Copyright © 1999 National Endowment for Democracy and the Johns Hopkins University Press. All rights reserved.

Journal of Democracy 10.4 (1999) 19-34

Federalism and Democracy:
Beyond the U.S. Model

Alfred Stepan


For those of us interested in the spread and consolidation of democracy, whether as policy makers, human rights activists, political analysts, or democratic theorists, there is a greater need than ever to reconsider the potential risks and benefits of federalism. The greatest risk is that federal arrangements can offer opportunities for ethnic nationalists to mobilize their resources. This risk is especially grave when elections are introduced in the subunits of a formerly nondemocratic federal polity prior to democratic countrywide elections and in the absence of democratic countrywide parties. Of the nine states that once made up communist Europe, six were unitary and three were federal. The six unitary states are now five states (East Germany has reunited with the Federal Republic), while the three federal states--Yugoslavia, the USSR, and Czechoslovakia--are now 22 independent states. Most of postcommunist Europe's ethnocracies and ethnic bloodshed have occurred within these postfederal states.

Yet in spite of these potential problems, federal rather than unitary states are the form most often associated with multinational democracies. Federal states are also associated with large populations, extensive territories, and democracies with territorially based linguistic fragmentation. In fact, every single longstanding democracy in a territorially based multilingual and multinational polity is a federal state. [End Page 19]

Although there are many multinational polities in the world, few of them are democracies. Those multinational democracies that do exist, however (Switzerland, Canada, Belgium, Spain, and India), are all federal. Although all these democracies, except for Switzerland, have had problems managing their multinational polities (and even Switzerland had the Sonderbund War, the secession of the Catholic cantons in 1848), they remain reasonably stable. By contrast, Sri Lanka, a territorially based multilingual and multinational unitary state that feared the "slippery slope" of federalism, could not cope with its ethnic divisions and plunged headlong into a bloody civil war that has lasted more than 15 years.

In addition to the strong association between multinational democracies and federalism, the six longstanding democracies that score highest on an index of linguistic and ethnic diversity--India, Canada, Belgium, Switzerland, Spain, 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, however, suggest 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.

Consider the case of Indonesia, for example. It seems to meet all the indicators for a federal state. It has a population of over 200 million, and its territory is spread across more than 2,000 inhabited islands. It has great linguistic and ethnic fragmentation and many religions. Thus it is near the top in virtually all the categories associated with federalism. If Indonesia is to become a democracy, one would think that it would have to address the question of federalism or decentralization. Yet at a meeting of Indonesian political, military, religious, and intellectual leaders that I attended after the fall of Suharto, most of the participants (especially those from the military) rejected federalism out of hand because of secessionist conflicts at the end of Dutch colonial rule. Indonesia should at least consider what I call a federacy to deal with special jurisdictions like Aceh or Irian Jaya. A federacy is the only variation between unitary states and federal states. It is a political system in which an otherwise unitary state develops a federal relationship with a territorially, ethnically, or culturally distinct community while all the other parts of the state remain under unitary rule. Denmark has such a relationship with Greenland, and Finland with the Aaland Islands.

A Misleading Picture of Federalism

In seeking to understand why some countries are reluctant to adopt federal systems, it is helpful to examine what political science has had [End Page 20] to say about federalism. Unfortunately, some of the most influential works in political science today offer incomplete or insufficiently broad definitions of federalism and thereby suggest that the range of choices facing newly democratizing states is narrower than it actually is. In large part, this stems from their focusing too exclusively on the model offered by the United States, the oldest and certainly one of the most successful federal democracies.

One of the most influential political scientists to write about federalism in the last half-century, the late William H. Riker, stresses three factors present in the U.S. form of federalism that he claims to be true for federalism in general. 1 First, Riker assumes that every longstanding federation, democratic or not, is the result of a bargain whereby previously sovereign polities agree to give up part of their sovereignty in order to pool their resources to increase their collective security and to achieve other goals, including economic ones. I call this type of federalism coming-together federalism. For Riker, it is the only type of federalism in the world.

Second, Riker and many other U.S. scholars assume that one of the goals of federalism is to protect individual rights against encroachments on the part of the central government (or even against the "tyranny of the majority") by a number of institutional devices, such as a bicameral legislature in which one house is elected on the basis of population, while in the other house the subunits are represented equally. In addition, many competences are permanently granted to the subunits instead of to the center. If we can call all of the citizens in the polity taken as a whole the demos, we may say that these devices, although democratic, are "demosconstraining."

