FEDERALIST THEORY AND PRACTICE

In this article I offer an understanding of federalism that matches normative arguments with institutional practices, so that the rules and policies recommended by the defense of federalism have something in common with federalism as it exists. To begin with I discuss what is missing from existing understandings of federalism based either on competition and exit or on democratic voice—or, to look at it from the other direction, what is problematic about federalism from the perspective of the leading normative theories that seem to address it. I then develop an alternative account, derived in part from arguments found in The Federalist Papers and elsewhere in the history of liberal thought—an account that helps us make sense of why federations take the shape that they do, with relatively few, relatively large, constitutionally entrenched provinces that are often ethnoculturally distinguished. I proceed from the thought that the real at least might be rational, that if the stable practices of constitutional democracies do not fit with our normative theories that we should at least try to find a new theory; accordingly, the article incorporates both normative and institutionalist perspectives. I suggest that federalism as it exists offers genuine benefits that might well not be attainable with either the more unitary or the more radically decentralized institutions typically called for.

Competition and Exit

Two complementary but analytically separable understandings of federalism dominate the exit-oriented contemporary literature on the subject. I refer to them as competitive federalism and Tiebout sorting.²

The second face of choice of jurisdiction, Tiebout sorting, looks to an ongoing heterogeneity of preferences, and a sorting of the “consumers” according to the packages of laws and policies on offer. Although everyone may prefer lower taxes ceteris paribus, people also have preferences on the spending side of public finance; they want public schools, police protection, public road provision, and so on. They may moreover have varying preferences about how taxes are levied. There is therefore a wide range of tax-and-spending packages that will attract residents. Even more dramatically, preferences vary about morals regulation, educational approach in the public schools, language policy, environmental and labor regulation, and other policy domains less easily quantified and commensurated than taxes and spending.

This approach to thinking about federalism has particular salience in the United States today as it pertains to morals legislation.⁴

Closely related is the image of provinces as providing so-called “laboratories of democracy.” It is, Louis Brandeis held, “one of the happy incidents of the federal system that a single courageous state may, if its citizens choose, serve as a laboratory; and try novel social and economic experiments without risk to the rest of the country;” and such experimentation may point the way toward solutions for social and economic problems that other provinces might wish to emulate (New State Ice Co. v Liebmann 1932, 310). In its most technocratic understanding the metaphor is basically unrelated to federalism. A unitary state can order competing pilot projects to identify the best way to solve an agreed-on problem, and (as scholars of quasi-experiments in public policy know all too well) the policies that provinces just happen to adopt on their own are rarely well designed to allow precise and straightforward technical comparisons. Brandeis meant the idea in a fairly technocratic way; but today it is widely used as a more all-purpose defense of the legitimacy of states pursuing divergent policies in general. Typical is the invocation of Brandeis' dictum by the U.S. Supreme Court in dissents supporting California's authority to allow medical marijuana and New Jersey's authority to forbid the Boy Scouts of America from discriminating against gay boys and men—neither a technical question of economic regulation.

It should immediately be apparent that Tiebout sorting and competitive federalism are tightly linked; indeed they are the same phenomenon, applied to relatively homogenous preferences in one case and relatively heterogeneous preferences in the other. Given that a federation's population is likely to have relatively homogenous preferences on some matters, but not on others, we will see both mechanisms at work in any given federation. Moreover, some policies can be accessed piecemeal without physical relocation and so allow for competition in some domains and sorting in others. Most laws and policies, however, are geographically circumscribed, so it will often be the case that we get Tiebout sorting just to the degree that we do not get competitive federalism, and vice-versa.

Really existing federalism is, however, a long way from an optimal fit with either the best institutions for competitive federalism or the best institutions for Tiebout sorting. To begin with, the states and provinces that make up really existing federalism are, in general, too large. Federations, to put the point in historical perspective, tend to be divided into jurisdictions the size of ancien régime French provinces, not the size of Revolutionary départements or of the principalities of the old Holy Roman Empire. Jurisdictional competition, across at least a significant range of policies, would often proceed better if the jurisdictions were municipalities or metropolitan regions. States the size of New York or California, larger than most countries in the world, are costly for persons or (usually) firms to exit from. In each of the United States, Australia, Canada, Germany, Spain, Mexico, Nigeria, South Africa, and Italy the mean state, province, or regional population lies somewhere in the 2 million to 6 million range; so does the population of Scotland, with Brazilian states and Indonesian provinces being a bit larger than this range. (The outlier among major federations along this dimension is India, with a mean state population of over 30 million.)The median citizen lives in a state or province that is larger still.

To put the point another way, the fact that there are so few states or provinces in any given federations—so few policy providers—means that policy consumers must choose from among a small set of complete policy packages. (Again, the numbers of jurisdictions are more like the tens of old French provinces rather than the hundreds of the Holy Roman Empire.) Sometimes the policy consumers can choose policies on an à la carte basis, as in the case of Delaware incorporation, Nevada divorce, or New York contract law. But for most purposes, persons and firms must choose a complete package of taxes, spending, antidiscrimination laws, marriage laws, labor and environmental regulation, and so on. This package character of the laws and policies impedes both harmonization via competition and the matching of varying preferences with policies via sorting. It leaves states with considerable leeway to retain or adopt laws and policies that do not track residents' preferences regardless of whether these preferences are heterogeneous or homogenous across the federation. Tiebout sorting perhaps provides the right analytical lens through which to view the plurality of cities and towns, and their ability to compete on dimensions of taxes, service-provision, and policies. It casts much less light on federalism as such.

In many modern federations, one of the most conspicuous features of federalism is its association with ethnocultural or, especially, linguistic pluralism. That is, it is routine for one or more of the component provinces to be understood as the province of an ethnic, cultural, or linguistic minority, as in Quebec or Catalonia. In states without an overall ethnocultural majority, such as Switzerland, India, or Iraq, the entire map of provinces may be shaped to track ethnocultural geographic cleavages; in states with such a majority, including Canada and Spain, there may be only one or a few provinces shaped in that way, with the majority divided up among provinces on other grounds. A variety of quasi-federal arrangements, from the partial self-government of the three Celtic kingdoms of Great Britain to the partial self-government of American Indian tribal reservations, Puerto Rico, and Canadian First Nations, also draw boundaries that correspond to ethnocultural cleavages of one kind of another.

This (common) kind of federalism provides a special, and especially stark, case of the policypackaging that impedes either competition or sorting. If living in a province dominated by one's co-linguists trumps other considerations—and it often does—then the whole range of policies and laws other than language law is more or less taken out of competition. When ethnocultural identity is not just emotionally but also legally linked to political membership, mobility, and competition are shut down entirely. Non-Indians cannot join U.S. tribal reservations (they may become residents, but cannot subject themselves to general tribal jurisdiction or become tribe members), and Indians cannot create competing polities that share the same tribal identity. We could describe ethnocultural federalism as a special case of Tiebout sorting, where the preference for living with one's cultural fellows dominates other collective goods. (Compare Kukathas [2003], who combines an understanding of sorting into cultural groups with a normative emphasis on the possibility of exit from them.) But this dominance impedes the sorts of decisions more commonly discussed under the Tiebout rubric.

Mobility and competition accounts of federalism, therefore, fit poorly with a world of federations that have relatively few, relatively large provinces many of which are defined linguistically or ethnoculturally. This is not to say that mobility and competition are intellectually or politically unimportant. Tiebout sorting matters tremendously for migration, politics, and public finance at the local and municipal level. Jurisdictional competition offers a powerful tool for studying interprovincial movement of liquid capital, and has significant effects in those areas of law where residents need not permanently move in order to access the law of another province. Neither form of exit, however, does much work in justifying federalism as we find it. Certainly, federalism may be preferable to unitary government with respect to exit, offering more opportunities for jurisdictional competition or Tiebout sorting than unitary centralized states. But there will typically be little reason, on exit grounds, to prefer federation- size provinces or states to municipalities or to Swiss-style cantons.

Participation and Voice

At least arguments from exit, whether imagining homogenous or heterogeneous policy preferences, necessarily justify some sort of two-level systems of law and government—a lower level of competing jurisdictions that provide the relevant policies, and a higher level unified jurisdiction allowing for free movement among lower level jurisdictions. They therefore look more like real arguments for federalism than do the alternatives sometimes invoked: arguments from voice.

