*415 Dewey and the pragmatists sought, but failed, to establish a relation between democratic institutions and the identity of the democratic citizen conceived on these lines. [FN464] Dewey, in particular, recognized the fragility of a democracy founded on the distinction between a benign elite of experts and an ignorant mass public held docile by manipulation of the symbols by which it affirmed its identity. [FN465] How would the mass, in its ignorance, recognize the benefits of manipulation, however benign? How could the experts, isolated in their expertise from all the experience of the mass, know that their designs were indeed broadly beneficial rather than merely self‑serving? The way to overcome the distinction, he thought, was to make the public expert by affording the citizens the means to acquire expertise. This democratization of expertise, Dewey thought, would go hand in hand with the creation of a system of government which encourages inquiry into the effects on the developmental capacities of the individual of "every institution of the community when it is recognized that individuality is not originally given but is created under the influences of associated life." [FN466] But in his programmatic writing, Dewey focused almost exclusively on the elaboration of a project of comprehensive educational reform designed to form the citizen experts of the new democracy. [FN467] Of the actual institutions of self‑ government he said little, preferring to exult instead at the prospect of a public of scientist‑poets, enlightened by the reading of good newspapers and enlarged in their sympathy with the multitude by their reading of Walt Whitman. [FN468]
*418 To be sure, the institutions of democratic experimentalism provide no guarantee that overworked, self‑interested citizens will, from the start, understand their individual interest to be inextricably linked with the good of the community, but then neither do our existing institutions. Experimentalism at least holds the possibility of providing the public mirror of mutual private self‑examination needed to make good on Dewey's promise. Its institutions enlarge, perhaps vastly, the circle of expertise by revaluing in the problem‑ solving setting, forms of knowledge that were previously discounted, and thus encouraging the participants to learn still more. They diminish the hold of great symbols by calling attention to and scrutinizing the many separate facets of life that the large symbolic tokens condense to the point of indistinguishability. They do this not by attempting‑‑as though such a thing were possible‑‑to substitute analysis, purged of poetry, for figurative expression. The method, instead, is to use the differentiations of benchmarking, with its metaphoric and analogic play of similarity and difference, to uncover possibilities that deductive analysis overlooks and that great symbols obscure. For persons, connected to one another through the institutions of pragmatist sociability, the result is a politics of continuing exploration of difference through acknowledgment of diversity. Such a pragmatic view of politics and identity is not higher, truer, or more prudent than the familiar romantic one. But it is no way lesser, either; and we invoke it to suggest the prospect that democratic experimentalism, far from betraying our deepest intuitions of the meaning of politics, may provoke a reexamination of them.
F. Constitutional Scope
We began this Article by lamenting the breakdown of the defining features of American constitutional government, using the discomfort of existing constitutional law as an indicator of the extent of the breakdown. Our program thus far has focused on what may appear to be, from the perspective of constitutional law narrowly construed, matters of no great moment: We have proposed new organizing principles to be adopted by Congress and implemented by national agencies, state and local bodies, and the courts. How does such a program respond to constitutional concerns?
*419 As a threshold matter, we must, of course, show that democratic experimentalism is consistent with our constitutional tradition‑‑for we have not proposed supplanting the existing Constitution with a new one. This is not a trivial burden, for our design principles challenge basic assumptions about the Constitution we have. For the most part, we treat states as, in principle, no different from localities in seeming violation of core notions of state sovereignty. We elevate direct democracy to a status equal with, and in some instances superior to, representative democracy, in seeming violation of the Constitution's system of checks and balances. And we authorize the courts no less than the legislature to adopt experimentalist attitudes, in seeming violation of the principle that the courts stand as defenders of the most deeply entrenched values against experimentalist onslaught by the political branches.
In the succeeding Parts‑‑on federalism, separation of powers, and judicial protection of individual rights‑‑we endeavor to show that our design principles do not contradict the best understanding of the Constitution we have; indeed, in each section, we argue that something akin to democratic experimentalism has been nascent in constitutional doctrine for quite some time. We do not claim that the best understanding of the Constitution requires democratic experimentalism. Ours is a program to be adopted by democratic means, not judicial imposition. Nonetheless, we do not merely contend that democratic experimentalism can be made to fit the procrustean bed of constitutional law. At its broadest, our argument is that democratic experimentalism would revitalize the central institutions of American government, and do so in a way that is in harmony with the most attractive features of those institutions.