Third, as a result of the federal bargain that created the United States, each of the states was accorded the same constitutional competences. U.S. federalism is thus considered to be constitutionally symmetrical. By contrast, asymmetrical arrangements that grant different competencies and group-specific rights to some states, which are not now part of the U.S. model of federalism, are seen as incompatible with the principled equality of the states and with equality of citizens' rights in the post-segregation era.

Yet although these three points are a reasonably accurate depiction of the political structures and normative values associated with U.S. federalism, most democratic countries that have adopted federal systems have chosen not to follow the U.S. model. Indeed, American-style federalism embodies some values that would be very inappropriate for [End Page 21] many democratizing countries, especially multinational polities. To explain what I mean by this, let me review each of these three points in turn.

"Coming-Together" vs. "Holding-Together"

First of all, we need to ask: How are democratic federal systems actually formed? Riker has to engage in some "concept-stretching" to include all the federal systems in the world in one model. For example, he contends that the Soviet Union meets his definition of a federal system that came about as the result of a "federal bargain." Yet it is clearly a distortion of history, language, and theory to call what happened in Georgia, Azerbaijan, and Armenia, for example, a "federal bargain." These three previously independent countries were conquered by the 11th Red Army. In Azerbaijan, the former nationalist prime minister and the former head of the army were executed just one week after accepting the "bargain."

Many democratic federations, however, emerge from a completely different historical and political logic, which I call holding-together federalism. India in late 1948, Belgium in 1969, and Spain in 1975 were all political systems with strong unitary features. Nevertheless, political leaders in these three multicultural polities came to the decision that the best way--indeed, the only way--to hold their countries together in a democracy would be to devolve power constitutionally and turn their threatened polities into federations. The 1950 Indian Constitution, the 1978 Spanish Constitution, and the 1993 Belgian Constitution are all federal.

Let us briefly examine the "holding-together" characteristics of the creation of federalism in India to show how they differ from the "coming-together" characteristics correctly associated with the creation of American-style federalism. When he presented India's draft constitution for the consideration of the members of the constituent assembly, the chairman of the drafting committee, B.R. Ambedkar, said explicitly that it was designed to maintain the unity of India--in short, to hold it together. He argued that the constitution was guided by principles and mechanisms that were fundamentally different from those found in the United States, in that the Indian subunits had much less prior sovereignty than did the American states. Since they had less sovereignty, they therefore had much less bargaining power. Ambedkar told the assembly that although India was to be a federation, this federation was created not as the result of an agreement among the states, but by an act of the constituent assembly. 2 As Mohit Bhattacharya, in a careful review of the constituent assembly, points out, by the time Ambedkar had presented the draft in November 1948, both the partition between Pakistan and India and the somewhat reluctant and occasionally even coerced integration [End Page 22] of virtually all of the 568 princely states had already occurred. 3 Therefore, bargaining conditions between relatively sovereign units, crucial to Riker's view of how and why enduring federations are created, in essence no longer existed.

Thus one may see the formation of democratic federal systems as fitting into a sort of continuum. On one end, closest to the pure model of a largely voluntary bargain, are the relatively autonomous units that "come together" to pool their sovereignty while retaining their individual identities. The United States, Switzerland, and Australia are examples of such states. At the other end of the democratic continuum, we have India, Belgium, and Spain as examples of "holding-together" federalism. And then there is what I call "putting-together" federalism, a heavily coercive effort by a nondemocratic centralizing power to put together a multinational state, some of the components of which had previously been independent states. The USSR was an example of this type of federalism. Since federal systems have been formed for different reasons and to achieve different goals, it is no surprise that their founders created fundamentally different structures. This leads us to our next point.

"Demos-Constraining" vs. "Demos-Enabling"

Earlier, I described American-style federalism as "demos-constraining." In some respects, all democratic federations are more "demos-constraining" than unitary democracies. There are three reasons for this. First, unitary democracies have an open agenda, as Adam Przeworski points out, while in a federal democracy the agenda of the demos is somewhat restricted because many policy areas have been constitutionally assigned to the exclusive competence of the states. 4 Second, even at the center there are two legislative chambers, one (in theory) representing the one person- one vote principle, and the other representing the territorial principle. Third, because jurisdictional disputes are a more difficult and persistent issue in federal than in unitary systems, the judiciary, which is not responsible to the demos, is necessarily more salient and powerful.