Arguments from voice (e.g., Weinstock 2001, 76–67) maintain that people are more able to govern themselves, more able to exercise collective self-determination, more able to exercise real control over their government when they are governed locally than when they are governed centrally. Here one might invoke ideas from participatory democracy, civic republicanism, or principal-agent theory (lower costs of monitoring local governments). The smaller the unit of government, the greater decision-making authority any one voter has, and the much greater likelihood any one active citizen has of being heard by and persuading enough of his or her fellows to shape policy.

“The smaller the better,” however, is no better a justification for polities the size of California, Uttar Pradesh, or Bavaria when derived from voice than when derived from exit. An emphasis on voice as the justification for preferring state/provincial government over its central/national counterpart also seems to dictate a preference for counties, cantons, and municipalities over states and provinces. A single voter, or a single activist, is only trivially more effective in California than in the United States, in Bavaria than in Germany. Moreover, arguments from voice give no special weight to two-level polities; they cannot be the basis for a “compound republic” as a preferred outcome.

Two Levels or Many

At this stage it might be objected, by the advocates of either exit-based or voice-based justifications for federalism, that neither theory mandates a simple rule of “the smaller the better,” that both are capable of recognizing that some policies are better made at higher levels of government.⁵

When this kind of thought is put in democratic theory terms rather than in economic terms, it is referred to as subsidiarity. This is the doctrine—derived from Catholic thought and prominent in European debates (e.g., Delors et al. 1991; MacCormick 1997; Van Kersbergen and Verbeek 1994, 2004)—that political decisions ought to be made at the most local level possible, so as to encourage participation and policies that better match local needs, desires, and conditions.

Subsidiarity may offer a useful critical language, but it fails as an institutional decision rule. Föllesdal (1998) rightly points to the “obfuscation” on which much of the apparent consensus and enthusiasm about subsidiarity rests. Subsidiarity calls for case-by-case, issue-by-issue determination of how local a level of government can make a particular decision; this is exactly the wrong way to approach jurisdictional questions in general and constitutional-level institutional design in particular. It presumes a fantastic level of competence, knowledge, and disinterestedness on the part of the body that allocates decision-making authority in each case—itself usually one of the contenders for the authority at stake.

In any event, the level of flexibility envisioned by subsidiarity, or by the account of multiple levels of government efficiently providing the geographically appropriate public goods, about which level of government should decide which policy questions also fails to match up with federalism as it exists in the world. Real federalism is marked by a very high level of stability, and usually constitutional rigidity, in the arrangement of states or provinces; and by an entrenchment of two, but only two, levels of decision-making authority. That is, federalism is typically an arrangement of one central government and a number of states or provinces, not of a center, states and provinces, counties, municipalities, metropolitan regions, and so on.⁶

Moreover, federalism is usually marked by a fixed constitutional allocation of competences between one level and the other. Over some issues the center and the provinces may have concurrent jurisdiction, but even that is often constitutionally specified. All of this is entirely overlooked by, for example, Mueller (1996, 81). Arguing for efficient decentralization in public goods provision, he writes that “[t]he levels within a federalist system can be described under the headings of local, city, regional, national, and world governments,” and goes on to add watershed-size polities for environmental regulation. His chapter on “federalism” prominently features policymaking at the level of San Francisco, Chicago, Berkeley, Utica, and Santa Barbara, as if nonfederalist states lacked policy-making urban municipal governments, and as if such governments were constituent parts of the American federation. His model is entirely innocent of the special status for provinces. This is sensible as an account of the efficient local provision of public goods, but not as an account of federalism.

Of course, the constitutional allocation of responsibilities can be and is bargained around. U.S. states, for example, have given up their autonomy over education policy or traffic regulations in exchange for federal education and highway funds (South Dakota v Dole, 1987). More-local levels of government than states or provinces do exist, even if they lack constitutional standing. When there is voter demand for (or efficiency to be had by) moving central or provincial decisions to some more local level, bargains can be struck and responsibilities can be devolved. But insofar as levels of government multiply, insofar as the allocation of responsibilities among these levels is renegotiated and rebargained for on an ongoing basis—that is, just insofar as the system resembles the one envisioned by either arguments from subsidiarity or arguments from the optimal polity size for the provision of each good—the attempt to constitutionally entrench a federal system seems pointless. A formally unitary state is as capable as a formally federal one of creating subordinate jurisdictions, of giving them authority and then taking it away again. Indeed, the constitutional entrenchment of province-level government and of a particular division of authority would seem to only slow down and interfere with this process. It can impede the generation of rational metropolitan-regional level governing institutions when (as in the case of New York City or Berlin) the region crosses state/provincial lines; it can do the same to the possibility of sound environmental policy, when, for example, a watershed crosses such jurisdictional boundaries.⁷

Uniformity and equality

The dominant tendency in contemporary normative theory has been to ignore federalism altogether; even treatments in terms of competition or participation are fairly rare. This is in part because, with some important exceptions (e.g., Goodin 1996), the dominant mood in political philosophy since the early 1970s has been one of disdain for questions of institutional design, a disdain that is only slowly being overcome. But federalism in particular seems an inapposite topic for arguments about what the best or most just policies and laws would be. These often implicitly assume that, once one has determined what justice requires, there can be no interesting argument for allowing it to vary from place to place. This is an old thought; Condorcet (1969 [1811], 274) wrote, in disdainful criticism of Montesquieu, “As truth, reason, justice, the rights of man, the interests of property, of liberty, of security, are in all places the same; we cannot discover why all the provinces of a state, or even all states, should not have the same civil and criminal laws, and the same laws relative to commerce. A good law should be good for all men. A true proposition is true every where.”

If such stark statements are rare in contemporary theory it is more because the thought is taken for granted than because it is rejected, not as much because we lack heirs of Condorcet as because we lack heirs of Montesquieu for them to bother arguing against. Accordingly, federalism is almost entirely absent from the work of John Rawls and from the generation of liberal theory he inspired. In this post-1971 intellectual landscape, it has sometimes been suggested, or casually assumed, that liberalism is synonymous with moral universalism applied to politics. Either particularistic attachments or jurisdictional differentiations have been treated as in obvious tension with liberalism as such. For a state to draw distinctions even between its citizens and others has been identified as treating noncitizens as less than moral equals, and again, therefore, as illiberal. We have often been left with an image of liberalism as applied Kantian moral philosophy, or as applied moral philosophy simpliciter, rather than as a distinct political or constitutional doctrine. If the task of political philosophy is simply to state what the morally best policies are, then (some seem to have thought) there can be nothing normatively interesting to say in favor of a plurality of jurisdictions. A true proposition is, after all, true everywhere.

So federalism seems caught between a rock and a hard place. If we think about federalism primarily in terms of competition or sorting, or participation or voice, there is a puzzling regularity to be accounted for in the ways that existing federalism departs from what these theories prescribe. Provinces are too few, too large, too tightly associated with ethnocultural and linguistic cleavages, and too rigidly fixed. They lack the smallness and flexibility that any of those theories prescribe. So it is unclear what benefits they might offer, to be weighed against the general universalist assumption of equality before the law, of treating likes alike. Why allow laws to vary from one province to another? Whether we approach federalism from the direction of normative theory, trying to understand what even might make it justifiable in order to be able to come to an evaluation, or from an institutionalist or functionalist direction, trying to understand what use federalism might have that would lead to the widespread adoption of these large, rigid, culturally defined provinces, there is some reason not to let matters lie.

DIVIDED PATRIOTISM AND THE SEPARATION OF LOYALTIES

In Federalist nos. 25, 45, and 46, Publius ([1788]2003)⁸

Publius's argument instead depends on the citizenry's natural loyalty and attachment to their states as against the federal center. That is, a prediction about the affective relationship citizens will have to states is built into the account of what will make the constitutional structure work. It is an account of counterbalancing, but crucially different from the kind of counterbalancing involved in interest checking interest, ambition checking ambition. The latter has an overtone of calculation and egoism that is missing from the counterbalancing developed in nos. 25, 45, and 46.