Resurgent constitutional debate about federalism, understood broadly as the doctrine regarding the proper distribution of authority between the federal government and the states, is both a leading example and a symbol of the disarray of our general understanding of the institutions created and recognized by the Constitution. Some affirmation of a sphere of activity reserved to the discretion of the several states seems necessary to sustain the commitment to check the potential menace of a powerful national government by defending a zone of prior, naturally vital state sovereignty which it cannot invade. But in practice, for half a century, the judiciary itself has criticized every effort to discern the boundaries of this prior and natural zone for capriciously limiting the capacity of the federal government to regulate some matter of national interest. Yet the judiciary has been unable to abandon the quest for a domain of state sovereignty not itself derived from the fugitive distinction between federal and state powers‑‑even as the flows of national and international commerce cut new channels in the landscape of authority. Hence, the Supreme Court oscillates, sometimes rapidly, sometimes slowly, between revisionary revivals of the distinction and weary criticism *420 of the revision. [FN469] Surely, it is a sign of deep disorientation repeatedly to assert a doctrine that is independently unsustainable merely because its negation would jeopardize a larger claim of which it is held to be part.
By contrast, in democratic experimentalism, the states and other subnational jurisdictions have (at least) a Madisonian centrality to public action. [FN470] Their purpose, we saw, is to act as the chief instrument of public problem solving given the manifest limits of central direction. At the limit, the national government supplements and assists the smaller units in this system, not the reverse. But in democratic experimentalism, in contrast to the successive judicial syntheses of the recent past, these subnational jurisdictions do not have natural boundaries to their power. Rather, the fluidity of the divisions of authority among them, and between them and the national government, are necessary to their purpose, not evidence of their irrelevance. For it is only by continually adjusting these boundaries that the jurisdictions can, in fact, be effective problem solvers. This reversal of perspective recasts and renders tractable the problem of federalism. Its central theme, from the vantage point of democratic experimentalism, would no longer be the (impossible) search for immovable boundary stones marking the limits of federal and state power, but rather the definition of general standards for determining the just and effective division of sovereignty with regard to particular public problems.
In this Part, we show that such general standards are inherent in the idea of experimentalism as we have defined it; that they establish potentially far‑ reaching (if self‑imposed) limits on the power of Congress to foist its will on the states; and that these limits include the requirement that Congress allow substantial latitude to subnational jurisdictions to determine for themselves how best to cooperate to realize experimental goals.
It is only a slight exaggeration to say that, but for a quirk of interpretation, the Supreme Court might have established the precedents for an experimentalist doctrine of cooperative federalism in New York v. United States. [FN471] The Court had before it legislation obliging the states to respond to the problem of disposal of low‑level radioactive waste, and authorizing them to form jurisdictions of the dimensions they severally or as regional groups found appropriate. [FN472] The Court invalidated the legislation because of the form of the obligation, [FN473] missing a chance, we will *421 argue, to connect limits on what Congress may order states to do with the articulation of standards for the division of jurisdiction between state and national entities. Criticism of that decision provides the occasion for us to sketch the rudiments of cooperative federalism here. As prologue, to fix ideas, we evoke the confusion of current federalism debate; as afterword to our counterview, we note again, that practice is outrunning preaching, in that crucial elements of a regime of cooperative federalism are anticipated in current legislation.
Finally, although our concerns in this Part are largely doctrinal, we are, in conformity with the architecture of our overall argument regarding the relations between an experimentalist Congress and local governments, agnostic in the end as to whether the judiciary should enforce any of the Constitution's federalism norms. Certainly there is considerable historical support for the claim that Congress is better suited to consider the interests of the states, even after the adoption of the Seventeenth Amendment. [FN474] The argument in this Part is that Congress would act consistently with the best understanding of federalism by fostering democratic experimentalism, and that if the courts decide to enforce federalism norms, they would do well to focus on the concerns we identify below.
