist violence, for there would be no such violence if they did not resist separatists’ demands. Indeed, such identification is problematic not only normatively, but empirically as well. Peaceful secessions are not unknown. That is true even for most of Yugoslav and Soviet secessions, which raises an interesting question: are secessions and post-secession environments more or less likely to be peaceful if the disintegrating multiethnic state is an ethnofederation? I would venture to hypothesize that, in fact, federally organized multiethnic states are less likely to generate violence when they disintegrate, because the seceding federal units have defined borders and governance structures, which should have conflict reducing effects, especially if those units’ borders are congruent with the demographic boundaries of their titular groups. Collapsing unitary multiethnic states do not have either such institutions or such borders, which should create more opportunities for friction and conflict.

I end with anticipating and responding to an important and predictable objection, which has to do with how I view nationhood. I have treated it as prior to the institutions of the state, which can easily trigger accusations of ‘essentializing’ it. That is not the case, for I firmly reject the essentialist position. But I also reject the assumption about the extreme malleability of national and ethnic identity that underpins much of the constructivist and ‘new institutionalist’ theory of nationalism. It is simply undeniable that nations with stable identities often exist independently of states and their institutions, and are often prior to them. Dismissing this fact as essentialism is intellectually indefensible, while advocating policies that are designed to circumvent it is sure to contribute to a world that is simultaneously less just and less peaceful.