Riker sees the demos-constraining aspect of federalism (and the weak politywide political parties normally associated with federalism) as basically good, because it can help protect individual rights from being infringed by the central government's potential for producing populist majorities. 5 But when examined from the point of view of equality and efficacy, both of which are as important to the consolidation of democracy as is liberty, the picture becomes more complicated. The deviation from the one citizen-one vote principle that federalism necessarily implies may be seen as a violation of the principle of equality. Overrepresentation in the upper house, combined with constitutional provisions requiring a supermajority to pass certain kinds of legislation, could, in [End Page 23] certain extreme cases, lead to a situation in which legislators representing less than 10 percent of the electorate are able to thwart the wishes of the vast majority. This raises serious questions for the efficacious and legitimate functioning of democracy. If one were interested only in creating a system that best reflects the demos and that functions as an effective democracy, a case could be made that the democratic values of participation, decentralization, and equality would be better addressed in a unitary system that has decentralized participation than in a federal system. But if a polity has great linguistic diversity, is multinational, and is very large, its chances of being a democracy are much better if it adopts a federal system.

If federal systems were forced to adhere to the Rikerian model, multinational democracies would be faced with a stark choice: If they wished to adopt a federal system to reduce ethnic, religious, or linguistic tensions, they could do so only at the risk of severely constraining majority rule. But if we look at the federal systems that actually exist in the world, we see that not all federal systems are demos-constraining to the same degree. American-style federalism is demos-constraining, and Brazil is the most demos-constraining federation in the world. Yet the German federal system is much more demos-enabling than that of the United States, and India's is even more demos-enabling than Germany's. We can, in fact, construct a continuum, ranging from federal systems that are demos-constraining to those that are demos-enabling. Where a particular federal system lies on this continuum is largely determined by the nature of the party system, which I discuss elsewhere, and by three constitutionally embedded variables: 1) the degree of overrepresentation in the upper chamber; 2) the policy scope of the territorial chamber; and 3) the sorts of policy issues that are off the policy agenda of the demos because they have been allocated to the states or subunits.

1) Overrepresentation in the territorial chamber. I think it is fair to argue that the greater the representation of the less populous states (and therefore the underrepresentation of the more populous states), the greater the demos-constraining potential of the upper house will be. The United States and Brazil follow the same format: In both countries, each state gets the same number of senators. Since Wyoming had a population of 453,000 and California had a population of 30 million in 1990, this meant that one vote for a senator in Wyoming was worth 66 votes in California. In Brazil, the overrepresentation is even more extreme. One vote cast for senator in Roraima has 144 times as much weight as a vote for senator in S~ao Paulo. Moreover, Brazil and Argentina are the only democratic federations in the world that replicate a version of this overrepresentation in the lower house. With perfect proportional representation, S~ao Paulo should have 114 seats. It actually has 70. With perfect representation, Roraima should have one seat. It [End Page 24] actually has eight. The Brazilian Constitution, inspired by the ideology of territorial representation, specifies that no state can have more than 70 seats in the lower house (thereby partially disenfranchising S~ao Paulo) and that no state can have fewer than eight.

Yet the principle of equal representation of each state in the upper house is not democratically necessary and may even prove to be a disincentive to multinational polities that contemplate adopting a federal system. Many democratic federations have quite different formulas for constructing their upper houses. In Germany, the most populous states (or Länder) get six votes in the upper chamber, those of intermediate size get four, and the least populous get three. Austria, Belgium, and India are still closer to the one person-one vote end of the continuum. If multilingual India had followed the U.S. pattern, it would not have been able to do some things that were absolutely crucial for political stability. Between 1962 and 1987, India created six new culturally distinctive states in the northeast, mostly carved out of Assam, a conflict-ridden region bordering Burma and China. If India had followed the U.S. model, these new states, containing barely one percent of India's population, would have had to be given 25 percent of all the votes in the upper chamber. The other Indian states would never have allowed this. Thus something democratically useful--the creation of new states, some of which were demanding independence by violent means--would have been difficult or impossible under the U.S. principle of representing each state equally.