With respect to the threat of military subversion, Publius maintains that a standing army at the center would be less dangerous to republican liberty than the alternative, standing armies in the several states, because

This inclination, this salutary jealousy, will prevent the buildup of any armed forces that exceed the federation's genuine needs for external defense. Such a buildup cannot take place all at once, and the people and the states (electing the House and the Senate, respectively) would certainly interrupt the process. But even if the buildup did take place, the greater attachment of the people to their states than to the federal center would allow the states to defend themselves successfully. Citizens in their armed resistance will be “united and conducted by governments possessing their affections and confidence. … Besides the advantage of being armed, which the Americans possess over the people of almost every other nation, the existence of subordinate governments, to which the people are attached, and by which the militia officers are appointed, forms a barrier against the enterprises of ambition, more insurmountable than any which a simple government of any form can admit of” ([1788]2003, 231).

Much of the argument of these numbers of the Federalist is in a conditional, hypothetical, or subjunctive mood. Madison and Hamilton were at the time united in their belief that it was the state governments that posed the greatest threats to liberty, and in their doubt that the federal government would prove to be a danger. They were certainly not inclined to aggravate any worries about the new government, at the moment when worries could imperil ratification of the new Constitution.

Still, the account appears to be cogent enough. It may have empirical premises that are subject to doubt, not least the supposed inability of the federal army to defeat unified state militias. But the core thought is that authority can be safely vested in the central government in part because, and perhaps just to the degree that, the people are inclined to be loyal and attached to their states rather than to the center. The Federalist Papers are sometimes read as an appeal to Virginians, New Yorkers, and the rest to think of themselves as Americans; they are treated as an early instance of so-called nation-building (Beer 1993). This is a fair reading of, for example, Jay's no. 2, and it sits well with what we know about Hamilton's personal views. But it is not, for the most part, the public stance of Publius.

Loyalty to the states does not only stave off military rule. It is the general protection against the new constitutional order going awry. If the anti-Federalists' worst fears come to pass, Publius insists, all will still be well, because the new government

In an argument of this sort, much that has been puzzling about federalism ceases to be so. “The smaller the better” is poor advice for generating the sort of loyalty that might lead people to take up arms, and worse advice for organizing an armed force that might successfully resist oppression. Even in the more usual conditions short of armed resistance, this theory calls for states of a pretty substantial size. The “first and most natural attachment” of the people is not likely to be to a unit that they enter and exit too easily, one to which their relationship is that of a mere policy consumer in a competitive market.⁹

Of course, the framers of the Philadelphia Constitution did not face a de novo choice about what size states should be or whether they should be equal in legal status to counties or towns. They were faced with the fact of 13 partially independent states, most of whose sizes had been determined long before. (Only three states retained western land claims by the time of the Constitution's drafting.) This is emphasized by Riker and others who see federalism as the outcome of a bargain among exogenously given units. But on the account given here, path dependence need not be viewed as arbitrariness, or even as wholly exogenous. Over a century and a half the English colonies had coalesced into these 13 polities. Particular towns and settlements had not remained the primary political units. Some fragmentation had occurred, but not very much, and plans for further consolidation such as the Albany Plan had failed. The 13 colonies appeared to represent communities to which people were attached, by whatever combination of inertia, interest, and loyalty or sentiment was in play. Although that would be a bad criterion on which to select polities for jurisdictional competition, it was a perfectly good one on which to choose polities that people would be loyal to in the future. That loyalty was indeed in evidence during the constitutional debate. For all the retrospectively vaunted communitarianism of the anti-Federalists, and for all their invocations of Montesquieu on the small size of republics, it was states rather than localities or counties that they saw as primary. Even the Federal Farmer, who was unusually enthusiastic about Switzerland, treated the 13 states as the proper loci of republican government.

Neither Publius nor the anti-Federalists imagined that loyalty could or should vest exclusively in the states. As Pettys (2003) notes, there are suggestions in these numbers of The Federalist Papers of a desirable competition for citizens' loyalty between the states and the center; if the states are badly governed, they may lose their sentimental hold on their people—and thus to lose their ability to rely on popular support against the federal government. Pettys, however, sees this as an unproblematic outcome and reads that back into Publius. The structure and text of the arguments indicate otherwise. If the states are entirely unable to hold the loyalty of their people, then the major freedom-protecting mechanisms of the federal system will be lost. States as bare institutions may not have the authority and weight to resist federal overreach; in this, federalism differs from the separation of powers at the center. Even an unpopular Congress retains significant de jure power and prerogatives with which to impede a president. Thanks to the Supremacy Clause, the same is not true of state governments. To fulfill their role, they must not entirely lose a competition for loyalty—a vertical competition with the center, very unlike the horizontal competition of competitive federalist theory.

No one in 1787 much worried about whether the states would be able to hold onto their citizens' affective attachment; the identities as Virginians, New Yorkers, Rhode Islanders, and so on were too strong. Indeed some state identities had distinctively religious characters, bolstering the sense of attachment to them for some time. But we, at least, can ask the question: what will tend to encourage durable affective loyalty to the provinces? One answer is certainly ethnic or cultural differentiation among the provinces; and so ethnocultural federalism, puzzling for competition accounts of federalism, ceases to be so. Provided that provincial loyalty does not become so intense, or central loyalty so weak, as to encourage secession or civil war, the distinctively strong attachments that come with ethnocultural identifications may be a real resource for federalism rather than a problem for it.

National Loyalty

This linkage of ethnocultural communities with a federalism grounded in loyalties that could counterbalance the central state was most fully developed by Lord Acton, the British liberal historian and theorist of the mid-to-late nineteenth century. A commitment to federalism, decentralization of authority, and institutional counterbalances runs right through Acton's life's work. His ultimately futile struggle against Ultramontanism and the emerging doctrine of papal supremacy and infallibility within the Catholic Church, his highly moralized histories of the growth and development of freedom, complete with a political sociology of what institutional conditions allowed for freedom, and his explicitly normative writings about politics were all animated by a dread of concentrations of central power.

In “On Nationality,” his rejoinder to John Stuart Mill's defense of nationalism in Considerations on Representative Government, Acton justified multinational states in much the same way that Publius defended the compound republic in the arguments discussed earlier. This account of the institutional bases of political resistance to central authority explicitly links decentralization with both ethnocultural pluralism and a large size of the component units. Extent, rigidity, and loyalty are precisely the traits which Acton says one would want in the partially self-governing communities within a state. In pluralistic states the nation may be “a bulwark of self-government. Private rights, which are sacrificed to the unity [of nation and state], are preserved by the union of nations. No power can so efficiently resist the tendencies of centralisation, of corruption, and of absolutism, as that community which is the vastest that can be included in a State, which imposes on its members a consistent similarity of character, interest, and opinion, and which arrests the action of the sovereign by the influence of a divided patriotism” (1985 [1862], 424–25, emphasis added).

Acton compared the presence of different nations in the same state to the separation of church and state, in its ability to prevent “servility” and “promote independence, ” “by forming definite groups of public opinion, and by affording a great source and centre of political sentiments, and of notions of duty not derived from the sovereign will. Liberty provokes diversity, and diversity preserves liberty by supplying the means of organization” (1862, 425).

National diversity is especially well suited to this task. National customs and national sentiments appear to be independent of the state in their origin, and vary from one nation to another. A state that begins to exceed its proper liberal bounds, one that moves “beyond the political sphere which is common to all into the social department which escapes legislation and is ruled by spontaneous laws”—will “provoke a reaction, and finally a remedy.” That intolerance characteristic of “absolutism is sure to find a corrective in the national diversities, which no other force could so efficiently provide. The coexistence of several nations under the same State is a test, as well as the best security, of its freedom” (1862, 425).

A “divided patriotism” ensures that power will not be concentrated in the hands of those of whom, as Publius had it, the people have the least suspicion. The unity of nation and state encourages an unhealthy veneration of the state; it joins together loyalties that are more safely separated.

