*341 A. Congress
The chief tasks of Congress in the experimentalist framework are to authorize, finance, and, if necessary, withdraw national support for state and local experimentation, as a means of reforming current solutions to entrenched problems or responding to emergent ones. By authorizing experimentation, Congress agrees to provide funds to local governments, or to administrative agencies collaborating with them, to pursue, by any constitutional means they see fit, such general purposes as Congress has determined, provided that all beneficiaries of the authorization agree to declare their goals and subject their activities to corresponding measurement by methods they agree upon.
Congressional authorization of subnational experiments would depend on a double determination. The first‑‑conventional in character‑‑is that the experiment identifies citizenship goods: goods that are important to the ability of citizens to act as such, yet that are unlikely to be supplied adequately to them unless the state provides at least some citizens directly or regulates markets to assure provision. Education is a familiar example of such a citizenship good, for it provides the foundations of self‑determination upon which citizenship itself is founded, yet, in the simplest case, citizens may lack the means to provide their children an adequate education. The second determination necessary to authorization‑‑reflecting the conditions propitious to experimentalism‑‑is that the citizenship good be complex. In this context, complexity means that the good consists of one or more services whose means and ends are continuously adjusted to account for the diversity and volatility of the needs of citizen users, and also to account for adjustments in services that are complementary to the main services. The difficulties of New Deal institutions suggest that many current programs are complex (their complexity overburdened the institutions); discussion of the contracting for new services points to the same conclusion.
There will be, of course, disagreements as to whether particular goods meet these tests, and how to frame experiments when they do. As in the case of disputes at the local level of governance councils, the freedom of action provided by experimentalism itself will often clear the way to resolve such disagreements. Some jurisdictions may believe, for instance, that because of moral hazards and other perverse incentive effects, provision of goods in the form of social welfare may lead to behavior that subverts their ultimate purpose of securing economic self‑sufficiency. A showing that economic self‑ sufficiency, as measured in publicly agreed ways, was increased more through training in workplace skills or through public employment or (contrary to our own strongly held view) by doing nothing at all than by any form of direct assistance, would count strongly for their views. In an experimentalist regime, these jurisdictions could pursue such alternatives on condition that they meet certain procedural requirements discussed below. Thus, whereas division now paralyzes the national legislature, and the resulting need for compromise *342 often so denatures competing proposals that their proponents cannot be held in the slightest accountable for the results, division in an experimentalist Congress would usually trigger authorization to act on the contending programs. Such actions would be faithful enough to the legislative purposes‑‑and sufficiently measurable by standards they outline‑‑to count as evidence of their plausibility.
The freedom of maneuver of an experimentalist Congress is further increased by the possibilities of authorizing experimentation incrementally or allocating resources to jurisdictions that could not, unaided, manage to put forward the kinds of coordinated proposals from which experimentalism begins. When it is unclear what good pragmatist methods will do or there is fear of disruption if they are introduced abruptly, Congress can dedicate a small, perhaps increasing, share of current allocations to a program for its experimentalist redesign. Congress could also finance experimentalism from increments to the program's budget. If the concern is, rather, that just those jurisdictions whose disorganization makes reform most urgent will initially be incapable of mustering the organization needed to meet the minimal, formal requirements of experimentalist concertation, Congress can provide separate funds to local governments and administrative agencies to permit them to build the capacity to do so. As we will see, the other institutions of the national framework can introduce experimentalism from their separate starting points and build the capacity to engage in it incrementally as well. [FN171]
This freedom of maneuver comes at the price of substantial self‑limitation in the direction Congress can impose on local governments participating in experimentalism. When Congress authorizes experimentation, it puts its own disagreements to one side on the grounds that the means to an end are only provisionally known, and the definition of the end itself will change as appropriate means are discovered. Therefore, it cannot simultaneously give preference to some means over others, or, equivalently, define the ends so narrowly that only certain means count as furthering them. In authorizing experimentation, therefore, Congress should state the publicly desired ends in abstraction from the means, and with sufficient generality to accommodate refinement through pursuit of effective solutions. Consider an example developed below: If an experimentalist Congress intends to increase highway safety, it should not prescribe an increase in the drinking age as the means. Rather, local governments should have the discretion whether to choose instead driver education, mandatory use of safety‑enhancing devices, or some other combination of means, provided only that the locales give reasons for their choices and expose those reasons to public scrutiny by benchmarking results.