A. The Arc of Federalism
Today, again, once‑moribund questions of federalism haunt judicial debate. After the 1985 Supreme Court ruling in Garcia v. San Antonio Metropolitan Transit Authority, [FN475] the notion that the judiciary would impose limits on national legislative authority in favor of state sovereignty appeared to be breathing its last breath. Garcia held that if an affirmative power delegated to Congress by Article I authorizes regulatory competence, the Tenth Amendment poses virtually no bar to exercising the delegated power in derogation of state sovereignty. [FN476] Because Congress enjoyed virtually plenary power under the Commerce Clause, Garcia was understood by the Court itself as abdicating judicial authority to enforce constitutional norms of federalism. [FN477] As the Garcia majority recognized, this left Congress, through the representation of the states in the Senate, as the chief guarantor of state sovereignty. [FN478] The Garcia dissenters, by *422 contrast, believed that the Court had left the fox in charge of the chicken coop. [FN479]
But shortly after Garcia, the Court began to revive judicial protection of state sovereignty. In 1991, in Gregory v. Ashcroft, the Court invoked many of the principles rejected in Garcia in holding that in the absence of utterly unequivocal statutory language, Congress would not be deemed to have intended the Age Discrimination in Employment Act [FN480] to apply to state judges, as that would constitute an infringement of state sovereignty. [FN481] Although Gregory was a statutory rather than constitutional case, [FN482] the writing was on the wall: All of the members of the Garcia majority remaining on the Court dissented in Gregory. [FN483] One year later, in New York v. United States, the Court announced that principles of federalism forbid the national government to "'commandeer' state governments into the service of federal regulatory purposes . . . ." [FN484] Three years after that, in United States v. Lopez, the Court invalidated a federal statute imposing criminal penalties for possession of a handgun in the vicinity of a schoolyard. [FN485] For the first time in over half a century, the Court ruled that Congress had exceeded its authority under the Commerce Clause by attempting to regulate an activity that bore an insufficient connection to interstate commerce. Stressing the states' traditional role in education and punishing crimes of violence, the Court appeared to revive for Commerce Clause purposes the distinction between traditional and nontraditional state functions that the Garcia Court had rejected for Tenth Amendment purposes. [FN486] And in 1997, the Court again invoked principles of federalism to invalidate portions of two popular *423 federal statutes: the Religious Freedom Restoration Act [FN487] and the Brady Handgun Violence Prevention Act. [FN488]
Thus, the pendulum of interpretation cuts its familiar arc. If the recent cases appear to undermine Garcia, it should be recalled that Garcia itself expressly abandoned the states'‑rights‑protective decision in National League of Cities v. Usery, [FN489] which in turn had overruled the narrower interpretation of the Tenth Amendment given in Maryland v. Wirtz. [FN490] The periodic doctrinal reversals mark time.
B. New York v. United States
To illustrate the potential of democratic experimentalism to escape such temporizing, we focus on the question at issue in New York v. United States. In New York, the Supreme Court invalidated a scheme with many of the hallmarks of democratic experimentalism because it held Congress had exceeded its authority in giving impermissibly direct orders to the states. [FN491] In this section, we will argue that the Court in this case was both too solicitous of state sovereignty and not nearly solicitous enough: Too solicitous because it focused on formal indicia of state autonomy and apparent insults to the sovereign dignity of the states; not nearly solicitous enough because it neglected to recognize that the challenged legislation invited the states to act as independent collaborators with the federal government and one another in developing solutions to their mutual problems. In our alternative framework, it is these substantive limitations, if any, that mark the boundaries of permissible government interference with state sovereignty.
The dispute in New York arose out of a crisis caused by the shortage of disposal sites for low‑level radioactive waste. [FN492] Because the nation possessed insufficient facilities for processing the waste, Congress enacted a number of provisions designed to encourage the states to increase their waste‑ processing capacity. [FN493] Based on proposals submitted by the *424 National Governors' Association, the congressional response placed responsibility on each state to develop its own disposal facilities or to join with neighboring states in a regional compact. [FN494]
Congress provided several kinds of incentives for compliance. States or regional compacts that developed adequate disposal capacity would be permitted to surcharge (and eventually exclude) waste generated outside the state or compact [FN495]‑‑a power that the dormant Commerce Clause would otherwise bar the states from exercising. [FN496] Some of the money collected through such surcharges would be transferred to the federal government, and then paid out to the states if they met a series of deadlines for developing disposal capacity. [FN497] Noncomplying states would thus be subject to serious financial penalties for failure to meet the statutory deadlines. [FN498] Finally, the statute provided that a state which neither joined a regional compact nor developed in‑state disposal sites by the 1996 deadline,
upon the request of the generator or owner of [in‑state] waste, shall take title to the waste, be obligated to take possession of the waste, and shall be liable for all damages directly or indirectly incurred by such generator or owner as a consequence of the failure of the State to take possession of the waste. [FN499] Although the Court upheld the other provisions, it invalidated the "take title" provision on the ground that it was the equivalent of an order by the federal government commanding a state to pass legislation, a power the Court believed the federal government lacks. [FN500] The Court stated:
The take title provision appears to be unique. No other federal statute has been cited which offers a state government no option other than that of implementing legislation enacted by Congress. Whether one views the take title provision as lying outside Congress' enumerated powers, or as infringing upon the core of state sovereignty reserved by the Tenth Amendment, the provision is inconsistent with the federal structure of our Government established by the Constitution. [FN501]
Under existing constitutional doctrine, the decisive fact in the above passage is that the state government is given "no option other than" implementing federal legislation. For, as New York itself makes clear, where Congress does give the states alternatives, even onerous ones, it can require *425 that the states implement federal legislation as the price of declining these alternatives. Thus, as alternatives to implementing federal legislation, Congress may offer states the possibility of forfeiting federal funds to which they would be entitled if they did choose implementation, or the (menacing) possibility of federal preemption. [FN502] Even at this general level of analysis, therefore, the requirement that the state be given the option of refusing Congress's orders appears to be little more than a formality. Closer examination confirms that this is so.