The range of variation among the world's federal democracies can be seen in Table 1 on the following page. This table also illustrates what I said above about most federal democracies choosing not to follow the U.S. model. The United States, along with Brazil and Argentina, which follow the same model, is an outlier on this continuum. The first line measures the degree of inequality of representation according to the Gini index. The values range from 0, which indicates perfect one person-one vote representation, to 1, which indicates that one subunit has all of the votes in the upper house. Belgium's upper house has a Gini-index value of close to 0. Austria's is not much higher. India's is .10. Spain's is .31. The U.S. Gini-index value is almost .50, and Brazil's is .52. This means that the best-represented decile in the United States has 39 percent of the votes in the Senate; in Brazil, the best-represented decile has 43 percent of the votes. In India, it only has 15 percent. The variations are [End Page 25] [Begin Page 27] immense. On this indicator, the United States is clearly on the demos-constraining end of the continuum.

2) Policy scope of the territorial chamber. Now let us turn to our second variable, the competences of the territorially based chamber. My proposition is that the greater the competences of the territorial house, the more the demos--which is represented on a one person-one vote basis in the lower house--is constrained. In the United States, the lower house has a somewhat more important role than the Senate in budget initiation, but if one takes into account the Senate's constitutionally exclusive prerogatives to advise and consent on judicial, ambassadorial, and major administrative appointments, the two houses come fairly close to policy-making parity. On this variable, Brazil has the most demos-constraining system in the world. There is no area that the Brazilian Senate does not vote on, and there are 12 areas where it has exclusive competence, including authority to set limits on how much states can borrow.

As we can see in Table 2 on the following page, however, other federal democracies do not give the upper house as much policy scope as they give the lower house. The German, Spanish, and Indian systems are less demos-constraining, because their upper houses are less unrepresentative and less powerful. While in Brazil senators representing 13 percent of the total electorate can block ordinary legislation (and in the United States, a committee chairman alone can at times block important nominations), in Germany important bills are seldom vetoed by the upper chamber. How can we account for such a difference? First of all, the upper chamber cannot participate in the two most important legislative votes, those for government formation and government termination. This power is the exclusive competence of the lower chamber. Second, the upper chamber can delay, but not veto, bills that do not directly involve the Länder. Third, on the approximately 50 percent of the bills that the upper chamber can theoretically veto because they do relate directly to the Länder, it seldom does so after closed-door reconciliation meetings are held in the joint committee representing both houses.

In Spain, Belgium, India, and Austria, as well as in Germany, only the lower house participates in no-confidence votes. In many countries, the upper house is largely a revisionary chamber, although it has a major role in anything having to do with federal intervention. In Spain, for example, if the government wishes to take action against a regional government that is in contempt of the constitution, the decision must be approved by two-thirds of the upper house. This, in my view, is entirely appropriate.

3) The degree to which policy-making authority is constitutionally allocated to subunits. The third constitutionally embedded variable on [End Page 27] [Begin Page 29] which democratic federations differ greatly is the powers that are given to the demos at the center versus the powers that are constitutionally allocated to the states. The 1988 Brazilian Constitution is so extensively detailed that a great deal of ordinary legislation can be passed only by a supermajority. In Brazil, many specific provisions on state and municipal pensions, state banks (all the states have banks), and the right of states to tax exports were constitutionally embedded. This is extremely demos-constraining. When too many issues are constitutionally embedded, the result is profoundly undemocratic, because these issues cannot be decided by a normal majority. Almost everything of importance in Brazil is constitutionally embedded. In order to change the constitution, 60 percent of the members of both houses (both those present and those absent) must vote in favor of an amendment twice. In a country the size of a continent, with bad transportation, it is hard even to get 60 percent of the legislature to show up.

At the opposite end of the continuum (see Table 3 on the following page), India has a very demos-enabling constitution. At the time of its drafting, its authors were painfully aware that there were more than 15 languages spoken in the country that at least 20 million people could claim as their mother tongue. The boundaries of the states did not correspond with linguistic boundaries. To get the government closer to the people, the framers of the Indian Constitution had to respect the linguistic principle, so they decided (Article 3) that the lower house, by a simple majority vote, could eliminate any state, carve new states out of existing ones, or change their names. That is the sort of provision that a "holding-together" federation can write. In a "states'-rights" federation like the United States, such a provision would be absolutely impossible. But if it had not been possible in India, the failure to realize the "imagined communities" of the country's hundreds of millions of non-Hindi speakers might have led to secession in a number of places.

The U.S. Constitution is even more difficult to amend than the Brazilian Constitution, but it is parsimonious, so the vast majority of legislation can be passed by ordinary majorities. In Spain, the main constraint on the majority at the center derives from the statutes of autonomy, which deal primarily with questions of culture and language. In Germany, many federal programs are administered by the Länder, but lawmaking and policy oversight remain the prerogative of the center.