Similarly, if the goal is to improve the possibilities for young people to pass successfully from school to work, experimentalist reform cannot *343 effectively be limited, for example, to the redesign of existing vocational training programs. This presumes again that the legislature knows more about the contours of a solution than the recourse to experimentalism warrants. For if the current institutions most directly charged with an official task are failing, it may be that the reason is a flaw in their very constitution, and such flaws are likely to remain hidden unless the failing program is contrasted with others constituted differently. Such contrasts, in turn, depend on an unencumbered survey of possibilities. Indeed, just as solutions often lie outside the realm marked out by habit, so the problem‑ solving institution best suited to solve a particular problem may have originated as a response to another one, and only local knowledge of its operation reveals the potential of applying it to a new task. Such are the investigations that the authorization to experiment must encourage.
The same logic requires restraint in the determination of subnational jurisdictions. The dimensions of effective government will change according to the particulars of the problem of governance; "local" actors, whatever their limitations, know best when "local" is improperly sized. Hence, the experimentalist Congress defers to local government in defining the jurisdictions that will be the protagonists of particular experimentalist programs. In reforming schools or cleaning up environmental hazards, Congress can assign the states responsibility for determining the jurisdiction‑‑local, statewide, regional, or jurisdictions wholly distinct from ordinary political boundaries‑‑to be established to treat the problem. Nor are Congress and state and local governments limited to combinations or subunits of existing governments in designating problem‑solving jurisdictions. Congress can authorize the provision of funds to administrative agencies or to local governments to be distributed in turn to groups (of citizen users, local governments, and providers) able to present promising plans for continuing collaboration (including long‑term consultation with others).
With much power delegated to subnational bodies, on what basis would the electorate choose national representatives? Campaigns for election to an experimentalist Congress would join debate about national and local strategies in novel ways, tending in time to establish closer connections between them. At the national level, debate would focus on the implications of large alternatives revealed by experimental exploration of current ones, especially in the form of proposals for new areas of experimentalism based, in part, on analogies to experiments in progress. This debate would go hand in hand with a new localism, as candidates and incumbents in democratic experimentalism challenge one another to say what use they would make of the freedom to experiment locally, and especially how they would use the vantage point of national office to encourage the particular experimental activities they favor. Thus, candidates would soon have to demonstrate mastery of current alternatives, express and motivate their preferences among these, and then pledge to advance the preferred alternative by facilitating its local development. In *344 doing this, they oblige themselves to support the information exchanges upon which success will in large measure depend. Hence, they pledge to cooperate with local officials in experiments within the experiment: efforts to scout out new possibilities, or to help direct participants to do so, before these are recognized by more formal systems of evaluation. All of this helps to break down the traditional distinction between national politics as the realm of questions so large as to be answerable in the end by professions of faith‑‑government, for or against?‑‑and local politics as service to constituents harried by the problems of wealth or poverty. The more localities in an experimentalist democracy act on familiar grand ideas, and the more representatives are accountable for the local consequences of this action, the more the familiar grand ideas come to be defined by their implications in everyday life, and the more everyday life is implicated in the articulation of novel ideas.
Such changes would not, of course, eliminate porkbarrel legislation. Even in periods of fiscal austerity there will be military bases, government laboratories, regional administrative processing centers, or dams that can arguably be located in one jurisdiction rather than in another. These are among the most prized trophies of representative democracy; experimentalism does not automatically lessen their appeal. But pork would be a residual category, not the emblem of the hidden truth of politics behind the incense of high principle. [FN172]
*345 B. Administrative Agencies
The chief purposes of administrative agencies in democratic experimentalism are to assist state and local governments in benchmarking, and experimentalism generally, especially in connection with activities carried out under congressional authorizations; to set‑‑again by a variation of benchmarking‑‑ regulatory standards for market actors; and to undertake such changes in their own activities and organization as cumulative self‑scrutiny indicates will further these purposes. In addition, certain experimentalist agencies provide citizenship goods themselves, such as the administration of public lands, as opposed to assisting local governments in providing the goods. Such agencies will have to organize and coordinate local benchmarking evaluation of the citizenship goods that they provide.