Take conditional exercises of the spending power. [FN503] Although theoretically subject to constitutional limits, the power to attach strings to federal grants to the states enables Congress to dictate state policy to a considerable degree. [FN504] In an era of scarce governmental resources, few states can afford to forego federal funds as the price of avoiding implementing federal imperatives. A state that decides to refuse federal funds essentially opts to subsidize the states that accept federal funds, because, of course, the state's refusal does not result in a diminished federal tax burden for its citizens. This is a steep price to pay.
Conditional preemption also exacts a high price. Conditional preemption provides that, if a state does not regulate according to federal standards, its citizens will be subject to direct federal regulation. [FN505] To be sure, the direct federal regulation must be of a kind otherwise in the federal government's power, typically under the rubric of the Commerce Clause. But even after Lopez, this is a minor constraint. Even if Congress is not the only enforcer of federalism norms (as Garcia suggests), it is certainly the principal one. [FN506] In this sense, each additional instance of federal (as opposed to state) regulation shifts the federalism balance away from the states. Conditional preemption thus forces the states to choose between two threats to their sovereignty: They must either accept the indignity of implementing federal regulation or acquiesce in the displacement of their authority by the federal government.
This is not to say that conditional spending and conditional preemption are cost‑free endeavors for the federal government. In the case of the former, the federal government must allocate sufficient funds to make the states a financial offer that they will have difficulty resisting. Similarly, conditional preemption requires the federal government to provide the funding to make credible the threat of regulation by a federal agency should the state decline to implement the federal regulations at *426 issue. Thus, both conditional spending and conditional preemption are effective tools only to the extent that the federal government puts its money where its mandate is. But this need to ante up has not prevented the federal government from using these tools effectively to discipline state behavior. [FN507]
Yet, the New York Court is prepared to accept these methods of discipline as the background conditions for modern federal democracy. It simply asserts that conditional funding and conditional preemption "offer [[ ] the States a legitimate choice rather than issuing an unavoidable command," adding that these mechanisms "have now grown commonplace." [FN508] The Court does not even ask whether, much less argue that, the states' choice is a real one. The Court essentially assumes that the greater power to deny funds or preempt entirely includes the lesser power to fund or preempt conditionally.