Constitutionally Symmetrical vs. Asymmetrical

Let us now turn to a final point concerning the U.S. model. The U.S. Constitution, as discussed above, establishes a form of symmetrical federalism, which is bolstered by a certain normative disinclination on [End Page 29] [Begin Page 31] the part of Americans to accept the concept of collective rights. With the exception of Switzerland (where none of the political parties strictly represents any one linguistic or religious group), all of the multinational democracies are constitutionally asymmetrical: In order to hold the multinational polity together, they assign different linguistic, cultural, and legal competences to different states. Under the symmetrical American model, many of the things that are most essential in a multinational context cannot be accomplished. With the possible exception of the special case of Switzerland, all federations that are constitutionally symmetrical--Austria, Germany, Australia, the United States, Argentina, and Brazil--are mononational. India, Belgium, Canada, and Spain are multinational and their federations are all asymmetrical. (The Russian Federation is also asymmetrical, but, constitutionally, it does not yet work as a democratic federation.)

The concept of collective rights is in tension with the traditional American way of thinking about such matters, which is based on individual rights. It is true that a polity cannot be a democracy unless the individual rights of all citizens are enshrined in the constitution and a countrywide system of horizontal and vertical controls is credibly established to support these rights. Whatever rights the national sub-units may possess, they cannot constitutionally or politically violate the rights of individual citizens. The enforcement of individual rights can be an obligation of both the center and the subunits, but the center cannot completely delegate responsibility for the establishment and maintenance of democratic rights and continue to be a democracy. Alexis de Tocqueville is very clear on this point. He admired the robust local associationalism of U.S. democracy but pointed out that the rule of law in the entire polity had to be guaranteed and enforced by the center.

In multinational polities, however, some groups may be able to participate fully as individual citizens only if they acquire, as a group, the right to have schooling, mass media, and religious or even legal structures that correspond to their language and culture. Some of these rights may be described as group-specific collective rights. Many thinkers in the liberal tradition assume that all rights are individual and universal and view any deviation from individualism and universalism with suspicion, but this assumption is open to question.

Let me conclude with four observations, partly drawn from studies of the historical development of democracy, about democratic group-specific rights (to use a term coined by the Canadian political philosopher Will Kymlicka). 6 First, individuals are indeed the primary bearers of rights, and no group rights should violate individual rights in a democratic polity. In democratic multinational federal states, this means that something like a bill of individual rights should be promulgated by the federal center, and any laws and social policies that violate [End Page 31] it must fall outside the constitutionally guaranteed policy scope of the subunits.

Second, while individual rights are universal, it is simply bad history to argue that in actual democracies all rights have been universal. Frequently, the struggle to reconcile the imperatives of political integration with the legitimate imperatives of cultural difference has led countries to award certain minorities group-specific rights, such as those given to French-speaking Quebec in Canada, to cultural councils in Belgium, and to Muslim family courts in India. The key point is that it is the obligation of the democratic state to ensure that no group-specific right violates individual or universal rights.

Third, while individuals are the bearers of rights, there may well be concrete circumstances in which individuals cannot develop or exercise their full rights unless they are active members of a group that struggles for some collective goods common to most of its members. If, for example, the Catalans had not been given certain group-specific rights involving the public status of their own language, I doubt whether as individuals they could have become full democratic citizens of Spain. Similarly, I do not think Kurds will become full democratic citizens of Turkey unless they are granted certain group-specific rights (such as the right to Kurdish newspapers and radio stations in the southeast of Turkey, where Kurds are a majority).

Finally, although such group-specific rights may not be consistent with some nineteenth-century tenets of Anglo-Saxon liberal democracy or with the French idea of citizenship in a nation-state, they are consistent with a polity in which group rights do not violate individual rights, and they permit effective democratic citizenship and loyalty to be extended throughout the polity. They offer, in fact, one of the few ways to craft democracy successfully in the difficult and populous world of multi-national states.

The Limits of the U.S. Model

The U.S. model of federalism, in terms of the analytical categories developed in this article, is "coming-together" in its origin, "constitutionally symmetrical" in its structure, and "demos-constraining" in its political consequences. Despite the prestige of this U.S. model of federalism, it would seem to hold greater historical interest than contemporary attraction for other democracies.