The agencies are thus the continuing organized link between the national and the local, helping to create through national action the local conditions for experimentation, and changing national arrangements accordingly. Experimentalist agencies dedicated to comparative evaluation of public and private actors must contend with the evasions and deceptions of those unwilling to submit to assessments whose outcome they do not control, and determined, therefore, to prevent the participation of actors interested in open examination of their situation. Conversely, experimentalist agencies must also contend with obstruction by those who may use participation itself to frustrate their efforts: inveterate opponents of government administration or regulation in any form, or activist citizens determined to paralyze administration, preventing approval of actions they disapprove. We focus here on benchmarking comparisons and regulation through benchmarking as the prototypical activities of the experimentalist agency. We then suggest why experimentalist administration can likely succeed at these tasks despite the inevitability of obstruction. Last, we indicate the distinctive organizational features of this novel form of administration that allow it rapidly to identify and generalize good practices as they emerge in relevant areas of activity, and to reorganize itself to better its ability to do this.
1. Benchmarking.‑‑In furnishing assistance in benchmarking, administrative agencies are almost literally creating the infrastructure of decentralized learning. Governments that want to learn from comparing what they do to the activities of others like themselves must first find each other, clarify the similarities of their activities, and define measures to rate outcomes. Assuming that superficial resemblance is not always or even often a good indication of deeper comparability of circumstance, they must‑‑by self‑ defeating circularity‑‑start with deep mutual knowledge to become useful acquaintances. Because of their ability to survey many jurisdictions from many points of view, administrative agencies can *346 break this circle. With regard to programs in, for instance, education, training, or child‑protective services, the appropriate agency convenes the local actors: to formulate suggestions for subgroupings of comparable jurisdictions (very crudely: urban and rural with further subtypes within each); to begin characterizing both individual programs (what services are actually provided and how?) and the architecture of decisionmaking (who participates and how?); as well as to devise performance measures by which to evaluate these. Participants in these meetings arrive with proposals elaborated by the relevant groupings of officials from governance councils, service providers, and citizen users, and they return to these to discuss the results of each round and prepare for the next.
The agencies must be able to take account of local diversity and resulting differences in the direction of local innovation in order to provide effective measures of performance in core programs‑‑those which in some form all jurisdictions are obligated to provide. As a consequence, the agencies' measures must themselves be diverse and composite. Such basic institutions of government such as schools or prisons, and many others besides, serve distinct and potentially conflicting ends. Prisons exact a penalty for wrongdoing against society, incapacitate the wrongdoer during his imprisonment, provide specific and general deterrence, and rehabilitate the wrongdoer for participation in social life. [FN173] Schools that enable young persons to flourish in a democracy teach respect for self and others, as well as the more or less specialized capacities needed for economic and political independence. Different jurisdictions will naturally differ in their emphasis on these purposes and accordingly prefer measures that record progress on the dimensions they prefer.
Nonetheless, common systems of measurement will be possible and widely valued because jurisdictions are unlikely to disavow the institutional purposes they do not favor. Therefore, they will be concerned with maximizing achievement of their preferred purposes while minimizing the sacrifice of other legitimate aims. Thus, those who view prisons primarily as places of rehabilitation are unlikely to think that rehabilitation is furthered if inmates literally or figuratively run riot in prison. [FN174] Conversely, those who see the purpose of penal institutions as inflicting deserved and dissuasive punishment are unlikely to prefer forms of discipline that encourage recidivism as against those that reduce it. [FN175] In the *347 school setting, those who emphasize the importance of the pupils' esteem for self and others will frequently consider this a precondition for acquiring the skills that afford self‑sufficiency; those who emphasize the latter will often treat it, conversely, as a precondition for the former. [FN176] In short, localities will be under pressure to measure their effectiveness at achieving a range of goals, not merely those they choose to emphasize.
Hence, despite the differences in emphasis from one jurisdiction to another, there is a common interest in learning more about the relation between various ends. In the foregoing examples, mutually comprehensible measurements reveal the trade‑offs or surprising complementarities between forms of discipline and rehabilitation, or between pride of identity and generally certified accomplishments. Jurisdiction X will not be able to say to its disgruntled citizens who favor policy goal P, "We cannot have P because we seek Q," if comparable jurisdiction Y pursues P as effectively as Y does, without sacrificing Q. The utility of mutually comprehensible measures creates incentives to agree to common, composite measures by which each jurisdiction monitors performance indicators that reflect not just its preferences, but the preferences of other jurisdictions as well. Beyond these composite, core measures will be ones devised by institutions and jurisdictions in distinct settings: rural schools, or schools for the deaf, or the unruly, and so on. Discussion and agreement on measures of these various sorts harness diversity of purpose to a common enterprise without imposing false uniformity.
Despite their manifestly demanding goals, these benchmarking proceedings can be effective even if their results are modest when judged by the exigent ambitions; the utility of modest results will encourage agreement that, in turn, allows the parties to learn enough to do modestly better the next time. Recall that the aim in benchmarking is simply to reveal sufficiently large differences in performance and approach to provoke local debate about the possibilities of improvement, and, subsequently, about the improvement of the groupings, characterizations, and measures themselves. Agreement on groupings, characterizations, and measures should be feasible because categorizations are understood as provisional‑‑in the case of novel programs, explicitly experimental‑‑and perfectible, not definitive. Recall, too, that benchmarking and the ensemble of learning‑by‑monitoring institutions of which it is a part, do not *348 aim to produce an exhaustive, fully replicable characterization of the products or processes to which they are directed. [FN177]
Benchmarking does not produce laboratory protocols by which successful experiments can be reproduced elsewhere. Rather, it reveals or leads to the discovery of unsuspected goals and indicates the guiding principles and related kinds of means for obtaining them. Error‑detection systems (which can themselves be benchmarked, and which can be combined with random‑assignment experiments and other familiar methods of evaluation) are then used to determine how to adapt the indicated means to the local setting to achieve the goal. Put another way, the benchmarking comparisons need only produce a usefully informative disequilibrium between current practices and prospects for improvement. Thus, even the early characterizations of programs and outcomes can produce enough learning to allow adjustment of the results of initial rounds according to the exchanges of local experience they help organize.
2. Obstruction.‑‑But the very feasibility of agency coordination of benchmarking will be an urgent reason to obstruct it for those who risk a bad showing in comparison. Their obstruction can take many forms. Jurisdictions that do little in an area can group themselves with others who do only that much or less and claim that their modest efforts and results are the most circumstances will allow. There are performance measures that notoriously conceal more than they reveal. [FN178] Insiders often find it easy to exclude outsiders, because part of being an insider is knowing which bits of the indigestibly large mass of information introduced into debate really matter for decisions.
Yet, this obstruction will succeed in the long run only on the highly unlikely condition that almost none of the local governments participating in the benchmarking and almost none of their constituents actually intend to learn anything from it. Assume, to the contrary, that there is a small group of governmental actors that does. The members of that group can find one another in the general meetings and establish the classification necessary to begin information exchange; the agency then publicizes their measurement scheme and their substantive results. If there are among the constituents of the obstructionists any who are discontent with the results their local governments provide, they can use the public information to suggest unfavorable comparisons between their home situation and that of the better performers. Thus, reversing the burden of proof, they can pressure their own government to prove‑‑by cooperative participation in benchmarking‑‑that the comparison is unwarranted. And to defend itself by cooperating, the local government *349 begins to provide information for assessments that are valid even if the initial, invidious one proves not to be.
Truculence would be sanctioned initially by the administrative agency, which would treat obstruction of benchmarking as a violation of the obligation to exchange information accepted as a condition for obtaining national funds for experimental purposes. In addition, authorizing legislation would confer on aggrieved citizen users a statutory right to participation. In judging the validity of administrative sanctions and citizen claims, courts would look for evidence that local governments and agencies had actually engaged in directly deliberative problem solving with regard to benchmarking and related activities. [FN179]
Within any jurisdiction, the discipline of those who use participation to frustrate the purposes of administration depends less on the shadow of the law and more on the pressures of competition for influence and place. Again, the crucial assumption is that there are some jurisdictions in which participants do want to cooperate in problem solving. So long as there are, and cooperation does produce results, aggressive participation as a means of obstruction will be open to criticism from within those groups in whose name it is being exercised. Industrialists or managers who quibble endlessly about providing any information or environmentalists who insist on having all that can be imagined will soon be confronted by compatriots who can cite examples where settling for less is the way to get much more. Indeed, there are some first signs that advocacy groups are in fact realizing that they have more to gain by participating in decentralized problem solving than by using strong‑arm techniques to set limits on centralized decisions. [FN180]
Experimentalist agencies would not merely coordinate the experimentalist methods of subnational jurisdictions; agencies themselves would adopt the new disciplines. Thus, in the new framework, agency regulatory standard setting similarly depends on benchmarking and error detection for the initial formulation and continuous adjustment of rules, and on a similar combination of incentives and legal sanctions for its enforcement. Subject to some important exceptions, until recently, standards in areas such as occupational health and safety, environmental protection, the transportation of hazardous materials, and the like have been implicitly premised on the fixity of the mass‑production world. They specify, in effect once and for all, means and ends simultaneously, and by reference to each other. A safe construction method for workers at risk of dangerous falls is one that provided them with safety lanyards or tethers, and other specified equipment to be worn under risky conditions; "acceptable" emissions of certain effluents are defined as those not exceeding certain concentration levels in the environment at any one moment, *350 or accumulating to more than specified amounts in any period. The actual standards are often the result of a typical compromise: The responsible administrative agencies and the advocates of those exposed to, for example, environmental and occupational hazards urge rules that would eliminate the risks. Producers countered that a regime that eliminated risk entirely makes production impossible: Workers completely secured against falls are so tightly tethered that they cannot construct anything and the only way to eliminate certain effluents entirely is to cease production of the good of which they are a coproduct. The obvious compromise is to define a list of obligatory protection measures or schedule of fines that reduce (or create financial incentives to reduce) risk without eliminating it, yet are economically acceptable. Because it is laboriously achieved, this compromise is seldom revisited, with the result that standard means and ends usually outlive the circumstances to which they were applied. [FN181]
Experimentalist regulation, in contrast, would connect rulemaking with the learning‑by‑monitoring institutions of firms. The fundamental link is that most hazards are joint products of conditions that produce waste or inefficiencies in general. [FN182] Identifying and eliminating the sources of the hazards both raises efficiency and creates the preconditions for subsequent efficiency gains. For example, a production system that leaks recurrently is discarding its own output while burdening the environment; it is, moreover, an unpredictable system. Efforts to improve it could be wasteful themselves because, so long as the system is out of control, there is no conclusive way to determine their effects. Similarly, many of the worst construction hazards result from flaws in the design of the building under construction (in the case of steel‑girder erection, poor design may produce torsion at the joints which can then spring apart with catastrophic effects for workers in the vicinity and the structure as a whole), or from poor training in construction methods (which endangers workers and reduces the value of the finished building). Thus, in experimentalist administration, the search for efficiency is incidentally a search for regulatory improvements and vice versa.
The administrative agency can, in turn, use this connection between regulatory goals and efficiency to promulgate regulations in the form of rolling best‑practice rules. Such rules require regulated entities to use processes that are at least as effective in achieving the regulatory objective as the best practice identified by the agency at any given time. In one variant, the current production method that creates the lowest level of *351 risk is the standard all producers must meet (within a certain grace period), either by adopting those methods or devising equivalents. In another, polluters are pushed from the bottom of the heap rather than pulled toward the top: The level of risk defined by the most hazardous operators defines a regulatory purgatory from which polluters must ascend (again within an agreed period); the acceptable minimum rises as the worst performers improve. In both cases, benchmarking establishes and periodically updates the standard to incorporate improvements, raising the ceiling in the one case and the floor in the other. Firms that achieved significant improvements as part of the first, rising ceiling type of programs of increasing efficiency, would gain a further advantage in establishing them as public standards. Competitors would have to incur the costs of adopting some version of the new methods sooner than market competition alone would have required, reducing risks without an offsetting gain in efficiency, or paying a fine. In the meantime, the innovator could be innovating again. Specialist producers of equipment‑‑pumps, valves, many kinds of machine tools, for example‑‑that reduce risks by increasing the reliability on which efficiency improvements ultimately depend, would, under such regulations, have an additional motive to do what they often do in any case: Make their current products obsolete by building more capable models.
Notice that this method eliminates the problem of information hoarding associated with many of the market‑based alternatives to traditional rulemaking. In the standard market‑based alternative, firms are assigned tradable rights to emit certain quantities of pollutants in a given period. [FN183] Those who reduce emissions below the level of their entitlements *352 can sell the rights to emit the difference to those firms who emit more, and the prospect of doing so is the incentive to cut emissions. [FN184] The difficulty with this system (other than the vexing problem of determining and constantly adjusting the price for the entitlements) is that it encourages firms to hoard the know‑how they acquire in pollution reduction. The greater the gap between their knowledge and their competitors', the greater the prospective proceeds from the sale of rights. But if this secrecy is rational for the individual firm, it is plainly irrational for the society as a whole to pay the costs of multiple, independent reinvention of waste‑reduction methods.
Rolling best‑practice rulemaking and the incentives it creates are not limited to circumstances in which the search for incremental efficiency gains leads directly to reductions in known hazards. The method can be used prospectively to create incentives for attractively safe products that do not currently exist. In some production methods, hazardous materials or circumstances do not coincide with inefficiency; instead, in these methods, the hazardous materials or circumstances are a necessary and irreducible byproduct or component of an efficient process or product. In these cases, by definition, efficiency gains pursued along the current trajectory of technological development will not substantially reduce the hazard. The use of fluorohydrocarbons in current refrigeration systems or as a propellant, or of gasoline in current automobiles are examples. The only way to eliminate the hazard is to find a substitute for the product or process of which it is a part. It is here that prospective rolling best‑practice rules play their most significant role. The traditional method of encouraging such substitution of safer products or processes is by technology‑forcing legislation that imposes fines on producers that fail to reduce hazards to a level achievable only by some (unknown) alternate technology by a distant date. [FN185] These rules touch off a game of chicken: Recalcitrant producers do nothing and encourage others to do the same *353 in the hope that their inaction can eventually be used to persuade courts or administrative agencies that the original goals were infeasible. Collusion against the legislative or administrative goal is easy because it is tacit, and because potential innovators will be deterred from developing alternatives for fear that the results will fall short of the requirement. Under a prospective rolling best‑practice rule, in contrast, the best alternative solution available by a distant date would (re)set the standard from that time on. Innovators are rewarded for outdoing the competition, and hence encouraged to outdo one another. Only explicit (and therefore detectably illegal) collusion could bind all from developmental activities that would spur the rivalry of the others.
Likewise, rolling best‑practice rules can be used potentially, to reduce sources of risk in novel or experimental products, even before the precise nature of those sources can be identified. Potential rolling best‑ practice rules are useful where product life cycles are short with respect to the time needed fully to test and improve the safety of a product under real‑ world conditions (computers, much software, and complex financial products) or where initial real‑world failures would be catastrophic (pharmaceuticals, foodstuffs, and products bound for space or the battlefield). The way to reduce risks under these circumstances is to characterize more and more precisely the sources from which hazards may derive and to reduce and monitor each precisely characterized source more and more effectively. Contaminants are much more likely to be introduced into batches of foodstuffs at some points in processing than at others: when harvested, during transport, when fermenting, etc. Precise accounts of potential hazards and countermeasures‑‑ including error‑detection systems for maintaining and improving the countermeasures themselves‑‑can then be developed to lower the possibilities of dangerous damage to products of various types. [FN186] In the production of foodstuffs, these are called HACCPs, for hazard analysis of critical control points; [FN187] the Federal Drug Administration maintains an analogous set of standard operating procedures with regard to the production of pharmaceuticals. [FN188] The potential rolling best‑practice rule benchmarks these prophylactic measures and establishes them as the initial production standards in the relevant industries. As these standards help producers increase the reliability of both products and processes, efficiency considerations again encourage compliance. Indeed, to the extent that risk reduction is a condition for creating markets in potentially hazardous products as diverse as sophisticated financial derivatives and foodstuffs, *354 regulation of this sort amounts to a form of codevelopment between producers, users, and regulators. The decomposition of as‑yet undeveloped products and processes enables producers and regulators alike to learn from real‑world experience in circumstances in which no experience with the completed product or process yet exists.
3. Novel Forms of Organization.‑‑Finally, to update and propagate benchmarks, and the background understanding they suppose, while assessing compliance with them, experimentalist administration introduces novel forms of organization. Experimentalist regulatory agencies recall the design and problem‑solving teams of learning‑by‑monitoring institutions in the way they pool various kinds of expertise in the evaluation of different situations. Recall that in learning by monitoring, collaboration among teams within a firm and between the firm and its suppliers breaks down the distinction between the individual actors' roles. [FN189] Similarly, from the point of view of the composition of their personnel and their personnel's career paths, the operations of experimentalist agencies blur the distinction between regulatory agency and regulated entity‑‑without obstructing public scrutiny of administrative activity. We call the ensemble of these forms peer, team, or participatory administration to emphasize two points. First, in their organization, agencies will come to utilize work teams in much the same manner that the new firms do, and thus the agencies will be structured as participatory units. Second, and more important, peer, team, or participatory administration refers to the close working relationship between regulator and regulated entity that will facilitate the agency's role as conduit of information. To an important degree, peer administration provides a mechanism by which agencies set rolling best‑practice rules. Agency staff, observing (or more properly, participating in the activity of) the regulated entities first‑hand, develop a strong sense of emerging processes, and by pooling knowledge of these processes with staff at other locations, agencies can identify emerging best practices.
The New Deal pattern of organization‑‑or, rather, the characteristic dilemma of that organization‑‑is useful as a point of contrast. On the one hand, New Deal agencies were meant to remain distinct from the social or economic worlds they regulated or administered; distance and detachment were thought to be requirements if the agency was to exercise its (delegated) lawmaking authority to establish general, enduring, and impartial rules. [FN190] Hence, agencies needed an extensive, professional, and independent staff, competent to gather information for rulemaking and adjudication. But the same agencies, on the other hand, had to deal directly and continuously with the interests they regulated. Otherwise, no matter how well staffed, they lacked the fine‑grained information about emergent possibilities or potential evasions required to exercise *355 their (delegated) lawmaking authority in the interest of the particular segment of the public within their jurisdiction. Hence, agencies needed to supplement or even supplant staff work with complex alliances with various interests, aiming to expose what normal research alone could not uncover. The results were continuing struggles of the agencies to be in the worlds they regulated, but not of them, and the concomitant oscillations of their leading officials between magisterial lawgiving and factional politicking. Thus, as an architect of the administrative state and founder of the Securities and Exchange Commission (SEC), James Landis, for example, likened the role of the administrative agency to that of a board of directors for an industry, able to use its fact‑finding powers and panoramic perch to reach judgments more balanced and farsighted than those accessible to more partial parties. [FN191] But as the senior operating official of the SEC, he was always playing off the independent accountants against the corporate treasurers to determine what types of corporate financial information could reasonably be disclosed for purposes of evaluating publicly traded securities. [FN192]
Experimentalism links benchmarking, rulemaking, and revision so closely with operating experience that rulemakers and operating‑world actors work literally side by side‑‑but, to repeat, in plain view of the public‑‑and thus, largely overcome the distinction between the detached staff of honest but imperfectly informed experts and the knowledgeable but devious insiders they regulate. Inspection by peer administrators is a characteristic institution for establishing these connections. Assume that initial regulatory benchmarks have already been fixed with regard to, say, forest‑fire prevention or the safety‑ related operations of nuclear power plants (to pick two examples to which we will return in detail in a moment) by some procedure of extensive consultation. Then the task of the peer inspectors is to determine whether particular operating units are in compliance with the benchmarks, and to grasp the general lessons, if any, regarding obstacles to compliance in cases where they are not, or regarding the need to set more demanding standards, when they are. Such evaluative learning is in effect a kind of higher‑order error detection: It aims to discover why the error‑detection institutions of a particular unit were either unable to detect and correct the disturbances that obstructed compliance or were so effective that they raised performance above currently established levels.
But of course the initial benchmarks and, more generally, standards for purposes of environmental or consumer protection, occupational health and safety, or the coordination of complexly interconnected products with aspects of public infrastructure (as in telecommunications) can all be set by similar means. Thus, whereas practitioners may currently testify at hearings or serve on ad hoc committees devoted to writing definitive *356 rules, in peer administration they would serve on the standing bodies that create the framework for rules that can be periodically updated as practice warrants, and help establish the forms of participatory review suited to keeping the rules up to date in various settings. Peer participation on the problem‑solving model could also be used occasionally to identify areas where the administration might change its own structures to facilitate experimentation: for example, by creating a service to help jurisdictions that lack the experience to formulate plans for experimental projects to do so or to aid others threatened by deadlock to advance. Again, these services would draw on the experiences of respected practitioners. Selection would favor those who had learned to break deadlocks not in some arbiters' way of splitting differences (which entrenches fundamental assumptions and thus, in time, obstructs wide‑reaching change), but rather by showing the parties how, in pursuing new projects that sidestep or clarify their differences, they might come to a new understanding of those projects. Selection would also favor those who became adept at bootstrapping planning, in which each project increases the jurisdiction's capacity to formulate a comprehensive and better directed successor. By such means, peer administration could become at once the frame of national experimentalism and, together with a new style of electoral politics, an instrument for connecting that frame to the local activity it regulates.
C. Antecedents and Lessons
Just as the institutions of learning by monitoring in the private sector have advantages which lead to their diffusion, so principles of democratic experimentalism in administration are often sufficiently attractive to both public and private actors to be adopted piecemeal in the public sector. That agencies designed on conventional principles have begun to reorganize themselves along the lines described above in recent years [FN193] argues for the robustness of peer administration as a general response to problems of regulation under conditions of volatility and diversity. There are, in addition, suggestive historical cases where measures of peer administration were adopted avant la lettre by American agencies operating in environments which were anomalous for their day, [FN194] but which approximated the situation administration now faces.
We begin with negative lessons by considering a classic failure, which upon closer inspection, reveals itself to have been a near success. This anticipatory history helps specify key preconditions for experimentalism to thrive‑‑most critically, some mechanism for forestalling litigation until after an experimental regulatory regime has had the opportunity to demonstrate its worthiness or lack thereof. We draw these lessons chiefly from the furor over regulatory attempts to require air bags on *357 automobiles during the last quarter century. We then reach back to a still earlier antecedent in the Forest Service and forward to contemporary success stories concerning nuclear power plant safety and environmental regulation to show how different regulatory regimes, beginning from nearly contrary starting points, are converging on the same experimentalist methods.
1. National Highway Traffic Safety Administration.‑‑The National Highway Traffic Safety Administration (NHTSA, or the Agency) provides both an illustration of administrative experimentalism that might have been, and a cautionary tale about the capacity of current practice and doctrine to stifle regulatory innovation in its infancy. NHTSA was created in 1966 explicitly to bring science and technology to bear on the problem of reducing highway slaughter. [FN195] But in the course of fifteen years, NHTSA's decisionmaking became so unwieldy that in Motor Vehicle Manufacturers Association v. State Farm Mutual Automobile Insurance Co., it was chastised by the Supreme Court for failing to give due consideration to manifestly promising safety devices. [FN196] The authoritative study of NHTSA by Mashaw and Harfst, The Struggle for Auto Safety, recounts the Agency's decline from putative avant‑garde to laggard. [FN197] From this history, Mashaw and Harfst draw a large conclusion. As they understand it, American legal culture is preoccupied with individual rights and permanent contests for authority between the President and Congress on the one side and the federal and State governments on the other. [FN198] Contrasting the Agency's promise with its performance, their study presents the history of NHTSA as an object lesson in the constraining influence of that legal culture on all forms of regulation. [FN199]
Yet beneath, or, rather, entwined with, the story of an agency that belied its own efforts to seek practical truth, the authoritative account also contains a counternarrative of an agency that might have adopted certain experimentalist methods but for the accidents of political happenstance; of an agency, moreover, that arguably did adopt an oblique form of experimentalism in response to the State Farm decision; and, indeed, of an agency, and of a whole legal regime more generally, that could by modest reform encourage experimentalism quite directly. Juxtaposing the actual outcome with the counterfactual alternative thus suggests how easily we might have stumbled upon a variant of administrative experimentalism without intending to and how we might achieve one if we did.