SELF-ENFORCING FEDERALISM

The problem is perhaps best illustrated by constitutional secession clauses. The general view of such clauses is that a constitutional right to secede “would increase the risks of ethnic and factional struggle; reduce the prospects for compromise and deliberation in government; raise dramatically the stakes of day‑to‑day political decisions; introduce irrelevant and illegitimate considerations into these decisions; create dangers of blackmail, strategic behavior, and exploitation; and, most generally, endanger the prospects for long‑terms self‑governance" (Sunstein 1991:634). But if, as Sunstein argues further, a constitutional provision prohibiting secession is best interpreted as a contractual agreement whereby federal units pre‑commit to strategies that preclude secession, what is the mechanism that sets any such precommitment in concrete? What renders such a precommitment credible? Why would anyone believe that a promise not to secede will be honored in the future or that the mere act of uttering or agreeing to the promise changes the likelihood of one’s future actions? Why should we suppose that a constitutional provision threatening punishment of those who attempt secession will be honored? If a decision about secession is itself a response to beliefs about the responses of others who also act out of self-interest -- a belief about the likelihood that secession will be punished or ignored -- then why would a constitutional secession clause influence their self‑interest and the likelihood that they will act in accordance with its terms. Finally, if, as some models of secession assume (see for instance Haimanko et al 2001, LeBreton and Weber 2000), people choose to form, maintain or dissolve a federation on the basis of its ability to resolve economic inefficiencies among otherwise sovereign states and to achieve a more or less `equitable’ distribution among those states of the gains of cooperation, then how does any constitutional provision influence economic calculations or outcomes?

These are not mere rhetorical questions, for here game theory upon which the notion of the incentive compatible institution rests, maintains as one of its fundamental lessons that there usually exists a vast multiplicity (if not an infinity) of alternative equilibria – of alternative sustainable outcomes – in essentially any `reasonably complex’ social process. Indeed, almost any feasible outcome, including those that benefit no one over the status quo, can correspond to an equilibrium. This multiplicity of equilibria generates a number of problems for the participants in a game, the most important one being that absent an initial set of mutually consistent set of beliefs as to what long term strategy each will choose, there is no guarantee that any equilibrium will prevail or that the equilibrium which prevails will be beneficial for any of them. Thus, to ensure a mutually advantageous outcome, or even an outcome that at least one participant finds desirable, beliefs and strategies must be coordinated.

This discussion is not as far removed from the more practical task of enforcing the provisions of a federal constitution. First, it gives theoretical meaning to de Tocqueville’s observation that America’s early political experience helped coordinate it to certain functionally useful values, procedures, and political institutional devices. For example, the reader will search in vain for the words `majority rule’ or their equivalent in the U.S. Constitution. That document is careful to specify the circumstances when something other than a simple majority was to decide an issue, but otherwise `of course the majority would rule’ and there was no need to coordinate society to that rule in the text of the Constitution. More generally, our discussion of coordination suggests the need for a profoundly different conceptualization of a constitution than is offered by the contractual argument. To see what we mean, consider the fact that it probably matters little in the life of the American democracy whether the House of Representatives contains 435 or 345 members, whether states are represented by two or three senators, whether an executive veto override requires a two-thirds or three-fifths vote, whether budgetary legislation must originate in one legislative chamber or another, whether the admission of new states to the union requires a majority or three fifths vote of Congress, or whether the Secretary of State or Defense stands highest in the order of presidential succession. We can also imagine the United States surviving with a different rule for admitting new states, different age requirements for members of Congress, a different rule for ratifying treaties, a different system of presidential impeachment and conviction, or even (within limits) a different procedure for amending the constitution itself.

This is not to say that American history would be the same had other alternatives been chosen, but it is far easier to imagine the Republic in peril if its constitution were wholly silent or explicitly ambiguous on each of these things. The early history of the United States consists of a number of critical junctures because of what its constitution did not say (about, for instance, the Supreme Court’s authority or the right of states to secede) rather than because of what it did say, or because what it did say left the door open for confusion and unintended consequences (about, for instance, electing vice-presidents versus presidents). Because the ongoing political process of any nation almost certainly allows for a abundance of equally `acceptable’ equilibria of rules, what is important is that society’s members be coordinated to the same set of rules so that whenever one or the other comes into play, its precise meaning is not the subject of self-interested and self-serving debate or that when debated, discussion is otherwise limited by what the document says, including the rule for conducting and terminating that discussion.