Since the emergence of nation-states on the world stage in the after-math of the French Revolution, no sovereign democratic nation-states have ever "come together" in an enduring federation. Three largely unitary states, however (Belgium, Spain, and India) have constructed "holding-together" federations. In contrast to the United States, these federations are constitutionally asymmetrical and more "demos-enabling" than [End Page 32] "demos-constraining." Should the United Kingdom ever become a federation, it would also be "holding-together" in origin. Since it is extremely unlikely that Wales, Scotland, or Northern Ireland would have the same number of seats as England in the upper chamber of the new federation, or that the new upper chamber of the federation would be nearly equal in power to the lower chamber, the new federation would not be "demos-constraining" as I have defined that term. Finally, it would obviously defeat the purpose of such a new federation if it were constitutionally symmetrical. A U.K. federation, then, would not follow the U.S. model.

The fact that since the French Revolution no fully independent nation-states have come together to pool their sovereignty in a new and more powerful polity constructed in the form of a federation would seem to have implications for the future evolution of the European Union. The European Union is composed of independent states, most of which are nation-states. These states are indeed increasingly becoming "functionally federal." Were there to be a prolonged recession (or a depression), however, and were some EU member states to experience very high unemployment rates in comparison to others, member states could vote to dismantle some of the economic federal structures of the federation that were perceived as being "politically dysfunctional." Unlike most classic federations, such as the United States, the European Union will most likely continue to be marked by the presumption of freedom of exit.

Finally, many of the new federations that could emerge from the currently nondemocratic parts of the world would probably be territorially based, multilingual, and multinational. For the reasons spelled out in this article, very few, if any, such polities would attempt to consolidate democracy using the U.S. model of "coming-together," "demos-constraining," symmetrical federalism. 7

Alfred Stepan, who was Gladstone Professor of Government at Oxford University in 1996-99, is now Wallace Sayre Professor of Government at Columbia University. This essay is based on a lecture he delivered at George Washington University in November 1998 in a series cosponsored by the Mário Soares Foundation, the Luso-American Foundation, and the International Forum for Democratic Studies. A more extensive version of this essay will appear in his book Arguing Comparative Politics, to be published by Oxford University Press in the spring of 2000. He is working with Juan J. Linz on a larger project on federalism that will culminate in a jointly authored book, Federalism, Democracy and Nation.


1. See William H. Riker, "Federalism," in Fred Greenstein and Nelson W. Polsby, eds., Handbook of Political Science (Reading, Mass.: Addison-Wesley, 1975) vol. 5: 93-172.

2. Ambedkar's speech is found in its entirety in India. Constituent Assembly, Debates: Official Report (New Delhi: Manager of Publications, 1951) vol. II, 31-44.

3. Mohit Bhattacharya, "The Mind of the Founding Fathers," in Nirmal Mukarji and Balveer Arora, eds., Federalism in India: Origins and Development (New Delhi: Vikas, 1992), 81-102.

4. Adam Przeworski, "Some Problems in the Study of the Transition to Democracy," in Guillermo O'Donnell, Philippe C. Schmitter, and Laurence Whitehead, eds., Transitions from Authoritarian Rule: Comparative Perspectives (Baltimore: Johns Hopkins University Press, 1986), 47-63.

5. See William H. Riker, Liberalism Against Populism: A Confrontation Between the Theory of Democracy and the Theory of Social Choice (San Francisco: W.H. Freeman, 1982), 247-53. As Riker acknowledges, however, federalism may also give the majority in the subunits the power to limit the freedom of some of the citizens (as the history of the southern United States shows), making it difficult for the federal government to protect them.

6. See Will Kymlicka, Multicultural Citizenship: A Liberal Theory of Minority-Rights (Oxford: Clarendon, 1995). For a powerful argument by a distinguished legal theorist that group rights are often a precondition of individual rights, see Joseph Raz, The Morality of Freedom (Oxford: Clarendon, 1986), 193-216. For a political and philosophically acute discussion of these issues in India, see Rajeev Bhargava, "Secularism, Democracy, and Rights," in Mehdi Arslan and Jannaki Rajan, eds., Communalism in India: Challenge and Response (New Delhi: Manohar, 1994), 61-73.

7. The tentative arguments made in these concluding paragraphs will be developed analytically and empirically in much greater depth in Federalism, Democracy and Nation, a book being written by Juan J. Linz and myself.