The Court likewise notes and appears prepared to tolerate more subtle insults to the states' sovereignty. In arguing on behalf of the anticommandeering principle of New York, Justice O'Connor observes that when the federal government issues directives to the states, state officials must divert time, energy, and resources from state priorities and redirect them towards federal priorities. [FN509] This perturbs states in the setting of their own agendas and blurs the lines of governmental accountability. [FN510] Yet, choices made as a result of such intentional or unintentional derangement of the states' agenda are no more‑‑and no less‑‑an expression of autonomy than choices made under the threat of losing funding or losing regulatory authority entirely. To the extent that the Court's federalism doctrine is not a mere formality, it appears that the Court is at once overly protective of, and callously indifferent to, the states' freedom of action. [FN511]
But if the Supreme Court's anticommandeering principle does not actually protect states' rights to set and execute their own agendas, what is it designed to do? As in much of the Supreme Court's federalism jurisprudence, the wellspring is not the desire to protect particular powers of *427 the states, but rather to respect their historical position with respect to the federal government. Otherwise, it would be hard to make sense of the judicial preoccupation with the dignitary interest of the states. Their prerogative to be shielded against any exercise of federal power that manifestly treats them as completely subordinate units of the national government apparently aims to preserve their status as the legatees of the thirteen original states that ratified the Constitution, acting as independent sovereigns, and reserving for themselves an impregnable residuum of autonomy. [FN512] According to this view, the federal government simply lacks the authority to invade that residual area of state authority no matter how expedient it may be. What looks like protection for the "feelings" of the state, [FN513] is thus really the consequence of treating state autonomy as an inviolate historical principle. This brings us full circle to the dilemma of modern federalism sketched at the outset of this Part: Fidelity to constitutional history requires, in extremis, defense of state sovereignty, yet nothing in the historical understanding provides a guide to substantive understanding of that defense that is workable under modern conditions. [FN514]
Further doctrinal oversights [FN515] and confusions concerning the Articles of Confederation aside, [FN516] all this points to the conclusion that *428 the New York Court focuses on the wrong criterion. The question should not be whether the federal government gives the states a nominal alternative to regulating according to federal standards. Rather, as we argue in the next section of this Part, the question should be whether the federal government has treated the states purely as instruments of its national will, or by contrast, as partners in policy formulation and implementation. The availability of alternatives matters in answering this question, but only when it counts as evidence of the choices that matter.
C. Toward a New Delegation Doctrine
Under existing doctrine, federalism acts largely as a side constraint on legislation of national scope. In contrast, in democratic experimentalism, federalism is an essential ingredient of the national framework. Congress authorizes and helps finance experimental elaboration of programs, and the state and local governments actually do the experimenting. Should such a vision become reality, a revised federalism doctrine‑‑whether judicially enforceable or not‑‑would defend a zone of active and substantive collaboration between the two by imposing a double limit on the power of Congress to delegate authority to the states and other local governments. The first limitation requires that when Congress delegates authority to the states, it must provide some guidelines as to the objectives that the states will pursue. It cannot simply pass the buck by declaring local governments responsible for attending to the well‑being of citizens. Rather, it must announce general, yet limited goals. The warrant for this, we find, in an extension of traditional nondelegation doctrine. The second limitation is that Congress may not provide too much guidance: It must leave some important aspects of policy setting for the states themselves, and do this without surreptitiously introducing preferences for some means over others. We find the warrant for this in a rectification and extension of the commandeering doctrine evoked in New York. It is in the band between these limits that democratic experimentalism thrives.
Nondelegation may seem an odd place to begin a discussion of experimentalist federalism. The traditional nondelegation doctrine prohibits Congress from delegating lawmaking authority to an administrative agency without specifying at least a broad policy objective the agency should pursue. [FN517] For a delegation to be effective, there must be a "declared policy by Congress and its definition of the circumstances in which *429 its command is to be effective." [FN518] In part because of its historical association with heightened review of economic legislation during the Lochner era, the nondelegation doctrine, since 1937, has rarely been invoked when administrative discretion has not affected protected private rights. [FN519] Nonetheless, its basic underpinnings rest on a widely accepted insight. Nondelegation doctrine seeks to serve democratic values by ensuring that Congress itself makes (and thus takes the heat for) important decisions of policy. [FN520] In the setting of democratic experimentalism, this requires that Congress must take seriously the obligation to connect the national and the local by setting priorities in the sequence of reform and innovation, by making use of emerging local results to reframe large problems (without precluding their subsequent redefinition through local adjustment). As the zone of democratic experimentalism is intended to be wide, the nondelegation doctrine would be evoked sparingly, to compel deliberation in Congress where political expediency might work against it. Below, we will see that the helter‑skelter decentralization of welfare reform to the states is an example of the abdication of deliberative responsibility to which this approach would apply. [FN521]
Whether or not New York v. United States was rightly decided, commandeering of state and local policymakers [FN522] is bad for the republic, whether the democracy is representative or experimental. Under what conditions, after all, might Congress prefer to instruct the states to enact legislation containing terms dictated by Congress itself, rather than simply *430 enacting legislation with those terms? Two troubling possibilities come to mind. First, Congress may believe that it lacks the affirmative constitutional authority to legislate directly. Second, in the case of legislative programs requiring funding, Congress may not have the political wherewithal to make the necessary appropriation. Neither justification is satisfactory. The first suggests that Congress is merely circumventing the limits on its affirmative powers, [FN523] the second that Congress abdicates its fiscal responsibilities.