So, to sum up, our argument here is that a theory of federal design, if not democratic design generally, needs to abandon the view of constitutions as social contracts. Admittedly, it may not be a radical conceptual departure to argue that things like voting rules, amendment rules, and succession rules are points of coordination. And it is probably true that certain provisions of a constitution are best describes in contractual terms – most notably, those Level 1 provisions that attempt to specify the legitimate prerogatives of the federal center and that oftentimes are the product of compromises reached among federal subjects at the inception of a constitutional order. But to treat the entire document as a contract leaves unanswered – indeed, leaves unanswerable – the question of enforcement. And if that is true, then we cannot address the issue of federal stability, which by definition requires constitutional stability. Our suggestion here, then, is that we ought to employ a different conceptualization of constitutions; namely, one that sees such a document as a mechanism of socio-political coordination – as an incentive compatible mechanism that is self-sustaining because it is in no decisive coalition’s interest to unilaterally defect from its provisions.

Voters versus Elites

The second constraint requires that we take a more abstract, even philosophical, view of design. Briefly, if a constitution is an equilibrium selection device, and if the game in which it is imbedded requires coordination to achieve specific outcomes consists, at least in part, of rules and institutions of conscious design, then once again we must confront the problem of enforcement – in this instance, enforcement of the rules of the super-game. It might seem, then, that like those who conceptualize constitutions as contracts, we have merely pushed the problem of enforcement back a stage, to yet another level of institutions that must somehow be enforced. However, we should keep in mind that even without imagining any additional institutional layers, a constitution is necessarily imbedded in a game that consists of society’s general structure – its system of norms, conventions, and those things one might choose to call culture. Here we prefer to assume that many of the components of this super game lie outside the realm of conscious design (although of course attitudes can be changed and preferences shaped by advertisement, propaganda, etc.) so that we can focus on formal rules and the question of whether choices exist that encourage federal stability regardless of culture, etc. Nevertheless, the issue of enforcement remains, and our solution in this instance must be a bit different than before. Specifically, taking all political institutions in their totality – constitutional or otherwise – we want those institutions to not only be incentive compatible, but also fully self-enforcing. That is, we want the full nexus of institutions, imbedded in society’s structure, to yield outcomes such that no one has an incentive to disrupt what exists – or at least not to disrupt it in ways other than what the rules endogenously allow. Thus, we want a nexus of rules such that not only will no one defect unilaterally from acting in accordance with any part of them – if we are allowed the freedom to interpret a constitution as a subset of rules – but that no subset will be changed except in a manner prescribed by the remaining rules in the system. We not only want a system of rules such that when acting in their self-interest, individuals make the appropriate (social welfare maximizing) choice, but also a system such that no subset of individuals will prefer a different procedure given whatever rules exist for making changes, and such that those rules of change are themselves endogenously supported.

A constitution creates a great many roles for individuals, including those of legislator, president, minister, judge, and so on. But all critical roles can be sorted into two categories – voters and elected officials (although at times we might choose to substitute `federal subject’ for `citizen’ and include `appointed official’ in the category of `elected’). And although we appreciate the great abstraction implied by this dichotomy, it is nevertheless adequate for our purposes – which in this case is to specify the relationship that needs to exist between these two classes of actors in a stable federal system.

With regard to the first half of this dichotomy, we can begin by supposing that voters are not concerned directly with the stability or instability, or even with such grand notions as the legitimacy of a constitution or its consistency with other social norms. This is not to say that once an institutional equilibrium is established, citizens in a democracy are impervious to arguments that one action or another on their part or on the part of any politician endangers the constitutional order. Nevertheless, when designing a federal state it is only prudent to proceed from the perspective of the Framers of the US Constitution — with the assumption that `the seeds of faction are sown in the nature of man’ — which we take here to mean that voters are primarily concerned with their myopic self-interest, expressed in simple constructs such as increased services, lower taxes, greater access to the governmental bureaucracy, and so on. In this scheme, then, the sole political act of a voter is to express his or her self-interest by voting for those candidates who promise, either on the basis of campaign platforms or past performance, the greatest immediate reward.

In this scheme, then, the burden for ensuring stability falls on the shoulders of elected political elites — or, more properly, on the game we assume these elites play among themselves and before their electorates. That is,

Principle 1. A system of individual-level incentives designed to ensure federal stability should apply not to individual citizens, but to political elites since it is they, even in a democracy, who coordinate and ‘lead’ a society to one equilibrium or another.