Now Acton did not, any more than Publius did, simply recommend loyalty to the smaller unit in place of loyalty to the larger. Indeed, like Publius, he considered the larger polity ethically superior in a fundamental sense, in part because of its ability to encompass diversity, and the smaller unit's tendency toward oppressive homogeneity. There are no genuine duties to the nation, only the duties of citizens to their states. “The Swiss are ethnologically either French, Italian, or German; but no nationality has the slightest claim upon them, except the purely political nationality of Switzerland” (1862, 429). Acton held our relationship to our ethnic nation, “a community of affections and instincts infinitely important and powerful in savage life, but pertaining more to the animal than to the civilized man,” to be “merely physical,” whereas “our duties to the political nation are ethical” (1862, 427–28).¹⁰

Here there is no romance of the ethnic nation! Yet these ethical and political duties do not translate into any normative injunction to commit our loyalty to the state, to foreswear the physical, animal, savage attachment to the ethnic group. The division of a state's population into nations, however morally arbitrary nations may be, stymies the tendency toward unity and toward the uncritical valorization of the state. Our loyalty to our nations as against our state, like the loyalty to American states rather than to the new United States, could encourage a healthy level of suspicion and jealousy. This spontaneous separation of loyalties ought to be made use of in institutional design, with a federalism built around the component nationalities. This was, Acton thought, “the theory which represents nationality as an essential, but not a supreme element in determining the forms of the State” (1985 [1862], 424–25, emphasis added). It opposes alike theories that treat nationality as supreme and so demand a state for every nation, and those that treat it as invisible or irrelevant and so leave no barriers against states.

Acton may be the purest exemplar of a theorist of “divided patriotism” and a separation of loyalties, a theory that treats substate groupings as important for the protection of freedom without ascribing any moral weight to them apart from the sentiment of loyalty that happens to attach to them. Moreover, no other major theorist is so explicitly and enthusiastically committed to a federalist project. Acton uses none of Publius's subjunctive-mood distancing from the compound republic; at several points in his work, he described federalism as the most important discovery of constitutional theory and as the key to sustaining democratic and republican forms of government. It is thus odd that, in “Nationality,” he at no point uses the word “federal” or any derivative. Nations may provide the means of organization; but he tells the reader nothing more specific about what those means of organization are, whether they include (or are limited to) a provincial level of government, and whether a state could have some provinces that were nationally defined and others not. He is commonly taken to be discussing, for example, the autonomy of Hungary within the Hapsburg Empire; he does not explain the connection between that and the republican federalism of the United States. When he discusses Switzerland, it is only with reference to the nationalities, and not at all with reference to cantons. Joining “Nationality” to his writings on federalism allows for fair inferences about ethnocultural federalism, but inferences they must remain.

Acton moreover must serve as a kind of warning to liberal federalists. His enthusiasm for federalism as a means, indeed as the signal means for checking modern majoritarian-democratic states, knew almost no bounds. “The true natural check on absolute democracy is the federal system … It is the one immortal tribute of America to political science, for state rights are at the same time the consummation and the guard of democracy” (1985 [1910], 211).

As sometimes happens, the enthusiasm for the means became so complete as to lead the means and the end of human freedom to be conflated. Acton's certainty that “the time has come for the extinction of servitude [slavery],” even combined with his recognition that slavery was at the very foundation of the American southern Confederacy (1985 [1866a], 274) could lead him to no better wisdom on the American Civil War than that slavery was “by one part of the nation … wickedly defended, and by the other as wickedly removed”(1985 [1866a], 277). Even that much condemnation of the Confederacy was grudging, for Acton finally believed that “slavery was not the cause of secession, but the reason of its failure” (1985 [1866a], 278). Far more important, he thought, was the southern attempt, both in the act of secession and in the constitution framed for the Confederacy, to insist on federalism and minority rights, against a northern Union (as he thought) committed to untrammeled majoritarianism. It was the Confederacy rather than the Union that carried forward the belief that “government is bound within certain limits and by certain laws,” the “repudiation of the doctrine that the minority can enforce no rights, and the majority can commit no wrongs.” The CSA's constitution was such as to “justify me, I think, in saying that history can show no instance of so great an effort made by Republicans to remedy the faults of that form of government” (1985 [1866a], 277). If only the Confederacy had freed the slaves (if only!), “they would have supplied the advocates of freedom hereafter with a peerless model” (1985 [1866a], 278). In the actual event, as Acton wrote to Robert E. Lee in a fawning postwar letter, “I mourn for the stake which was lost at Richmond more deeply than I rejoice over that which was saved at Waterloo” (1985 [1866b], 363).

Of Loyalties and Interests

It is a fixed point of the pluralist school of American political science, and the interest-group liberalism that builds on it, that a liberal civil society depends on identities and allegiances cross-cutting rather than on stacking. The separation of loyalties account in Publius and Acton presented here is in direct conflict with this aim. If the identity “Quebecois” is just another interest-group identity, cross-cut by, similar in kind to, and no more important than economic, occupational, hobbyist, or other memberships, then the identity “Quebecois” cannot do the work envisioned here. For an institution and the loyalty it engenders to really provide a counterweight to the central state, it must not be a membership that is easily traded off against other interests—just as citizenship in the general state is not. Acton seems to suggest that the plurality of churches and the plurality of nations within a state are different from the run-of-the-mill plurality of interests. One's church and one's nation may be strong enough, important enough to rival one's state as an object of loyalty, but ordinary cross-cutting associational memberships will presumably not be. Indeed, that is the claimed merit of cross-cutting cleavages: that no single cleavage is felt too strongly or allowed to structure too much of social life; “multiple and politically inconsistent affiliations, loyalties, and stimuli reduce the emotion and aggressiveness involved in political choice” (Lipset 1960, 88). Insofar as that is true, the cleavage won't do the desired work of creating a rival to the central state for citizens' loyalty and attachments. As Margaret Canovan (1997, 72–75) memorably puts it, ordinary self-interest cannot offer the “battery,” the stable long-term “reservoir of power that can slumber for decades and still be available for rapid mobilization” (1997, 73) that perceived nationhood can. Canovan recognizes that collective power and mobilization can take many forms, but maintains—plausibly—that national sentiment and loyalty is a particularly effective battery, whereas interest as such tends not to be.

In this light I suggest that ethnocultural and linguistic cleavages stacked with provincial ones are particularly effective at separating loyalties—and speculate that American federalism today may not be very effective at all. Cleavages that cut across states are the rule; and if there are cleavages that tend not to be cross-cut they are not state cleavages but rather race, education/class, and religious-regional culture. No more than a handful of states might have the kind of distinctiveness and popular loyalty as states that could actually generate a divided patriotism—perhaps Texas, Utah, Hawaii, and Louisiana, plus Puerto Rico. There are broad regional and cultural cleavages in the United States that get entangled with federalism, and that can generate distrust of the center when the center seems to line up on one side of the cleavage. These cultural cleavages are perhaps successors to state-level loyalties, but they should not be conflated with them. The cultural cleavages are both larger and smaller than states—they group together southern states, or coastal states, or rural states, while also dividing the cultural “border states” such as Ohio, Pennsylvania, Missouri, Indiana, and Washington. Today we rarely expect, and rarely see, sentiments of state-level loyalty or identity as such doing much work in American politics. The United States in this sense remains a situation classically described and analyzed by Calhoun (1851 [1992]). The real social cleavages do not track those implicitly or explicitly assumed by the state-loyalties argument of The Federalist Papers, and the governing structure may not have any particular merit in managing the existing conflicts. It is all too appropriate that Calhoun in particular offered the diagnosis, because for him as for us the cleavage between the white South (and its cultural offshoots in the Midwest) and the Northeast (and its cultural offshoots in the west) is of central interest, perhaps rivaled only by the white–black cleavage that Calhoun so carefully and conspicuously skirted.

The centrality of the North–South and white–black cleavages to American federalism has contributed to the general neglect or disdain for federalism as a topic in political theory. Insofar as an image of federalism has lurked in the background of mainstream Anglo-American normative social theory, it is an image decisively shaped by the era of Brown v Board, southern “states' rights” defenses of Jim Crow, and the Warren Court's enforcement of liberal rights against moralistic, paternalistic, or (especially) racist state action and legislation. I think that among American academics of the era, Riker (1964, 155) was unusual only in his extreme bluntness for holding that “if in the United States one approves of Southern white racists, one should approve of American federalism [… If] one disapproves of racism, one should disapprove of federalism.”