One problem with traditional federalism doctrine, as we suggested above, is that it does not go far enough. A prohibition on outright commandeering may block flagrant circumventions of congressional authority and evasion of responsibility, while also banishing explicit affronts to the idea of state autonomy. But it still tolerates control of the states by conditional federal spending, threats of preemption, and interference with agenda setting. In an experimentalist democracy, dependent on the innovative vitality of autonomous local government for learning, these intrusions would be prohibited by an anticommandeering doctrine as well. In drawing the line defining impermissible commandeering, Congress (or, in the justiciable variant of this theory, the Court) would focus on whether the states have been given a substantial role in shaping the federal policy they are to implement, and, in particular, whether the conditions imposed on their receipt of federal funds unduly limited their freedom to experiment. This would be a substantive standard for federalism, not a formal, historical one.
Notice, too, that federal programs defined by a broad policy goal with the obligation to benchmark do not pose as great a threat to state agenda setting as does a highly detailed directive. One of the principal advantages of performance standards (including broad federal directives to states) over design standards, we saw, is that the former enable the regulated entity (including states) to use simultaneous engineering methods to achieve the stated goal in the manner most consistent with the current, effective disposition of the entity's other operations. [FN524] Thus, under democratic experimentalism, congressional delegation that meets the noncommandeering tests also responds directly (if partially) to the concern about the disruptive effect of federal agendas on state agendas, because the directive is sufficiently open and the agenda sufficiently adaptable to minimize disruption.
The anticommandeering principle, as we would see it practiced, finally, does not require that all discretion be ceded to the states. Congress could often choose, for example, between delegating a matter to an (experimentalist) administrative agency or to local governments, provided that the one collaborate with the other in either case. Our anticommandeering principle only requires that when the federal government *431 does find it attractive to enlist the states directly in its regulatory programs, it does so by offering them the possibility of true cooperation.
For example, in South Dakota v. Dole, the Court upheld a federal statutory provision that denied federal highway funds to states that permitted persons under the age of twenty‑one to buy liquor. [FN525] New York v. United States expressly relies upon Dole to uphold the conditional spending at issue there. [FN526] Assuming, however, that the states cannot generally afford to forego federal highway funds, the provision at issue in Dole works just as effectively to commandeer the state legislative process as did the take‑title provision in New York. In practice, it acts as a directive from Congress to the state legislatures, instructing the latter to set a minimum drinking age of twenty‑one. Under the version of the anticommandeering principle that we propose, this statute should have been rejected by Congress for treating the states as servants rather than as partners of the federal government as surely as if Congress had simply enacted a statute requiring each state to enact a statute setting the drinking age at twenty‑one.
In the experimentalist alternative, Congress would have established a broader goal: highway safety. [FN527] Consider a hypothetical statute requiring each state to reduce its rate of highway fatalities to no greater than a rate determined by reference to the experience of other states or lose a fraction of its highway funding reflecting the degree to which the state falls short of the target. States would be free to set the minimum drinking age below twenty‑one, but only if they found some equally or more effective means of avoiding highway fatalities. Such legislation would encourage learning, as different locales experimented with various solutions and benchmarked the results. The states would, finally, be true laboratories of democracy because many eyes would be turned to the outcome of the experiments.
Experimentalist criticism of the condition upheld in Dole shows further that requiring Congress to delegate experimental policymaking as well as policy‑ implementation authority can help purge legislation of surreptitious, even impermissible preferences, and ensure that Congress is actually pursuing a power delegated to it by Article I. Suppose that Congress's real reason for seeking a national drinking age of twenty‑one was the belief that consumption of alcohol is always immoral but especially *432 so for persons under the age of twenty‑one. Such a policy objective may not be within the scope of the Commerce Clause and arguably contradicts the Twenty‑first Amendment. Yet, under traditional principles, the tying of the condition to the spending power essentially suffices to insulate it from review. By contrast, the relatively broad policy of reducing highway fatalities more clearly falls within the rubric of the Commerce Clause. Moreover, under a flexible mandate such as the one hypothesized above, the states are not required to regulate alcohol at all, so long as they find some effective means of reducing highway fatalities. [FN528] If a state, convinced of the ill effects of alcohol consumption, preferred to reduce highway fatalities by restricting the driving age rather than by other means, and could show that this was among the effective ways of achieving the overall goal, then, as argued above, it would be allowed to do so.