When proceeding in accordance with Principle 1 we are, in effect, trying to satisfy two objectives. First, and quite directly, we want to follow the Anglo-American concept of democratic constitutional design and acknowledge that such documents ought to constrain only the state and its agents, and not the sovereign — the people. In this case, we must proceed under the assumption that the sole institutional variables at our disposal are those that apply to political elites and not to democracy’s citizens. Second, we need to acknowledge that regardless of their particularistic motives, the motive that all politicians share in a democracy is the desire to win elections and that absent any contrary incentive, they would, like their constituents, readily challenge or overstep any constraint on federal bargaining if doing so serves that interest. Abiding once again by another precept of The Federalist – namely that `if men were angels no government would be necessary’ – we want to exclude the idea that we can rest the stability of the state on the fragile foundation of the good intentions of those who would lead it. The search for stability, then, requires that we design a game that motivates officeholders to resist or somehow redirect constituent pressures to overstep the rules.

Principle 1, though, can be divided into two more specific sub-principles. First, note that our earlier discussion of bargaining highlights the fact that the goal of providing incentives for political elites to reach `accommodating solutions’ is especially problematic when speaking of regional elites, whose electoral support is tied to specific geographic constituencies. It is they, after all, who directly respond to `narrowly construed’ electoral pressures. A second Principle, then, is:

Principle 2. Any mechanism intended to establish a Federal constitution as an effective coordination device must give local and regional political elites an incentive to uphold federative constraints even when their constituents prefer otherwise.

Principle 2, though, is only one half of the equation. The other half concerns the incentives of national elites – those who we might associate with the center or at least who can direct the center’s authority to coerce federal subjects via the application of any constitutional supremacy clause. That is,

Principle 3. Any mechanism intended to establish a Federal constitution as an effective coordination device must create political (office-related) rewards for national elites that dissuade them from overstepping their constitutionally prescribed authority and to acquiesce to the legitimate authority of regional governments .

It is sometimes argued, of course, that new democracies need `firm central direction’ and a period of benevolent authoritarian rule before a firm constitutional order is established. It is not unreasonable to believe, in fact, that oftentimes a state’s chief executive or head of state will seek to augment their authority – or at least the authority of the federal center itself – on the basis of noble motives. Too often, though, such a prospect sets in motion a chain reaction of responses in which federal subjects seek to preempt the center by augmenting their authority, which in turn makes the center’s demand for firmer control seem more legitimate and imperative, which in turn ... and so on and so forth. Such is the process observers fear they see in Russia today. The result can either be a federalism that becomes a mere facade for unitary rule or the dissolution of the state itself.

Principles 1 through 3 are, nevertheless, merely restatements of much of what we have already said in this volume – of the necessity to maintain a balance between the center and federal subjects, to keep bargaining within the boundaries of constitutional constraint, and to have the maintenance of this balance and these boundaries in the self-interest of political elites even when their constituents might prefer otherwise. To these Principles, however, we need to add a fourth; namely

Principle 4. Federal stability requires that regional and national political elites maintain some (possibly evolving) consensus over the definitions of `constitutionally prescribed’ and `legitimate authority’.

In summary, then, Principles 1-4 throw the burden of maintaining federal constraints on the shoulders of political elites. These principles stand in sharp contrast, then, to those a pluralist would emphasize: citizen involvement, citizen values, and a civic culture. Moreover, despite their emphasis on elites, these principles also contrast sharply with those promulgated by consociationalists. A consociational view also emphasizes the need for a cooperative consensus among elites, although the net they cast is not limited to political elites, but includes business and intellectual elites as well. However, because it is the state alone that possesses both the authority and power to coerce — by force if necessary — Principles 1 - 4 focus on political elites, and, more specifically, on their electoral incentives. Thus, rather than await a consensus based on vague and varied sources of self-interest — shared values, shared economic interests, shared ideology, or whatever — these principles identify as the target of Level 3 design the relationship of political elites to those who would maintain or preclude their official positions, their constituents.

Principles 1 - 4, though, pose a paradox, since an assumption implicit in them is that electorates cannot be expected to treat Level 2 institutions that describe government structure any differently than they do those that comprise Level 1 which allocates prerogatives to different governments. Constitutions may inspire fondness in people who live under them, but if a group of sufficient size can achieve its preferred revision of the rules, it will do so once the opportunity presents itself. Examples of a mass-based support for the revision of even a wholesale disruption of the constitutional order are well known. For example, while Ukranian voters chose to overwhelmingly support the Union in March 1991, once Ukraine’s political leadership championed independence and expressed its dissatisfaction with the USSR’s economic performance, the electorate quickly fell into line, and awarded an elite-sponsored referendum on independence in December nearly the same margin of victory given to unchallenged communist party officials in previous decades. Similarly, following his rise to power through an internal party coup in 1986, Milosevic immediately used nationalist rhetoric to argue that Serbia was discriminated against within Yugoslavia, both politically and economically. Following the boost to his rhetoric given by intellectuals within the Serbian Academy of Sciences, and aided by the rise of the anti-Serb mobilization on the part of the Croatian elites, the public was readily mobilized to take up arms against their countrymen and to abandon any pretense of civilized conduct despite years of seemingly harmonious relations. The role of elites as entrepreneurs of ethnic conflict in the pursuit of political advantage is, of course, well documented. But even in less extreme cases, we can find the same forces — those forces calling for a wholesale revision of the constitutional order — in operation, as Canada well illustrates with Quebec’s campaign for separatism.

These examples illustrate a common phenomenon — that of elites acting as `political entrepreneurs’, engineering support for themselves by encouraging preferences that allow them to act contrary to Principles 1-4. In the Canadian context, for example, because politicians "have a vested interest in provincial status and power which the several provincial electorates perhaps do not fully share" (Corry 1968, p. 53), we see those politicians attempting to move beyond the preferences of their electorates so as to make federal stability more problematic than what it might otherwise be. In perhaps one of the most divisive instances of political entrepreneurship, Riker (1986) offers the example of the issue of slavery and its role in precipitating the American Civil War. Briefly, prior to 1860, a nascent Republican party saw itself as a permanent minority vis-a-vis the Democratic Party if competition persisted on `traditional’ issues. The issue of slavery, however, divided Democrats between North and South, and thus afforded Republicans a wedge with which to win office. Of course, we would not argue that the Republican Party alone accounts for slavery’s ultimate salience or that their strategy of encouraging that salience violated any explicit Level 1 constraint — although the states of the Confederacy did argue and truly believe that federal interference in their `unique institution’ constituted a violation of states’ rights and agreements implicit in the Union’s founding. But, in addition to being a moral issue, slavery was a redistributive one as well, and Republicans saw in the chance to encourage its increased salience among the electorate, a unique opportunity to wrestle power away from their political opponents.

When elite collusion occurs to preclude political competition on divisive issues as opposed to the deliberate engineering of those issues is, of course, a question that should be central to any understanding of democratic stability. The specific formulation we want to address here concerns the conditions under which elected representatives can be expected to bear the burden of Principles 1 - 4 and resist the temptation to either shape citizen preferences to their own ends or, what is even a graver challenge, to follow those preferences when it suits their purposes regardless of the implications of their actions for political stability.

The problem with sustaining Principles 1-4, absent an institutional mechanism of some sort in addition to a constitution — and, therefore, the problem of sustaining federal constraints in a constitution generally — can be illustrated by a simple game. Consider a polity of three voters or, equivalently, three federal subjects and assume that whatever interactions occur among constituencies on federal matters, occur through their elected representatives. Referring now to the strategic form game in Figure 1, notice that we design the payoffs there so that `abiding’ by the rules is an equilibrium — no player (constituency) has an incentive to defect and `ignore’ whatever rules we might have established to preclude destructive federal bargaining. But notice also that payoffs are arranged so that this equilibrium is not what game theorists call `coalition-proof’. That is, if any two players coordinate their defections, then both can move the situation to a different equilibrium in which each realizes a gain at the expense of the non-defecting player, thereby capturing the notion of a critical subset of players capable of toppling a constitution. Of course, the coordination required for such a move may be difficult or impossible to achieve in unguided mass electorates. But the coordination of political elites and, subsequently, their constituencies, is precisely what a federal constitution hopefully encourages. That is, federal institutions give constituencies, through their elected representatives, the ability to choose outcomes that they might not be able to realize through unilateral action – which was precisely the problem, writ large, that characterized the Soviet federation prior to its collapse. Thus, even if we somehow render abiding by federal constraints an equilibrium in a narrow sense, it need not be an equilibrium if circumstances allow for coordinated action among elected representatives.

Unit 3

abide

ignore

Unit 2

Unit 1

abide

ignore

abide

ignore

abide

1, 1, 1

.9, .5, .9

abide

.9, .9, .5

0, 1.2, 1.2

ignore

.5, .9, .9

1.2, 1.2, 0

ignore

1.2, 0, 1.2

0, 0, 0

Figure 1

Fortunately, the world is not populated only by the likes of a Milosevic, and elites do sometimes play a positive role in maintaining federal stability. In addition to the removal of anti-Semitism from the American political agenda, we can also cite the 1992 Constitutional Referendum in Canada, which, although ultimately unsuccessful, attempted to direct voters to a pro-federal position and which was supported by all provincial prime ministers as well at the national government. Similarly, the gradual evolution of the European Union has been largely elite led, despite evident political risks.^[1] The question, then, is whether we can find a source of such motives in a game like the one portrayed in Figure 1. To see that we can, suppose we change this game as follows: If a representative is `instructed’ to ignore federal bargaining constraints by his constituency, he does so only with probability 1-p, and continues to abide with probability p. Ignoring for the moment the source of such probabilities, consider the strategic form game in Figure 2, which is the appropriate transformation of Figure 1 under this probabilistic assumption:

Unit 3

abide

ignore

Unit 2

Unit 1

abide

ignore

abide

ignore

abide

1, 1, 1

.9+.1p, .5+.5p, .9+.1p

abide

.9+.1p, .9+.1p, .5+.5p

1.8p-.8p²

1.2-p+.8p²,

1.2-p+.8p²

ignore

.5+.5p, .9+.1p, .9+.1p

1.2-p+.8p²,

1.2-p+.8p² 1.8p-.8p²

ignore

1.2-p+.8p²

1.8p-.8p²

1.2-p+.8p²

p(2.4-2.5p+1.1p²)

Figure 2

Notice now that for our transformed game, the strategy triple (abide, abide, abide) for voters remains an equilibrium. But consider the cell (ignore, ignore, abide). In Figure 1 this cell corresponds to an equilibrium and confers an advantage on constituencies 1 and 2. Now, however, notice that if 1.2-p+.8p²is less than 1(if p is greater than or equal to .25), then (ignore, ignore, abide) is no longer an equilibrium and the cooperative equilibrium (abide, abide, abide) is coalition-proof – the constitution has acquired popular legitimacy. Thus, if the likelihood of abiding by federal constraints is sufficiently great for all representatives simultaneously even when their constituents indicate an opposite preference, then somewhat paradoxically, sustaining those constraints is an unavoidable equilibrium among constituents.

This is not to say that each representative will prefer unilaterally to choose a value of p greater than 0. Indeed, if all other representatives have p=0, then the constituents of the defecting (imperfect) representative will, on average, fare less well than all other constituencies. Nevertheless, this game demonstrates the potential advantage of having politicians that only imperfectly represent their constituencies. Specifically, even though constituents prefer to `break the rules’ of restrained federal bargaining, that imperfection in politicians leads constituents nevertheless to act as if the representation they receive is perfect. As simple as it is, then, our example highlights two important questions. First, how do we establish rules — a super game — so that political elites not only find it in their self-interest to act with a high enough p, but also to sustain those rules, whatever they might be? Put differently, how do we find a way to make it advantageous for elected officials to act contrary to the myopic self-interest of their constituents? And second, how do we ensure the legitimacy of those rules among the electorate — rules that lead to an apparent thwarting of their will – where by legitimacy we mean simply that voters are willing to issue or at least accept mandates to their elected representatives to act in accordance with constitutional constraints? In effect, then, we are asking the following question:

Design Question: How, in a democracy, do we render representatives imperfect agents of their principals, their constituents? At the same time, how do we design our institutions so that those principals (voters) will nevertheless reward their agents (elected officials) for being imperfect — reward them somehow in the long run, despite the possibility (probability) that they will send signals to them to act contrary to these motives in the short run?

Desirable Imperfection and a Democratic As If Principle

That we might not want public officials to adhere strictly to the myopic demands of their constituents is, of course, reflected elsewhere in the literature on democratic process and design. Carey (1997: 69), for example, in describing the problems of finding a suitable balance between cohesive and incohesive parties, cites the case of Brazil, whose electoral system “creates strong incentives for legislators to cultivate personalistic loyalties among voters, even when doing so means ignoring their parties’ broader agenda. One result is that presidents cannot count on stable legislative coalitions or negotiate policy ... Another is that legislators’ relentless pursuit of government patronage and pork-barrel spending contributes to rampant political corruption.” On the other hand, if, for the sake of the state’s ostensible governability, we sought to reign in `excessive’ constituent attachments, we somehow gave “central party leaders ... the power to impose discipline on elected politicians, responsiveness to minority groups and local interests is necessarily compromised” (Carey 1997: 69).

At this point, however, we need not address these tradeoffs. Before it is important to note the important difference between what our design question implicitly requires (along with Carey’s concerns) and the usual approach to principal-agent problems that we find in the literature. Whether we look to the Public Choice literature on rent seeking or to the more general theory of principal-agent interactions, the problem considered is how to get agents (e.g., elected representatives, attorneys) to act wholly in the interest of a principal (e.g., voters, clients) — how to structure choice so that for the particular circumstance under consideration, an agent’s self-interest corresponds to the principal’s. Instead, our design question turns this problem on its head and seeks to render elites imperfect agents of their principals. However, for reasons that should become clear shortly, we must do this in a way that is not a simple back-tracking of the problem that the design of perfect agents is intended to solve. Briefly, the source of the problem in the more traditional principal-agent literature is that principals are assumed to be able to monitor their agents’ actions or the circumstances these agents confront only imperfectly, in which case agents can take advantage of these informational shortcomings to realize private gains at the expense of their principals. The most evident political example of this problem is corruption, but it in fact has broader application, including the issues we address here. For example, in deFigueiredo and Weingast’s (1997) model of negotiations between Serbs and Croats over the future of Yugoslavia, citizens can only observe whether negotiations succeed of fail, in which case if negotiations fail (in the interest of those conducting them), the population in general can never really know why: "perhaps it failed because genocidal Croatians sabotaged peace; or perhaps it failed because a treacherous Milosevic sabotaged it" (p.3).

The problem we confront is different than the one the principal-agent literature tries to solve. There, the task is to structure the game agents play so that they act `perfectly’ and sincerely advance the principal’s interest. Here, in contrast, we want politicians to be imperfect agents of voters. Moreover, the type of agent imperfection we want cannot be based on secrecy and imperfect information. We do not want our institutions at the outset to be democratically `flawed’ in their design, or biased in terms of who can learn what about events and processes. We still strive to secure a political structure that satisfies some admittedly idealistic notions of democracy and fairness, and that provides reasonable guarantees of individual rights. Thus, we want each agent's actions to be fully observable and fully accountable. But at the same time, and perhaps paradoxically, we want their `imperfect’ actions to be supported ultimately by those who judge them in the voting booth.

Our difficulty in doing this, of course, is that in searching for a mechanism that accomplishes this objective, we cannot look either to the constitutional provisions we want to protect or to moral (utopian) imperatives. In the first case we would merely end up with an infinite regress of reasoning: If provision A enforces provision B, and if Provision C enforces A, then what enforces C? If enforcement is provided by provision D, then what enforces D, and so on? And in the second case, the difficulty is that we cannot engineer moral or social imperatives – or at least we do not know how to do so. And even if we suppose that stability is achievable through an appropriate elite consensus, then we have merely restated Principle 4 without offering any guidance as to how that consensus is achieved — without identifying the specific institutional forms that will sustain that consensus.

Our search for a way of satisfactorily answering our Design Question while simultaneously satisfying Principles 1‑4, then, must lie elsewhere. And here we suggest that since we are concerned solely with democratic federalism, we should once again begin at the beginning – by examining more closely the fundamental motive of elected officials, which we assume is to win and maintain office. This, of course, is but another way of saying that we cannot escape the paradox of wanting our elected officials to function within a democratic state while being rewarded by their electorates for imperfectly representing them. We can, however, begin to see our way to a resolution of this paradox if we assume, as a starting point, that politicians act as if they care not only for the constituency they currently represent but also about other constituencies, including, but not necessarily limited to, those they hope to represent in the future. Our appeal here to `as if’ and to the admittedly ambiguous `but not necessarily limited to’ is not intended to obscure the solution to federal stability we seek, but is instead designed to leave open the possibility, elaborated in subsequent chapters, that elected elites can be led to serve the interest of a constituency different than the one they currently represent, and that voters can be induced to act as if they approve of such action.