The conservative American federalist Michael Greve (1999, 8) has sought to make a federalist virtue out of the relative decline of state loyalties, understood in a jurisdictional-competition way. “Choice and competition have always been among federalism's chief attractions. But those virtues are now much more attractive than at any time in American history. The modern trends that … seem to weaken the case for federalism—the citizens' mobility, their lack of attachment to their home states—greatly strengthen its appeal. Highly mobile and prosperous citizens will be particularly inclined to vote with their feet and to choose among competing jurisdictions. They will particularly appreciate that option, … when (and because) they are not terribly sentimental about their place of birth or residence.” By now it should be clear what is so strange about this. The thought that competitive federalism was “always” the underlying theory of American federalism sits oddly with the acknowledgment that such competition was until recently greatly impeded by local attachments; and there is no explanation as to why institutions developed in an era of such attachments happen to be the right ones for a new era of competition. Moreover, Greve does not, and cannot, explain where the political energy will come from to fight for the decentralizing reform agenda necessary to unlock federalism's competitive potential. That lack of sentimental attachment may have drained the system of a kind of political will and enthusiasm that would be necessary for his policy proposals to become plausible—and, of course, if the attachment were somehow restored, then by his own admission the policy proposals would lose some of their attraction.

By contrast, in Canada, Spain, and India, to name a few, one or more ethnocultural provinces command significant popular loyalty; Catholic Bavaria does in Germany as well. Those federal systems are thereby stabilized; the sentimental attachment that undermines jurisdictional competition keeps in place a much more robust counterbalance against centralization.

Counterbalancing and Opposition

Daniel Weinstock (2001, 76) identifies, as one of the main types of normative argument for federalism, “the liberty argument … which holds that every level of government is a threat to individual liberty, and thus sees the proliferation of levels of government and the counterweights so created as favouring liberty.” He is certainly right to single that theory out. There is, however, a nonsequitur in the argument as it stands. Political bodies do not automatically counterbalance, and political elites are far from averse to cartelization.

One of federalism's potential virtues in a democratic system is the capacity of the provinces to be oppositional. For example, the opposition party at the center is likely to be the majority party in one or more provinces. This provides the party with both the discipline associated with responsibility and the opportunity to prove itself capable of governing; it can brake the spiral of opposition parties that become more extreme as they shrink and shrink as they become more extreme. It often provides a training ground for future leaders of the party at the center. This virtue, however, depends on there being provincial citizens who are willing to resist the incentives the center can provide, and able to monitor provincial officials to keep them responsive to local electorates. Without a separation of loyalties, federalism may well become simply a vice in a political system. Provincial governments can easily be captured by the center. The central government's various spending powers, if nothing else, often give it the wherewithal to bargain around the constitutional allocation of authorities, to leave the provinces financially dependent on it, and to lavish or withhold funds as partisan rewards or punishments. The less de facto autonomy the provinces have, the fewer decisions they make that have significance in voters' lives, the easier it is for voters to disregard them. Such disregard in turn means less monitoring of them, lower turnout in provincial elections, and so an even-greater capacity for the provincial governments to be captured by the center and its majority party. Many democracies have some level of government that suffers this fate of capture, clientelism, irrelevance, and invisibility all reinforcing one another—in much of the United States, counties are like this—and it's far from unheardof for provinces to go this route. Mexico during the era of PRI dominance might serve as a prominent example. Although Russia never became a fully fledged democratic federation, the federal aspects of its system were clearly vulnerable to the central government's powers of purse and patronage; Moscow was able to buy and bully away what might otherwise have been institutions of resistance and counterbalance.

Affective loyalty to the provinces may not be able to arrest these processes entirely. It does, however, seem very likely to slow them. The sort of lively interest in state politics envisioned by Publius, or (even moreso) the intense interest in provincial politics in Quebec, stave off domination of the provinces by the center and its majority party. Without such loyalty, self-interested voters would have little reason to refuse the financial carrots the center could offer for following its line. They might come to view provinces like they view wards in a city characterized by machine politics—the unit to which services are provided or withheld, depending on the level of obedience.

To put it another way: provincial governments can serve as institutional focal points around which those who share identities or ideologies can coalesce; they can, as Acton had it, supply “the means of organization” (Levy 2003b, 2003c, 2004) This presumes, however, some association between the population of the province and the group seeking to organize. If the population of one province is pretty much like the population of any other, if there are no groups that could sensibly coalesce around their governments, then matters look quite different. Provincial officeholders and potential officeholders will have an interest in perpetuating the local institutions; but the provinces might serve no function other than providing employment and resources to a small class of local political elites. Rather than federalism in any interesting sense, a polity might be left with a vestigial layer of government that merely provides sinecures, patronage possibilities, and some possible rents to capture.

A sufficiently strong provincial concentration of interest might serve as well as identities or ideologies here. If the interests don't crystallize into something like an identity or ideology, however, it seems unlikely to suffice. If one province is overwhelmingly more urban and commercial (or more rural and agricultural, or richer, or poorer, etc.) than the rest of the state, then its voters will likely have a distinctive voting pattern. They will vote for different parties, or different branches of the same parties, than their fellow citizens elsewhere. This does not necessarily translate into support for federalism, or sufficient interest in provincial governments to sustain federalism; nor does it necessarily translate into support for any institution that will be robust enough or sticky enough to resist centralization. At best it translates into a geographic pattern of partisan support; at worst, to an invitation to the center to alter the balance of interests with selective subsidies.

Ethnocultural Provinces and Symmetry

On the other hand, it perhaps need not be the case that all provinces in a successful federation can successfully command loyalty against the center. If there are either structural or constitutional rules that keep all provinces on an equal footing, or a general sentiment that keeps the loyalty-inspiring province from being able to bargain for privileges that are radically different from those of other provinces, then a few such provinces, or even one, might serve as an anchor of the whole system.

A generally federal system in general might be able to check undesirable kinds of centralization in ways that one province, no matter how loyal its members, could not. That one province with deeply loyal members, however, may be able to bolster the system as a whole. In a hypothetical Canadian state divided simply into Quebec and the rest of Canada (“ROC,” as the tongue-in-cheek acronym has it), the power imbalance between the two would be too great for Quebec to be able to sustain its autonomy. This is borne out by Hale (2004, 166), who finds that ethnofederalism is associated with state collapse when it contains “a core ethnic region—a single ethnic federal region that enjoys dramatic superiority in population,” such as Russia in the Soviet Union; among other reasons this is because such arrangements “reduce the capacity of central governments to credibly commit to the security of ethnic minority nations.” The core ethnic region is so out of balance with minority provinces that the latter can't effectively protect themselves. By contrast, in a generally federal system, Quebec (or, e.g., Catalonia) is buffered by the general division of the majority group into a number of provinces. In turn the political and constitutional efforts of the minority province can then go into bolstering decentralization on a system-wide basis—or if, as sometimes happens in Canada, one province negotiates for additional authority on its own, it pulls along the other provinces whose voters and leaders dislike the asymmetry. Although those who are loyal to Quebec insist that it is not a mere province like other Canadian provinces, the reluctance of voters in the ROC to see the asymmetry between Quebec and the other provinces grow beyond a certain point aids the cause of Quebecois autonomy in the federation. In short, the intense loyalty of voters in one province—say an ethnocultural province—can protect the whole system of provincial autonomy, which in turn protects the ethnocultural province better than an asymmetrically autonomous province could protect itself against a unified majority.

This means that the intensely loyalty-inspiring provinces, the Actonian ethnocultural provinces, should not be understood as belonging to some wholly different order or normative category from the other provinces in a federal system, and neither should they be understood as wholly anomalous. They may be important factors in keeping the system as a whole federal. They can only serve this function if there are some political forces keeping the treatment of all provinces roughly symmetrical. Whereas such symmetry may be unpalatable to the ethnocultural minority, in the long run it is much better off facing a federated majority than a unified one. Moreover, because the theory places no inherent moral weight on ethnocultural self-government, it denies that the majority has any grounds for complaint in having its group broken up into a number of provinces.

“ODIOUS IN THEMSELVES:” UNIFORMITY AND PRIVILEGES

The argument for federalism based in divided patriotism and the separation of loyalties is a subset of a more general type of argument common to liberal thought about institutional design in the eighteenth and nineteenth centuries. Legal distinctions, inequalities of citizenship and privileges of various sorts, could become the basis for loyalties other than loyalty to the central state, and thereby provide both an emotional focus and an institutional support for resistance to the state. According to this style of argument, this separation of loyalties was a benefit of those institutional designs that encouraged it, not a cost; and it was a benefit even notwithstanding the moral arbitrariness of the legal distinctions.

Montesquieu maintained that intermediate bodies of various sorts were necessary for the protection of freedom in a complex society. He offered an extended defense of a moderate monarchy in a large commercial state—a monarchy that was moderate both insofar as and because it respected the rights and privileges of such corps intermédiaires. The defining observable difference between monarchy and despotism was the former's respect for, and the latter's suppression of, cities, provinces, religious bodies, and the nobility. This is not Montesquieu's more-famous-to-Americans account of federation among republics. It is a defense of provincial and urban autonomy, yes, but as just two pieces of a complicated patchwork of jurisdictions, authorities, ranks, and estates. These serve to hem in the royal executive power. There is a first-order moral interest in preventing religious intolerance, torture and cruelty, and violations of the rule of law. But this generates a very strong second-order interest in, for example, protecting the prerogatives of the nobility and the aristocratic-judicial parlements against executive interference. The interests and loyalties of such aristocrats are well aligned with the need to keep royal power in check by law. “The nobility should be hereditary. In the first place, it is so by its nature; and, besides, it must have a great interest in preserving its prerogatives, odious in themselves, and which, in a free state, must always be endangered” (1989 [1748], 160–61; emphasis added).

Destutt de Tracy (1969 [1811]), in his rationalist, libertarian, republican treatise on political theory, Commentary and Review of Montesquieu's Spirit of the Laws, responded with incredulity some 60 years later. “It would seem a more natural conclusion, we should think, that being odious they should rather be abolished.” It is indeed a more natural conclusion, and has in the course of history proved an almost irresistible one. But Montesquieu understood the conclusion's force; it was why the privileges must, in a free state, “always be endangered.” He thought, however, that that logic of uniformity was a danger to liberty—both directly, because uniformity was likely to be coercively enforced and diversity stamped out, and indirectly, because uniformity tended to weaken the loyalties and attachments that could perpetuate the intermediate bodies that guarded the fundamental laws against despotic rulers or assemblies. It is the latter argument that is at stake in the “odious” passage. The variety of local customs and laws at risk from uniformity are not inherently odious, and he defended this legal diversity and pluralism in the chapter Condorcet complained about with his comment that “A good law should be good for all men.” But the privileges of a hereditary nobility are odious, according to Montesquieu, as are special ecclesiastical rights. This odiousness, the very fact that the privileges can be defended only with appeals to the traditional constitution and the limits it places on royal power, ensures that a powerful constituency will be committed to such limits.

More than a century after The Sprit of the Laws, Alexis de Tocqueville expanded this thought into a major pillar of his comparison of ancien regime and nineteenth-century France. In The Old Regime and the Revolution, Tocqueville (1998 [1851]) was centrally concerned with the effects of homogenization, centralization, and an egalitarian ethos on freedom. After the Bourbon centralization was underway but before the Revolution and Napoleon brought it to its endpoint, “some institutions …, some ancient customs, old mores, even abuses” stood in its way. They provided both psychological impetus and organizational focus to the resistance to absolutism. They “preserved the spirit of resistance in the soul of many individuals [… ] and inclined them to stiffen their necks against abuses of authority,” and thus had desirable systemic effects even when they were based on unjust privileges (1998 [1851], 171–72).

The privileges of the parlements, for instance, Tocqueville agreed to be odious. Yet their insistence on pronouncing on the constitutionality of royal edicts, their resistance to such innovations as administering the provinces by centrally appointed intendants rather than locally elected Estates, and their opposition to tax reform (including the taxation of the nobility) without the consent of the national Estates, “sometimes served as the guardian of men's liberties: it was a great evil which limited a greater one …. The parlements were doubtless more interested in themselves than in the public good, but it must be recognized that they always showed themselves bold in defense of their own independence and their own honor, and they communicated their spirit to all who came near them.” Tocqueville celebrates the resistance of the jurists of the Parisian parlement when that body was dissolved in 1770. Not only were the magistrates themselves unanimous in their refusal to submit to royal edicts, but also the lawyers who practiced before it and judges of other courts all stood by their side. “I know nothing greater in the history of free nations than what happened on the occasion;” he proclaims. Parlementaire privilege was joined to a legalistic mindset. A socially powerful group was therefore deeply habituated to thinking of procedures, formalism, public hearings, and law—and was motivated to remain committed to legal and constitutional norms even when the crown punished them for it. “All these [legalistic] habits, all these forms, were so many barriers against the arbitrary will of the ruler” (Tocqueville 1998 [1851], 177).¹¹

By the mid-nineteenth century, he thought, such a defiant stance in defense of law against the state was unthinkable. What he calls the “old mores” had been replaced by a generalized servility before bureaucratic officials. Likewise, “[t]he priests, whom we have often since seen so slavishly submissive in things civil to the temporal power, whatever its form, the government's boldest flatterers as long as it pretends to favor the Church a little, were at that time one of the most independent bodies in the nation.” Tocqueville maintains that the clergy as an Estate were among the strongest defenders of civil liberties, procedural rights and protections, and the need for representatives to approve taxation—that is, they aimed to protect liberal rights for all Frenchmen, not only their own privileges. The unjust organized privileges of the pre-Revolutionary Church provided an institution to which the clergy would be loyal against the central state, and this loyalty committed them, like the judicial aristocracy, to the defense of law and constitution. Where the bourgeoisie had surviving privileges, especially where the freedom of cities and towns to govern themselves had not been wholly taken away, its members also showed this independence of spirit, this jealousy of privileges that made them better defenders of freedom than Tocqueville's contemporaries. As Tocqueville had worried, “in abandoning our ancestors' social state and throwing their institutions, ideas, and mores pell-mell behind us, what have we put in their place? We … have destroyed those individual powers which were able singlehandedly to cope with tyranny, but I see that it is the government alone which has inherited all the prerogatives snatched from families, corporations, and individuals; so the sometimes oppressive but often conservative strength of a small number of citizens has been replaced by the weakness of all” (Tocqueville 1969 [1833], 15). In America, Tocqueville thought, habits of association, commitments to religious sects, and to a lesser degree federalism had staved off the French turn to an undifferentiated and servile mass society—though he had little confidence that America would resist the democratic drive toward homogenization, centralization, and bureaucratization forever.

Returning for a moment to Acton, his apparent perversity on slavery ran one step deeper than the discussion of federalism and the Confederate constitution mentioned earlier. He suggested that the commitment to an unjust institution did for the South what Montesquieu and Tocqueville said the commitment to odious privileges had done for pre-Revolutionary France.

BULWARKS AND SEPARATIONS

“Bulwark” is a word that is much thrown around in constitutional and political theory without much thought. A bulwark is a defensive structure; it protects those who hide behind it. The bulwark is not itself the object of concern; it is a quintessential case of something that is only instrumentally valuable. The genus of theories under consideration here, the divided patriotism and jealousy and suspicion justifications of federalism, are bulwark theories. For such theories of federalism or institutional pluralism to make sense, the bulwarks must be capable of defending something other than themselves. This distinguishes the separation of loyalties from, for example, the separation of church and state or market and state that Walzer (1984) had in mind in saying liberalism was characterized by the “art of separation.” The separation of church and state has, as one of its purposes, the protection of religious freedom, the protection of churches' autonomy; it counts as a success if the institution has that result. If federalism is effective at protecting states' autonomy and prerogatives from federal intervention, but has no other effect on legislation or policy, then a bulwark theory fails. (Obviously, the bulwark theory fails if federalism somehow has an actively perverse effect on the center's policies, making them more invasive of private rights than they would otherwise be.) In Acton's words, it is the fact that “private rights” that are “preserved” by multinational federations, and imperiled by unitary nation-states, that justifies the former, not any rights of the nations as such ([1862] 1985, 424).

Merely showing that the provinces themselves sometimes infringe on the freedoms of which they are alleged to be protectors does not show that a bulwark theory fails. They might well both threaten freedom internally and provide some genuine protection for freedom as against the central state, and then the bulwark benefits must be weighed against the costs of local oppression. Indeed, we should expect this to be the normal case. Insofar as the intermediate institutions succeed in generating loyalty, it may well be because they are less internally diverse along some dimension than is the state as a whole. This homogeneity, as Publius insisted, may well lead to less freedom for local minorities. If a bulwark theory is correct, then this local diminution of freedom is not decisive; it must be weighed against the benefits gained by checking the central state. That is not to say that the weighing will always be in favor of the provinces; it is hard to conceive of an improvement in freedom from the central state that can outweigh local jurisdictions maintaining slavery. But it is to say, at least, that not every provincial-level restriction on freedom should be centrally abolished on analogy with slavery.

In all of these bulwark arguments there is an implicit or explicit claim that there is some worrisome social-political tendency toward centralization that is enduring powerful enough to take into account in our institutional design. For Tocqueville, Acton, Constant, and (in a somewhat different way) Montesquieu, this is a tendency toward centralization, homogenization, and coerced uniformity, a tendency toward a social condition defined by an all-powerful unitary central state confronting a mass of equal and relatively powerless individual persons who lacked connections to each other that were not mediated by the state. These theorists all considered this a terribly perilous state of affairs. Such a state might govern in ways that respected individual negative liberties, but it was unlikely to do so for long, and there would be no institutions that could stop it or slow it down when it moved toward illiberalism.

Madison did not embrace this political sociology, or at least did not treat it as a central and live worry for America. (Madison wrote much less about the French Revolution and European affairs than did Adams, Paine, or Jefferson, and it is hard to be too precise about when his political science is specific to America and when it is not.) His concern was not the characteristically nineteenth-century one, albeit adapted from Montesquieu, about the centralized and democratic national state. It was rather the eighteenth-century worry (also from Montesquieu) about the military subversion of republican governments—a military subversion led by and concentrated at the center. Each of these theories identifies a way that a despotic central state could come about without being deliberately chosen; as well a way to diminish this likelihood through citizens' loyalty to some stable intermediate institution, with its privileges and legal distinctiveness, instead of being simply and uncritically loyal to the state.

The credibility of either of these accounts in light of modern social science is beyond the scope of this article. Indeed, neither is necessary for a separation of loyalties justification of federalism. The generalized tendency at stake might, for example, be that of nationalizing states described by Brubaker (1996)—the propensity of states competing with each other, or preempting potential secessionists, to seek to socialize their citizens into viewing the state as a nation, and therefore as a primary object of loyalty. The tendency need not be toward the center accumulating all power in its hands, much less a tendency for it to do so in association with social homogenization. It might be only a tendency for the center to acquire a bit more authority than is optimal, sparking a hope that an overcorrection will approximate the preferable outcome.

If there is no secular tendency toward centralization, then the normative conclusions will not follow. Pettys (2003) analyzes the Federalist's account of state and central competition for the loyalty of the citizenry, emphasizing the distance between that account and contemporary ones based on Tiebout sorting. He neglects, however, the hypothesized secular trend toward centralization, and so treats the account simply as one in which citizens simply allot affection to the states or to the center based on which level performs better. That competition requires that the two levels often have concurrent jurisdiction, so that they can be directly compared. In other words, without a centralization thesis, Pettys plausibly derives a model of competition for loyalty that prescribes relatively weak constraints on federal legislative authority, and that views with equanimity the federal government's gradual victory in the competition.

If the various centralization theories are false, they are false as a matter of social fact; they are not simple moral mistakes. If these accounts or anything like them is true as a matter of social fact—if, for example, there is a tendency toward excessive or unjust ethnocultural homogenization in modern nation-states—then separation of loyalties arguments ought to have a central place in our thought about how to resist the dangers they expose. These arguments will begin rather than end our evaluations, because such centralization is not the only danger in politics, and intermediate groups as well as states can be oppressive. But they will need to be present.

Although these sorts of argument are largely unfamiliar in contemporary political thought, there is one instance of an argument that is similar in kind that has received sustained attention—one that stands normatively opposite to the argument under consideration. This is the theory that sentiments of national unity and solidarity, however unjustifiable in themselves or from a universalist perspective, are instrumentally necessary for the political stability of democratic government, the liberal constitutional order, and/or the welfare state. Here, too we see attention to questions about what institutions will generate what affective sentiments of loyalty, and which sentiments of affective loyalty contribute to normatively desired outcomes. The pedigree of this argument about solidaristic sentiment dates at least to one of the works that Acton was writing against: the chapter on nationality in John Stuart Mill's On Representative Government; its most influential contemporary proponents are Miller (1995) and Canovan (1997). Those who have tried to puzzle out whether a multinational state can have sufficient sources of civic unity, thereby conceding the nationalist position that the relevant question is how to generate or maintain enough unity or loyalty, include Kymlicka (1995). By contrast, Kukathas (2003) and Galston (2004) argue against such an emphasis on unified political-civic loyalties; as Weinstock (1999, 291) puts it, “unity, in and of itself, is a morally ambiguous property.” This article of course will not resolve a century-and-a-half old dispute; it merely aims to draw renewed attention to the separation of loyalties tradition, as Miller and Canovan have drawn renewed attention to the solidaristic one.¹²

Institutional Forms

I have argued throughout that there is a close fit between the virtues of separations of loyalty and the institution of federalism as it exists. This is not true for more-decentralist and more-flexible arrangements of local government—but there might be other institutional forms for which it would be true. Federalism may, however, be the most attractive and plausible such arrangement.

Nonterritorial federalism of the sort vaguely envisioned by the Austro-Marxists Otto Bauer and Karl Renner—a sort of combination of a millet system like that in the Ottoman Empire and guaranteed group representation at the center, elected on a nonterritorial basis—has often appealed to political theorists as a preferable alternative to the sort of federalism discussed here. In my view such nonterritorial federalism has been subjected to fatal objections, indeed has been shown not to be coherent, by Bauböck (2004, 2005). Although some elements of law and regulation can be conducted on a nonterritorial basis, most cannot. Territorially intermingled peoples might each be governed by a religiously specific family law, and voluntary contracts may be entered into according to a chosen body of law that need not be the law of the local region. Much of the essential business of governing in a modern state is not of this character, though. Criminal law, tort law, real property law, emergency and disaster response, environmental law, and zoning and land use regulation must bind all within a given territory to be effective. Most taxation, spending, and economic regulation are similarly bound to territory. A nonterritorial federalism that allowed members of various cultural groups to elect nonterritorial governments that lacked jurisdiction over any of these matters would be a nice piece of symbolism at best, a triviality or a sop to each group's rent-seeking elites at worst. The pathological case of federalism discussed earlier—wherein provincial governments are vestigial sinecures rather than real sites of government and opposition—seems likely to be the standard case for such nonterritorial pseudo-governments.

Moreover, the religious model of the Ottoman case generalizes very poorly to the common case of linguistic diversity. Neighbors may live according to different family legal codes with little effect on one another. There are, however, real pressures toward local linguistic homogenization in an era of mass literacy and democratic government (Levy 2003b), and having the general business of local or regional government conducted in a language is a major influence on whether people go on speaking that language or assimilate to another. That is, in cases of linguistic intermingling we are not even faced with the prospect of having most government business conducted territorially with a nonterritorial, multicultural part of government conducted in parallel. Even the multicultural part can't be conducted nonterritorially. If federalism has been far more common in practice than it has been examined in theory, nonterritorial federalism has been far more favored in theory than it has been seen in practice. Bauböck's and other criticisms indicate that this is not an accident. The authority that can be effectively exercised nonterritorially in a modern and democratic state is too attenuated for this to be likely way to organize separation-of-loyalty-based institutions, even if in the abstract it fits into the category.

More generally, within the universe of bulwarks, of arbitrary distinctions that might generate loyalties and attachments to rival the attachment to the state, federalism has some very attractive features. One of these is the possibility of exit and jurisdictional choice. Unlike a structure of classes defined by birth and blood, and unlike any system that distinguishes the population into self-governing groups by ascriptive identities, in the case of federalism it is legally permissible and possible to exit one intermediate group and join another. It need not be easy or costless; as we have seen, if it were costless then federalism couldn't have any of the advantages of a separation of loyalties. But insofar as it is possible, federalism is preferable to systems in which it is not. Although jurisdictional competition and Tiebout sorting may not be able to justify really existing federalism as such, they do seem to be crucial advantages of federalism over a range of other bulwark-based systems.

Another attractive feature, related to the first, is that the legal distinctions are not so “odious in themselves.” Of course, this may undermine the ability of provinces to serve as bulwarks at all. But insofar as there is free mobility among the provinces and legal equality within them, the legal differences between them lack many of the deeply unjust characteristics of, for example, aristocratic privilege. Moreover, the privileges of the provinces are ordinarily miniature and weakened versions of the privileges of states themselves in a world of sovereign states. If the latter are justifiable, then it seems odd to think of the former as being ruled out as a matter of first principles.

This kind of justification of federalism cannot rest all the way down on an account of the desirable consequences of freedom of exit. The argument under discussion here depends on loyalty. The exit-based argument about federalism and freedom suggests that the jurisdictions that are in competition with one another will be obliged to respect liberty in order to retain their members, and therefore, for example, the provinces or states of a federation will be obliged to remain internally liberal. But the arguments discussed here offer a mechanism for keeping the central state in check, not the smaller jurisdictions. That mechanism depends on the members of the smaller jurisdictions being loyal to them, which suggests that they will not readily avail themselves of the right to exit.

A loyalty or bulwark argument might, however, allow us to see what work competition or sorting (or even voice and self-government) could do in an argument about really existing federalism. Suppose that there is some correct version of the argument that power tends to be centralized, and intermediate bodies that rival the state tend to be suppressed. Suppose too that a fully worked-out understanding of jurisdictional competition, Tiebout sorting, and the moral interest in democratic participation recommended, as a first-best option, a system of many small local jurisdictions (e.g., towns or cantons) within an overall state, with a given set of competences. Given the tendency toward centralization, this first-best option is probably not stably available. The cantons will tend to be weakened more than they should be; policy domains will tend to be taken away from them and given to the center; their jurisdiction or even their existence will be fragile. A separation of loyalties argument would then recommend that these tendencies be counterbalanced with a thumb on the other side of the scale, an institutional design that led the voters to affectively overweight the importance of their local jurisdictions and that created a somewhat-excessive rigidity and entrenchment in the existence and authority of the local polities.

Increasing the size of the polities, aligning them with social cleavages people consider important, and according them more constitutional weight than would be the case in the first-best system might lead to a decent long-term approximation of the ideal. Such changes would not maximize competition, sorting, or voice compared to an ideal baseline; but they might well maximize these compared to a system in which the local jurisdictions constantly gave up ground to the center. The social cleavages and the constitutional weight perhaps operate clearly enough. Size, as we have seen in the discussion of the American case earlier, need not—even large provinces do not automatically generate any sentimental attachment and may not hold onto the attachment they once had. Large provinces are, however, more likely than cantons to be places where even very mobile persons spend much of their lives. They will exit the larger units less often than the smaller. Moreover, the larger provinces may feel more like complete regional economies, and so “communities of shared fate,” in Rawls's phrase. They may have cultural and political-cultural characteristics that are more distinctive, and durably distinctive. And they will certainly have greater institutional and organizational resources with which to provide the focal point for resistance that has been discussed.

Institutional Design and Political Theory

A general summary of these bulwark and separation of loyalty arguments in the history of liberal thought might be that there can be uses, for liberal ends, of nonuniversalistic and nonrationally justifiable attachments, institutions, and formal inequalities. Some will see these as cases in which illiberal means are being put to the service of liberal ends. But this seems to me question-begging. I want to suggest that there is nothing intrinsically illiberal about federalism at all. Federalism as a limit on the power of the central state is a case of illiberal means in the service of liberal ends only if liberalism is identified from the outset with a strict formal universalism, rationalism, and egalitarianism. I mean to weaken the automatic association between liberal principles, universalistic at the level of justification and ultimate morality, and the design of liberal institutions, where there need not be universalism either at the level of jurisdictional shape or at the level of personal sentiments or attachments. Institutional design is not a process of deduction from moral first principles; and if liberalism is to be a political theory rather than simply a restatement of universalistic moral philosophy, it must give institutional design its due.

The kind of argument developed here cannot do the work, in any particular case, of providing a final, conclusive argument in favor of really existing federalism. This is because of the arbitrariness in both the boundaries and the allocation of rights. If we abandon the claim that provinces represent natural communities each of which has a moral right to govern themselves, if we instead treat federalism as a source for institutions and loyalties that can generate needed counterbalancing or resistance to the state, then how to divide the state up into provinces becomes a matter of contingent sociological facts—which units are large enough or connect well enough with people's sentiments to generate the requisite loyalty? (Ex hypothesi, this is not how the citizens themselves view matters; their loyalty would seem incompatible with a full embrace of this contingency.) This will certainly not generate a unique and precise division, and so it does not generate a sufficient answer to the inevitable challenge: why these boundaries and not those? why units of this size and not that size? Yet a constant renegotiation of boundaries and unit size undermines the stability of the provinces on which the argument depends. Constant (1989 [1814]) noted the difficulty of squaring this circle, remarking that the constitutional advantages of a diversity of local laws and jurisdictions would not overcome the apparent strangeness of taking a unified and homogenous state and assigning different persons to different laws. It seems that federalism depends to some substantial degree on the givenness of the provinces; but this givenness is itself imagined or constructed, and there is no reason to think that this will escape notice.

Constitutional entrenchment is often a blunt instrument, a precommitment not to simply follow what will seem like the balance of reasons at some later moment; it always involves a decision not to decide everything anew on a case-by-case basis. It therefore always exists in some tension with what seems like the weight of reasons at the later moment. But this problem seems particularly acute in the case of federalism. The provisions of a bill of rights, for example, are typically meant to vest entrenched legal rights in the holders of the analogous moral rights. The justiceable individual right of freedom of religious belief maps more or less closely onto that same individual's moral right of freedom of religious belief. Constitutional entrenchment of separation of powers provisions isn't like that; it is not judges-qua-persons who have a moral right that legislatures not pass bills of attainder. But there are second-order normative arguments that are directly applicable, e.g. the liberal understanding of the rule of law that holds that persons should be convicted of crimes only when they have violated general and previously-promulgated laws. Similarly, arguments from democratic theory directly bear on the protection of an elected legislature's prerogatives to legislate. Normative theory not only offers reasons for some separation of powers, but for separating particular powers in particular ways.

The relationship of normative theory to the institutions of federalism is third-order at best. The entrenchment of provincial authority has to be seen as self-justifying in a sense; it is entrenched so that it will remain entrenched, and the way that the entrenchment stabilizes the system is by channeling the loyalty of the citizenry in directions that will tend to maintain it. Not only are provinces not themselves the holders of moral rights to which their constitutionally entrenched legal authority might correspond, but also it seems difficult to generate normative reasons for any detailed allocation of particular powers to provinces or to the center. There may be a few exceptions; if the provinces are linguistically differentiated, the argument may dictate provincial control over schools, and if they are religiously differentiated it may require provincial control over family law. Such cases are, of course, particularly hazardous ones for internal freedom. The provincial majority that is a statewide minority may well restrict the freedom of cultural insiders and outsiders alike, in the hope of protecting the identity that it finds intrinsically valuable (Levy 2000). To liberals who find that identity only instrumentally valuable, because of its ability to sustain the federalist structure and check the center, that short-term and local cost must be imprecisely weighed against the hypothesized benefit.

Otherwise, however, the theory doesn't offer a decisive answer to the question of why any given power is allocated to the provinces, in the face of the inequalities among citizens created by provincial diversity. A federalism justified in these terms is an enterprise of approximations and blunt instruments, and such enterprises have a way of giving way when pushed for rational justification of their details. Distinctions among citizens that are odious, or even irksome, in themselves are difficult to maintain in the name of a general procedure or institution that staves off a long-term danger; the “natural conclusion” tends to be concluded. That, after all, is one of the mechanisms of the tendency toward uniformity that called for federalism in the first place.

These somewhat gloomy reflections do not amount to a renunciation of the project of offering a normative-theoretical understanding and justification of federalism. They are predictive rather than analytical: third-order reasons tend to give way in the face of first-order ones. There is no normative reason to endorse that tendency in any general way (unless one is an act-utilitarian). If the reasons adduced for a separation of loyalties are sound, then there is a normative case to be made for the rigid, constitutionalized, large-province, often-ethnocultural federalism that we see in the world—even though that case may be an uneasy and fragile one.