This first articulation of a revised anticommandeering principle is necessarily tentative and sketchy. To repeat, we recognize the strength of the arguments against any judicially enforceable federalism limits. However, whether or not it is judicially enforceable, an anticommandeering principle that is sensitive to the distinction between treating states as servants and treating them as partners seems, from the vantage point of experimentalism, especially appropriate.
Moving to a regime of democratic experimentalism dissolves what, from the perspective of the Court's existing doctrine, must be a puzzle. The puzzle is why Congress may not commandeer the states, or why‑‑put the other way around‑‑ it must give them at least a formal role in national programs, if it may take the more intrusive step of direct preemption. This is, at bottom, the question with which the Court struggled in New York v. United States and again in Printz v. United States. For us, however, preemption no longer appears as an alternative to implementation through the states; for, under an experimentalist regime, preemption is simply another word for federal regulation, which in turn gives to states and localities a substantial role in policy formation and implementation. Hence the solution to the problem of the legitimacy of administrative agencies converges with the reformulation of principles of federalism.
D. Experimentalist Federalism in Existing Legislation
Even as the Supreme Court has been adopting rules of constitutional law designed to protect a residuum of state sovereignty, Congress has been devolving power to the states at an accelerating pace. In the case of *433 the Clean Air Act [FN529] and the educational‑standards bill called Goals 2000, [FN530] congressional devolution of authority to the states can be construed as an incipient form of experimentalist federalism based on models of cooperation among levels of government better suited to problem solving under current conditions than the dignity‑based model of federalism on which the Supreme Court appears to rely. In the case of welfare reform, devolution amounts more nearly to abandonment by the federal government of responsibility for the poor. Contrasting the former cases with the latter illustrates how the ensemble of conditions on experimentalist lawmaking elaborated above distinguishes forms of decentralization aimed at encouraging local and joint learning from those where that aim is pretextual. This contrast points towards the kinds of debates over federalism likely to erupt in coming years as the tension between judicial and (different) congressional understandings of decentralization are articulated under the pressure of events.
Consider first the Clean Air Act and its amendments. Regulation of air quality might be thought to be a quintessentially federal function. Pollution generated in one state causes harm in other states (and indeed nations). This is especially true of air pollution, which once released, cannot be locally contained. Yet, state democratic processes will undervalue out‑of‑state harms, because the harmed are not typically represented in the state political bodies. For example, legislatures in industrial states may be more willing to permit manufacturers to discharge pollutants into the air that come back to the earth in other states as acid rain than to allow pollution of local streams and rivers. National (or international) regulation appears to be the appropriate solution. [FN531]
Nonetheless, effective pollution regulation must also take into account local variations. Externalities aside, differences in topography and population density, as well as in the sources, and hence, the fluctuations in the level of pollution, may make standards that are reasonable for one region unreasonable in another. Thus, an effective regime of air pollution control should consider the problem at a variety of jurisdictional levels. That is what Congress has done through the Clean Air Act.
Under the Clean Air Act, the federal government establishes a broadly defined emissions standard, but leaves to the states authority for creating plans for "implementation, maintenance, and enforcement of such primary standard." [FN532] The program of cooperative federalism utilizes key features of democratic experimentalism. For example, the state plans must establish means of collecting and making available information about the program's success. [FN533] Strikingly, the Act provides that as a prerequisite to an agency's eligibility for funding as a regulator of air *434 pollution, it must assure the federal administrators that the "agency provides for adequate representation of appropriate State, interstate, local, and (when appropriate) international, interests in the air quality control region." [FN534] The Clean Air Act thus establishes a program of joint federal/state responsibility that functions largely according to principles of democratic experimentalism.