Our appeal here to the notion of as if is somewhat different from its common use in the rational choice paradigm. There, as if refers primarily to a method of investigation. Because such things as motive, preference, and perception cannot be observed directly, and because we must limit testable hypotheses to what we can observe — the choices people make — theories of individual decision making must operate under the assumption that these things are, much like electrons, postulated constructs. We can attempt to infer people’s motives from their actions, but we can never know what those motives `really’ are. Nor can we test hypotheses about these things other than in terms of the individual choices they imply in the context of specific models. If people’s choices are consistent with the predictions derived from one set of postulated motives versus another, then we cannot say for certain that they in fact hold those motives; instead, it is more correct to say that they are acting as if they hold them. Admittedly, this linguistic twist might seem like hair-splitting, but it serves to emphasize the methodological problems associated with discovering true motives. Instead, those who operate within the rational choice and individualist paradigm infer any motives necessary in order to maximize the consistency between observed choices and postulated models. And just as there are no true motives to be discovered, there are no true models — only models that fit our data better than others.^[2] If, for example, a model assumes that politicians maximize votes, then their responses to surveys and questionnaires are irrelevant insofar as the model’s `validity’ is concerned. Instead, if a model employing this assumption predicts the decisions of politicians with sufficient accuracy, the as if principle can be invoked to say that they are acting as if they maximize votes, without implying `true’ motives or requiring that the objects of investigation be consciously aware of the motives that `explain’ their actions.

Here, when addressing the issue of constitutional design, we want to adapt the as if notion to mean that instead of concerning ourselves with whether or not politicians see themselves as representing their legally defined constituencies and regardless of whether voters see themselves as rewarding their representatives for championing their interests, politicians can be led to act as if they represent broader interests and voters can be induced to act as if they support those actions. Conscious motives and `true’ intentions notwithstanding, are irrelevant and we might even prefer that politicians justify their actions in terms of representing narrow constituency interests when they are in fact acting otherwise, and that voters suppose they are choosing on the basis of narrow self-interest when they are in fact sustaining more global concerns.

At least with respect to politicians, it might somehow seem strange, perhaps even utopian, to assume that representatives can be led to act as if they serve a constituency other than the one election law assigns them. We can do this in any number of ways, but an especially important one is to assume that elected officials act with an eye to moving up a ladder of electoral advancement from local to regional to national office – to offices of greater national visibility, prestige and power. In this scheme, then, the objective to be realized through Level 3 (individual incentive systems) design is to mute the reflection of a single constituency’s particularistic demands by somehow giving its legislative representative an interest in the welfare of other constituencies.

This is no mere cosmetic addition to an squabble about federalism’s nature. When seeking an institutional solution to the problem of dysfunctional bargaining, our argument here is that we must consider more than the jurisdictional agreements a constitution might offer in the hope that they provide an economically justifiable and politically saleable allocation of responsibilities (Level 1 constraints) or on those provisions that define the national government’s structure and which try to provide different levels of government with ways to defend their position (Level 2 constraints). Our search for the sources of federal stability must also consider the institutions that make the otherwise incoherent `voice of the people’ coherent for political elites and that transform the short term concerns of the average citizen into long term motives for those who scurry about for their support. This discussion, then, suggests that the provisions – constitutional or otherwise – relevant to federal design are not limited to those commonly labeled `federal’ within a constitution or to the agreements negotiated among federal subjects when forming a federation. They also include electoral provisions that are generally established with other purposes in mind, such as ensuring a `fair’ representation of society’s varied interests or facilitating the reelection of incumbents.

On the surface of things, transforming the popular will into something other than what citizens might consciously intend might seem a perversion of democracy. But we argue that is precisely what an appropriately configured political party system ought to do in a federal democracy. Indeed, our argument here is that the federation’s party system and the institutions that help shape it are the critical components of sustaining a federal constitution. Notice that the game in Figure 2 upon which our discussion is based lacks one critical element – a competing cadre of elites who promise values of p = 0 by promising to reflect faithfully the wishes of the electorate whenever that electorate lays claim to some particularized benefit. To ignore this possibility is to ignore the lesson taught us by our experience with federal failure. Consider, for example, the successive civil wars that accompanied the dissolution of the Socialist’ Federal Republic of Yugoslavia: “In the political milieu of ethno-federalism, where concessions could be attacked by political rivals as betrayal of one’s own (ethnic) nation and where there was no longer any central institution authorized to decide disputed issues, escalating threats, displays of armed might and ultimately the use of force could not be avoided” (Hayden 1992: 25). Similarly, in describing the logic of ethnic conflict Horowitz (1985, 1997, 1991) emphasizes that although there are instances in which elites held more extreme positions than their followers, in country after country, politicians that began as leaders inspired by ideals of pan‑ethnic nationalism, socialism and inter-ethnic compromise, were reduced, against their wishes, by the sheer calculus of electoral competition to the more limited status of the leaders of a particular ethnic or otherwise defined group. And this same story of moderately inclined elites finding themselves outflanked by populist extremists is repeated by Rabushka and Shepsle (1972: 80-6).

It might seem from these examples that resting the stability of a federation on the shoulders of some electoral scheme and the political parties it engenders is folly, and it is not surprising that some scholars conclude that the impetus for a successful federation must come from other things ... from the ideological commitment of charismatic leaders who then transmit their federal commitment to the people: “where there is no paramount ideological commitment to the federal ideal”, the mere creation of federal institutions will not resolve the conflicts which will arise within the federation (Franck 19xx: 174). For example, then, scholars of the European Union frequently cite the “pro‑European idealism of heads of government” as the decisive factor of successful integration (see for example Moravcsik 1991) — an idealism in which political elites overcome and perhaps even ignore their and their constituencies’ short-term interests (hence the label `idealism’). We appreciate, of course, that behind such idealism lies strong material interests, including the pressure exerted by multinational firms and exporters who lobby for larger markets, common European regulation, and a stable monetary policy (Sandholtz and Zysman 1989, Moravcsik 1998). Nevertheless, despite a general acceptance of the concept of pan-European integration, citizens at large play at best a marginal role in determining the scope and the pace of integration (a fact commonly given the catchy label of the EU’s “democratic deficit”). Elites there have enjoyed significant freedom to act, provided only that integration remains a limited, incremental affair. The suggestion here, moreover, is that although members of the European Parliament are elected in ways that leave them unconstrained by voters, if and when those members are required to be more responsive to the demands of their constituents, the Parliament may choose to slow or even derail the pace of integration (Tsebelis 1999).

The European example suggests that an appeal to parties and their promise of political advancement is not a necessary motive for political elites to act in accordance with the game portrayed in Figure 2. However, with its members having been ravaged by two world wars, with its basic structure initiated in the shadow of a Soviet military threat, with its economy confronting a perhaps more permanent economic threat from America and Japan, and having been initially pushed in the direction of integration by the hegemonic authority of the United States, the European Union’s circumstances may be somewhat special and not a basis upon which to plan other federations. Thus, when considering states elsewhere, we would not be surprised to see a sympathetic appraisal for somewhat less-than-democratic suggestions of Rabushka and Shepsle (1972) for ensuring political stability – suggestions that include(1) a denial of independence, such as through a reimposition of colonial rule; (2) extreme decentralization so that few issues are allowed to become national (as in the case of the newly imposed constitution of Bosnia); (3) restrictions on free political competition, e.g., secrecy in government action (but see Sened 1995 for the value of information for the spontaneous development and enforcement of rights); (4) restrictions on the overall scope of government; (5) creation of homogeneous societies, e.g., the dissolution of heterogeneous ones; and (6) the creation of permanent external enemies. This is not to say that Rabushka and Shepsle would seriously regard any one of these proposals as feasible or even truly desirable. They are offered more in the spirit of emphasizing the difficulty of securing political stability in plural societies. The substantive suggestion we offer is that by restricting a politician’s vulnerability to populist challenges and allowing elected officials the luxury of giving the future sufficient weight, an appropriately configured party system plays a critical role in encouraging federal stability.

^[1] For example, when in 2000 Danish voters rejected the euro 53% to 47%, it was a brutal blow for Social Democratic Prime Minister Paul Rasmussen, who had led the country since 1983 (Economist October 2000).

^[2]Admittedly, the as if principle seeks to resolve an additional problem within the rational choice paradigm. Much of that paradigm relies on models that appear to assume that people are able to make complex mathematical calculations ... indeed, that they know all that there is to know of advanced mathematics. If, for instance, we sought to predict the actions of the expert billiard player within the paradigm, we would most likely offer a model of someone who, regardless of education, appears to be wholly familiar with trigonometry and the physics of friction and inelastic bodies, and who holds in their head all the values of sines, cosines and the like for every angle and with great precision. Despite appearances, however, our model would not concern itself with whether the pool player actually knew these things. Instead we would state our hypotheses in the form of asserting that the player’s actions were, in this circumstance only, consistent with such knowledge — that the player was acting as if he knew these things. Whether he actually knew them or whether his actions derived from instinct learned from years of practice would not be our concern (unless we were interested in the process of learning itself); our concern would simply be whether the model predicted his actions (and the actions of other expert players) with sufficient accuracy. For a more extensive discussion of this example and the as if principle in general see Friedman (1953).