Congress has used the model of cooperative federalism in other areas as well. Education, traditionally a state function, is a prominent example. Under the Goals 2000 legislation, states are encouraged to meet broad education goals in core academic subjects and to promote "bottom up" reform of education. [FN535] Here, states play a vital role in the formulation of goals. Federal standards are quite broad: States must improve their content standards and must set specific benchmarks by which to measure progress. [FN536] The principal substantive statutory requirement for the Secretary of Education's approval of a plan is that it hold "reasonable promise" of promoting student achievement. [FN537] On the other hand, participation rights are guaranteed in mandatory terms: "Each State improvement plan shall describe strategies for how the State educational agency will involve parents and other community representatives in planning, designing, and implementing the State improvement plan." [FN538]
A plan like Goals 2000 would survive a challenge under New York v. United States, but only, it appears, because it uses the mechanism of conditional funding. Were it a direct command of the federal government it might well run afoul of the Court's anticommandeering principle. Indeed, it might then also be suspect under Lopez, because like the statute invalidated in Lopez itself, Goals 2000 represents an intrusion into the traditionally state‑controlled domain of education. [FN539] Yet, in our view, the validity of a program like Goals 2000 should not turn on the formalism of counting it as an exercise of the spending rather than the commerce power. Instead, Goals 2000 and other programs of cooperative federalism in the area of education [FN540] ought to pass constitutional muster because they advance rather than impede the goals of federalism understood as experimentalist collaboration between the states and the federal government.
The 1996 reform of the nation's welfare laws is in many ways the evil twin of legislation like Goals 2000. It has been advertised by its supporters as devolving power to the states. In a technical sense, this may be correct: States play a larger formal role in the formulation of welfare *435 policy than they did under the old regime, and they appear to have considerable policy‑setting discretion. However, because the new legislation imposes arbitrary (and unrealistic) goals with inadequate funding, the powers of self‑ determination it gives states are illusory. Indeed, it gestures at experimentalism as a pretext for punitive moralizing that restricts the freedom to innovate while excusing the national government of responsibility for the results.
The 1996 Welfare Reform Act abolished the principal cash aid program for poor families, Aid to Families with Dependent Children (AFDC), along with a variety of other federal programs that we shall not discuss directly here. [FN541] Under AFDC, states were required to set statewide eligibility requirements and benefit levels as a condition for reimbursement by the federal government of a portion of the resulting expenditures. In this sense, AFDC was an "entitlement" program: All persons meeting criteria in a state plan that satisfied the federal limits were entitled to receive payments. [FN542] Within fairly strict federal standards, [FN543] AFDC thus also offered states an incentive to provide at least minimal benefit levels for their most destitute residents. [FN544] Even with such incentives, however, the inflation of the 1970s and 1980s substantially eroded the value of AFDC. Measured in 1993 dollars, the average monthly AFDC benefit declined from $676 in 1970 to $373 in 1993. [FN545]
The new legislation replaces federal reimbursement of state entitlement payments to qualifying families with block grants to the states. These grants are determined by a funding formula that is not adjusted for inflation. [FN546] Thus, population growth plus any increase in the price level will reduce the real value of benefits per recipient in the absence of offsetting savings in program expenditures. Congress expects the savings to come from a shrinkage in the welfare rolls as recipients move into the workforce. Indeed, the statute requires that by the year 2002, fifty percent of recipient families must include an adult working at least thirty *436 hours per week, with transition percentages in effect until then. [FN547] Congress apparently assumes that most of these people will obtain private sector jobs, as the Act appropriates very little money for job creation. Other requirements, such as the provision that a family may receive federal funds for no longer than a total of five years (with permission to states to exempt only twenty percent of the caseload) [FN548] also point to a cost savings achieved either by a transition to work or, more ominously, by exiting the state or sinking into destitution. While early experience provides a basis for cautious optimism that, at least while the national economy is booming, states will find effective means of moving recipients from welfare to work, [FN549] the new legislation provides little in the way of funding for programs that accomplish this, nor does it address the group action problem that welfare presents at the national level: the possibility that states with a substantial number of persons who appear unemployable will abandon them in the (vain) hope that they will migrate to neighboring states with more generous programs.
Even putting to one side questions of the adequacy of funding under the new legislation, and its likely consequences should the robust economy slacken, welfare reform is decidedly less experimentalist than the regime it replaced. Under section 1315 of the Social Security Act, the Department of Health and Human Services had the authority to grant waivers of federal eligibility rules to experimental state programs designed to move people from welfare to work. [FN550] The Clinton Administration was especially aggressive in granting such waivers. Prior to the 1996 legislation, it had approved seventy‑eight welfare reform experiments in forty‑three states. [FN551] The new law permits states with valid waivers to continue their approved programs until the waiver's original expiration date, [FN552] but at that point the states will have to comply with the very stringent requirements described above.
Similarly, the Family Support Act of 1988, which sought to move AFDC recipients to work, also provided for the creation of the kind of information infrastructure upon which democratic experimentalism thrives. It is worth quoting one section in full: