*358 First the balance of failure. NHTSA was the first of the agencies such as the Occupational Health and Safety Administration, the Environmental Protection Agency, and the Consumer Product Safety Commission, formed at the high noon of confidence in centralized government, to complete and perfect New Deal institutions by fusing democracy with science in a new way. [FN200] Operating under vague and general statutes, and formulating rules incrementally, by adjudicating individual cases through collegial decisionmaking independent of executive direction, the previously established agencies had often seemed impervious to the very interests they were meant to protect and all too susceptible to those they were meant to domesticate. [FN201] With more specific mandates, decisionmaking processes aimed not at adjudication but rather at making general rules under conditions open to the participation of potential beneficiaries, and directed by more powerful, hence more acceptable administrators, the new agencies were intended to reverse this relation. [FN202]
This general reorientation went hand in hand with the reconceptualization of automobile accidents from an epidemiological standpoint. [FN203] In the traditional view, accidents and injuries were coincident: Drivers and passengers were injured or killed when cars collided with each other or with fixed objects. For the epidemiologist, the human host is only harmed upon contact with the agent of rapid energy transference in the environment of the collapsing car. [FN204] Hence, anything that altered the environment to prevent the agent from contacting the host prevented the accidental harm. This, in turn, suggested that to improve highway safety the agency should focus less on preventing car crashes‑‑which required laborious and unreliable efforts to change driver behavior‑‑and more on reducing the hazards of the automotive environment by using regulation to encourage use of passive safety devices (self‑securing seat belts or air bags) or active ones (manually secured seat belts) to protect car occupants from the effects of the "second," internal collision. [FN205] As seat belts were widely used in the late 1960s, and air bags, under development since *359 the early 1950s, were on the verge of broad deployment, [FN206] there was nothing technologically extravagant about the approach. Nor was there anything rash in the agency's decision to promote these changes wherever possible by moving gradually from design standards based on equipment in current use to performance standards that would eventually specify the characteristics of whole automotive subsystems. [FN207]
Two closely related problems thwarted the agency's projects and turned many of its efforts down the path of self‑parody. First, absent any evidence as to whether firms could comply with agency standards if they made reasonable efforts to do so, courts simply did not know how to evaluate the claims of those manufacturers who asserted that the standards as written were unreasonably burdensome. The resulting ambiguities were judicial accidents waiting to happen, and several did. To take only the best known example, in Chrysler Corp. v. Department of Transportation, the United States Court of Appeals for the Sixth Circuit acknowledged that NHTSA was authorized by statute to induce firms to introduce new technology, but found that Standard 208, establishing the performance criteria for passive restraints eventually to be required on new automobiles, was defective because of three ambiguities in the specifications of the dummy with which the safety device was to be tested. [FN208] The court found that these ambiguities violated the statutory requirement that agency standards be "objective," in that manufacturers using different but equally compliant test dummies might market passive safety devices with different performance characteristics. [FN209] Eventually some might be unjustly penalized for these differences. [FN210] In fact, the specifications for the dummy had been elaborated by a committee composed of industry engineers; more important, as the agency observed, no manufacturer would be penalized for choosing one interpretation of the specification over another. [FN211] But since the court believed otherwise, and since it erroneously believed that manufacturers develop equipment only after they have perfected the corresponding test devices, it ordered that implementation of Standard 208 be delayed until the ambiguities were resolved and manufacturers could adjust to corrected specifications. [FN212] This and similar decisions that overturned standards just because they were rigorous enough to allow precise identification of ambiguities [FN213] devastated *360 NHTSA's ability to promulgate performance standards of any kind, emboldening reluctant manufacturers to try intransigence under all possible legal pretexts before compliance, while intimidating the rulemakers through the prospect of endless challenges. [FN214]
The second problem that thwarted NHTSA's efforts was that it found no way to assess reliably whether the driving public would welcome its regulatory interventions as an advance in public safety, disregard them as inconsequential nuisances, or rebel against them as violations of an elemental freedom of movement. [FN215] Public reaction was for the agency as unpredictable as judicial reaction, and this second unpredictability, symbolized in the disastrous episode of the interlock rule, was as intimidating as the first. [FN216] In response to complex technicalities related to the Chrysler decision, and in revision of its own earlier determinations, NHTSA in effect required that 1974 model‑year cars be equipped with an interlock device that disabled the ignition unless the seatbelts of occupied seats were engaged. [FN217] The public revolted. There were terrifying stories of women unable to flee rapists and amusing ones of parking attendants driven mad by incessant buckling. [FN218] Senators rose on the floor to tell of constituents strapping turkeys and dachshunds snugly into the front passenger seat. [FN219] Congress then added injury to insult by amending the Agency's enabling statute in 1974. [FN220] Instead of insulating NHTSA from judicial review, it subjected any future passive‑restraint standard to a legislative veto, and prohibited the Agency outright from requiring any interlock device. [FN221]
Given these obstacles, the Agency sought politically and judicially acceptable alternatives to rulemaking as a way of demonstrating vitality if not efficacy, and temporized issuing those standards‑‑208 above all‑‑that it could not abandon without publicly disavowing its original hopes. [FN222] The search for alternatives led quickly to a massive program for forcing manufacturers to recall and repair cars found to have dangerously defective parts. [FN223] Politically, the campaign was welcome because as car prices rose during the inflation of the 1970s, consumers became less and less tolerant of defects, and let their representatives know it; [FN224] judicially, it proved unexpectedly acceptable because courts treated the Agency in this regard with the leniency accorded consumer plaintiffs in *361 product liability cases rather than the stringency applied to it as a defendant in administrative law matters. [FN225] Taking account of the response rate to recall programs, the costs of accidents occasioned by the trip to or from the local recall center, and the risks of introducing new defects while repairing the known one, it is likely that the diversion produced on balance more harm than good; but there is no doubt that it was successful, for a time, in diverting attention from the near breakdown in rulemaking. [FN226]
Consistent with this retreat from rulemaking, NHTSA eventually concluded that detachable automatic seat belts would not result in a significant reduction in highway deaths or injuries, and accordingly rescinded Standard 208 in October 1981. [FN227] The decision to abandon the standard was in part a reflex response to the advent of the Reagan Presidency and the era of deregulation it promised, but, perhaps in equal measure, also the product of NHTSA's despair at finding a solution that was both innocuous and minimally effective. [FN228] Given the constraints, the logic behind the rescission order had a compelling aspect: If the public, out of concern for freedom of movement and fear of entrapment, would at most accept passive devices that could be detached by a release button in the manner of the quintessentially active device, the manual seat belt, why require passive restraints in the first place?
But the Supreme Court reasoned differently. In State Farm, the Court found NHTSA's rescission of Standard 208 arbitrary and capricious because the Agency had failed to give adequate consideration to requiring air bags or nondetachable automatic seat belts as an alternative. As willing to punish the Agency for the ambiguities of inaction as the lower courts had been to penalize the ambiguities of rulemaking activity, the Supreme Court criticized the decision for assuming, on no evidence, that the driving public would defeat passive devices just because it was technically possible to do so, and wondered why air bags, which were unintrusive to the point of invisibility, had suddenly disappeared from the menu of possible regulatory solutions. [FN229] The controversy was eventually mooted when, responding to consumer demand, automobile manufacturers equipped new vehicles with air bags as standard equipment. Only then, in an ironic if not pathetic gesture, Congress responded by including a provision in the Intermodal Surface Transportation Efficiency Act of 1991 directing NHTSA to amend Standard 208 to require air bags on *362 all new vehicles. [FN230] The new requirement became fully effective for passenger vehicles on September 1, 1997‑‑just after NHTSA promulgated a final rule authorizing the depowering of air bags to prevent injuries to children and small‑statured adults. [FN231]
Now consider an alternative, experimentalist outcome that is much closer, it turns out, to actuality than this account of nearly insurmountable obstacles would make it appear. The regulatory result would very likely have been different had just one car manufacturer been willing to build a fleet of vehicles equipped with air bags and had NHTSA been willing to defer establishing standards until the fleet was tested on the road. For the experience of building and using this fleet would have put to rest questions about the acceptability of the device to the driving public while addressing judicial concerns about ambiguity. If consumers were willing to drive the cars it was unlikely that they would rise in protest against regulations making air bags standard equipment; if manufacturers sincerely feared that they would be penalized for misinterpreting ambiguous performance standards, they could adopt the solution of the pioneer.
In fact, this experiment was almost run. The then‑Secretary of Transportation, William Coleman, proposed the idea of a cooperative demonstration project in 1975, and General Motors, long a proponent of air bags, was willing to participate. [FN232] Indeed, General Motors had already begun in the preceding years to use its problem‑solving capacities to provide working solutions to regulatory problems. As the only car company ready to produce cars equipped with air bags at the time of the Chrysler decision, it helped the Agency develop dummy standards nine months after the Court demanded clarification. [FN233] In the end, only the accidents of national politics prevented construction of the demonstration fleet. Coleman left office when President Ford lost the election of 1976 to President Carter; under its new Secretary Brock Adams, the Department of Transportation simply saw no need for experimentation, arguing with almost cynical consistency both that consumers would surely take no notice of seat belts, [FN234] and that demonstration projects were a "weak" regulatory strategy. [FN235]
But had this test been conducted it seems likely that the Court in State Farm would not have felt obligated to imagine all the ways, however improbable, that the development of a novel technology might prove infeasible or unjust, and the apparent limits of American legal culture *363 would have been extended without having been traversed. Mashaw and Harfst say as much: Besides advancing Standard 208, they write, "[a] similar demonstration approach might well have saved Standard 121, the truck antilock braking standard, and yet another might have enlisted the support of the tire industry for tire performance standards." [FN236]
Observe as well how easy it would have been to pass from a series of demonstration projects that almost were, to explicit modification of the legal regime along the lines suggested above. Generalizing the experience acquired through demonstrations, all organized by utterly conventional agreement under current law, NHTSA could have developed rolling best‑practice standards in tandem with pilot projects, encouraging competitors to pool standard‑relevant knowledge (as General Motors did in the case of the dummy specifications) along the way. The Agency, and others following its example, would have thus created a de facto regime in which pre‑enforcement challenges to the feasibility of rules and standards would be pointless because rules and standards would not exist, let alone give rise to enforcement actions, until their feasibility was established. Should further clarification prove necessary, it would be a short step from this regime to a statute making forbearance from pre‑enforcement litigation the rule in experimentalist litigation, thus insuring that courts decide administrative law cases against a detailed backdrop of fact.
In one regard such a shift would restore, ironically enough, the state of affairs that had obtained in administrative law in the period before Chrysler and related decisions‑‑when agencies imposed order by applying rules case by case, and courts, when invoked, reviewed the agency decisions in light of the facts. [FN237] The difference is that in the experimentalist regime, courts would be reviewing decisions and factual scenarios produced by the articulation of performance standards, potential and prospective, inconceivable in the epoch before Chrysler. If this is the path we take, the confusions of the last decades will come to seem more a detour than a necessity; and doctrinalists of the future will wonder how, for a time, courts could have inverted the very idea of thought experiment from a technique of imagining a new possibility from an imaginary vantage point to a method of finding legal obstacles‑‑some restricting agency action, some obligating it‑‑by imagining all manner of possibilities. As Paul Verkuil argued in the mid‑1970s, and as Mashaw continues to argue persuasively today, elimination of pre‑enforcement litigation would go a long way towards reducing the arbitrariness of judicial review. [FN238] Moreover, as Mashaw presents no deeply entrenched obstacles in American legal culture to doing so, it is hard to see why that culture is, as argued in The Struggle for Auto Safety, itself an obstacle to significant *364 reform. [FN239] In short, the counternarrative may, in the long run, prove more important than the farce that obscures it. Before turning to those reforms of the courts, we consider three more examples of past and present experimentalist administration.
2. Antecedents in the Forest Service.‑‑The Forest Service of the United States Department of Agriculture (the Service) attracted public and scholarly attention in the late 1950s for its ability to adjust complex policy goals to extraordinarily diverse local settings, largely through controlling, and learning from, the exercise of discretion by its lowest level operating agents, the forest rangers. As this was the period in which the view of bounded‑rationality institutions [FN240] was formulated, academic efforts to grasp the success of the Service were at pains to portray its organization as a centrally directed hierarchy of this type. Indeed, the leading study of the subject, Herbert Kaufman's The Forest Ranger, draws heavily on the work of Herbert Simon, a progenitor of the bounded‑rationality idea. [FN241] Here, using mainly the evidence reported in Kaufman's study, we show that this success, and especially the reciprocal reshaping of general rules and local practice that made it possible, rested instead on methods that foreshadow peer administration introduced into the Service at its founding. Thus, this brief example does double duty. It buttresses the claim that experimentalism is a general system of problem solving under conditions of diversity and complexity by showing that where problems are addressed with notable success under such circumstances, the solution is by this and not other means. It also points to the possibility that American public administration contains an organizational tradition, rooted in a variant of Progressivism, that might serve as one operational precedent for an extensive system of experimentalist administration. We conclude our discussion of the Forest Service by briefly recounting the problems that the Service has encountered in the decades since Kaufman completed his account.
The modern Forest Service was established in 1905, when responsibility for protecting and managing the country's national forests was transferred from the Department of the Interior to the Department of Agriculture, which until then had been limited to gathering historical, statistical, and technical information concerning forestry and forest products. [FN242] The Service was charged from its inception with pursuing various, often conflicting goals in disparate settings, and the complexity of the tasks and diversity of the settings increased rapidly in the following years. [FN243] By the late 1950s, the Service had jurisdiction over 151 million acres scattered through forty‑two states and Puerto Rico, including remote *365 regions of the Rockies, relatively developed lands on the East Coast, the coastal mountains of Alaska, and the tropical jungles of Puerto Rico. [FN244] Today, national forest lands total 192 million acres in all regions of the country (although concentrated in the Western States and Alaska), or nearly nine percent of the United States landmass. [FN245] Within these areas, the Service manages timber logging and sales, reforestation, fire control, grazing, the use of watersheds and natural habitats, and recreational activities. [FN246] The balance of these activities varies widely from region to region, and even where it is approximately the same, differences in soil composition, vegetation, climate, and accessibility require that similar goals be pursued by different means. [FN247] Once arrived at, local and national policies must be frequently revised to take account of, among many other things, changes in product markets that make, say, lumbering more attractive than grazing, or developments in labor markets that change the attractiveness of a career in the Forest Service itself to potential recruits with differing levels of education and professional ambition. [FN248]
From the beginning, day‑to‑day management of local activities was delegated to Forest or District Rangers, each responsible for a single district within a national forest, and reporting to a forest supervisor (who in turn reported to a regional forester with jurisdiction for several national forests). [FN249] Prior experience under the Department of the Interior had shown that efforts to control the use of public lands directly from Washington resulted in interminable delays in granting permits and other decisions, and obstructed conservation. [FN250] The result was a founding and enduring principle of delegation subject to review that was clearly announced in a letter from the Secretary of Agriculture to the Chief of the Forest Service in 1905: "In the management of each reserve [now national forest] local questions will be decided upon local grounds . . . . General principles . . . can be successfully applied only when the administration of each reserve is left largely in the hands of local officers, under the eye of thoroughly trained and competent supervisors." [FN251]
From the vantage point of the organizational understanding of the 1950s, the puzzle was how the necessary exercise of local discretion could be sufficiently controlled to ensure conformity to the central direction. The answer, as provided in Kaufman's account, was comprehensive written regulation, policed and updated by a system of inspections and clearances, *366 and complemented by personnel policies that encouraged compliance by building esprit de corps. [FN252] In the late 1950s, the regulations were contained in a seven‑ volume Forest Service Manual, four volumes of which were issued to District Rangers. [FN253] Continually revised, and supplemented by insertions at the regional and national‑forest levels, as well as by technical manuals prepared by the Service staff in Washington, these volumes established procedures and standard responses for each class of problem the Ranger might encounter. [FN254] In addition to uniform, Service‑wide guidelines, specific local plans for implementing policies for fire prevention and timber management were required of all districts. On their own initiative, other districts could implement plans for recreation, grazing, and other major Service objectives. [FN255] Such plans included quantitative performance goals which set target levels for numerous areas, including timber sales, fire control, wildlife preservation, and the number of visitors to the forest, among others. [FN256]
Conformity with the rules and plans was secured, to begin with, by the requirement that all decisions involving the redisposition of anything more than trivial amounts of resources be cleared in advance by superiors (and in turn their superiors). [FN257] Deviations were detected after the fact by regular and extensive inspections by superiors of the work done by their subordinates. [FN258] For purposes of the functional inspections looking into all aspects of forest management, District Rangers were required to maintain diaries recording their activities by thirty‑minute intervals, as well as detailed records of all expenditures and income, so that results could be compared to the efforts and resources that produced them. [FN259] Homogeneity of interpretation was underpinned by a homogeneity of outlook secured by filling entry‑level professional positions with persons sharing a common background in forestry studies, rotating recent entrants through various regions (to build loyalty to the Service rather than particular locales), and filling supervisory jobs through internal promotion. [FN260] That, at any rate, is Kaufman's account, seen through the lens of hierarchical organizational principles.
But the statements of officials and District Rangers interviewed by Kaufman, together with administrative rules and practices under which they operated, suggest an interpretation of much of this directive machinery as an instrument of peer administration. "Again and again," Kaufman writes, "the researcher is told by officers in the field that they do the bulk *367 of the work even though others sign the papers, and their superiors freely acknowledge this dependency." [FN261] It was the District Rangers who, for example, recommended the issuance or denial of permits, established the feasibility of land transactions, and furnished the plans on which production quotas and targets were based. [FN262] Headquarters relied so openly on the judgment of local officials that new procedures and equipment were only introduced after field testing in pilot ranger districts, and in consultation with those who participated in the experiments. [FN263] In sum, "leadership decisions about what the Forest Service can and should do rest in the last analysis on what the field men tell the leaders." [FN264]
This dependence on local information in turn calls attention to a lesser role for the Manual than it plays in Kaufman's account and transforms the role of inspections and diaries. Thus, because field officers objected to its unwieldiness, the Manual remained incomplete. Indeed, plans to add three additional volumes were abandoned in the late 1950s in favor of a project to reduce and simplify the existing ones. [FN265] Inspections, accordingly, appear to have been less occasions for verifying compliance with a master plan than a method of pooling and evaluating experience from the whole Service. Because of the policy of internal promotion, most inspectors had been District Rangers, [FN266] and this experience, combined with the activity of inspection itself, put them in a good position to identify and propagate good practices and criticize deficient ones; and as a rule the emphasis was more on the former than the latter. "[T]he stress in inspections is on training," Kaufman writes, "and the inspectors may be said to constitute an itinerant school . . . ." [FN267] The highest officials of the Service, moreover, were quite explicit in distinguishing investigation, defined as the search for "something that's dishonest or otherwise wrong," from inspection, whose purpose was "to see how together we can do a better job." [FN268] Given this distinction, it is unsurprising that field officials, "rather than fearing inspection, tend[ed] to welcome the opportunities it afford[ed] them to keep abreast of developments in *368 the organization . . . and to give their own ideas to their superiors at first hand." [FN269]
In a setting where inspection was a rudimentary form of information pooling, activity diaries documented de facto organizational routines and allowed comparative evaluation of their effectiveness in the context of local circumstance. [FN270] In other words, defects in reporting often revealed defects in operation. Thus, the sharpest criticism was reserved for District Rangers whose confused or incomplete procedures, as reported in their own logs, made risky situations more dangerous still, whatever the actual outcome, [FN271] and conversely, commendations were directed to those who planned what they did and did what they planned. [FN272]
Such benchmarking was further facilitated by the Forest Service policy of transferring Rangers from one district to another every few years. [FN273] This policy facilitated error detection as the new Ranger might see mistakes that the previous Ranger, entrenched in habit, had failed to notice. This practice also allowed the Service to benchmark the performance of each Ranger against the Ranger previously responsible for that district. Furthermore, the exposure of each Ranger to a variety of local conditions built a base of diverse experiences which could be called upon when responding to new or fluctuating conditions.
In sum, despite Kaufman's disposition to understand the Forest Service according to the paradigm of a large‑scale hierarchical organization, the facts he discloses tell a different story. The Service discovered that it could best coordinate national policies with local circumstances by a decentralized experimentalist system of error detection through information pooling and benchmarking.
Yet, if the early experience of the Forest Service is an antecedent of experimentalist public administration, its more recent experience could be read to suggest limits of experimentalism, in so far as experimentalism is inspired by the Progressive experience. Since its inception, the Service has been charged with managing its land "for six renewable surface uses‑‑outdoor recreation, rangeland, timber, watersheds and water flows, wilderness, and wildlife and fish." [FN274] The potential for conflict among these uses is obvious, and that conflict has been realized in recent decades: Timber production on Forest Service land has increased dramatically. *369 [FN275] Meanwhile, other developments have stressed competing uses. Most notably, the enactment of environmental protection statutes such as the National Environmental Policy Act, [FN276] the Endangered Species Act, [FN277] the Clean Air Act, [FN278] the Clean Water Act, [FN279] and the Wilderness Act, [FN280] placed higher priorities on the nontimber uses of Forest Service land. Recreational use has also increased dramatically in just the last decade (although timber production has experienced a sharp decline in that same period, partly due to court orders based on environmental protection statutes). [FN281] In principle, the Forest Service's historical commitment to decentralized decisionmaking should have positioned it well to respond to the new conflicts‑‑for its own and other agencies' experience demonstrate that local flexibility plays a key role in accommodating conflicting demands by uncovering new possibilities. But in practice the new priorities were less amenable to such solutions for a number of reasons.
First, Congress's continued insistence on emphasizing timber production often crowds out other uses. [FN282] Second, the underlying concerns of environmental protection legislation are generally best managed by focusing on ecosystems as the relevant geographic unit. [FN283] Yet, national forests under Forest Service management typically constitute fragments of larger ecosystems that are divided among private landowners and other federal agencies [FN284]‑‑typically the National Park Service, the Bureau of Land Management, and the Fish and Wildlife Service. [FN285] The Forest Service has experienced severe difficulties coordinating its activities with other agencies, partly because of failure to consult from the outset of projects, [FN286] and partly because the various agencies collect data in noncompatible formats and have not yet developed the means for meaningful pooling of information. Finally, the sheer procedural burden of complying with (or in some cases failing to comply with) Congress's often *370 contradictory mandates exacts a heavy toll on the Forest Service: "[C] onducting environmental analyses and preparing environmental documents consumes about 18 percent of the funds available to manage the national forests and approximately 30 percent of the agency's field resources." [FN287] The net effect of all this confusion has been to dissipate the efficiency and creativity gains that decentralization promises. Although the Forest Service remains highly decentralized, [FN288] it has, of late, exhibited signs of paralysis more typically associated with rigid centralized bureaucracies. A recent General Accounting Office study concludes "that the Forest Service's decision‑making process is clearly broken and in need of repair." [FN289] More ominously, a cover story in the June 1997 Harper's portrays the Forest Service as the corrupt servant of a timber industry that lobbies Congress to provide large subsidies for logging in Forest Service lands, with expedients such as fire‑salvage and disease‑control used to justify clearcutting in forests that would otherwise be protected by environmental laws. [FN290]
These developments might be read as an inevitable consequence of Progressivism's faith that scientific management will produce a single, best solution, when in fact conflicting goals will be pursued by conflicting interest groups. [FN291] Whatever the merits of this critique as a general matter, however, it hardly applies to the particular case. In his administration of national lands, Pinchot pursued an approach nearly the opposite of this portrayal of Progressivism. Recognizing that conflicting goals and changing knowledge would render once‑and‑for‑all rules ineffective, he sought a corrigible system: Pinchot favored public ownership of lands, short‑term leases rather than long‑term leases or outright sales of resource‑ rich property, an obligation by leaseholders to exploit resources immediately as a means of preventing speculative occupation, periodically adjustable fees, and a ban on forms of use, such as overgrazing, that would lead to irreparable harm to the environment. [FN292] Moreover, anticipating the dangers of overlapping administrative authority, Pinchot advocated integration of all public land questions under the supervision of a single federal department. [FN293] Thus, the current crisis in public land management seems more nearly a consequence of disregard for Pinchot's *371 principles of corrigibility [FN294] than an indictment of other‑worldly technocratic optimism.
3. Nuclear Power Plant Safety.‑‑We turn now to a more recent and more nearly complete example of experimentalist administration. Recently, there has emerged a new system of benchmarking regulation of the nuclear power industry housed in the Institute of Nuclear Power Operations (INPO, or the Institute), founded and financed by the utilities themselves in 1979‑‑nine months after the Three Mile Island disaster‑‑to reduce the potential for catastrophic accidents in the industry. [FN295] A memorandum signed in October 1988 between INPO and the Nuclear Regulatory Commission (NRC) creates the framework for a "continuing and cooperative relationship" between the two "in the exchange of experience, information, and data related to the safety of nuclear power plants." [FN296] Under this agreement, the NRC, in effect, retains the formal authority to promulgate regulations, but either adopts the standards in training, maintenance, and other matters elaborated by INPO, or simply acknowledges best practices defined by the Institute without further formalizing them. [FN297] There are no civil or criminal penalties for noncompliance with INPO standards, [FN298] but INPO can suspend the membership of uncooperative utilities and has found means, as we will see, to ensure that the operating goals it sets are carefully respected. [FN299]
In practice, INPO's chief activities consist of pooling the industry's operating experience, establishing benchmarks that distill the lessons it contains, and then evaluating individual power plants according to their ability to meet the relevant benchmarks. [FN300] Operating information is gathered initially through the Significant Event Evaluation‑Information Network, or SEE‑IN. [FN301] This is "an industry‑wide effort to systematically collect, analyze, and share the industry's experience with safety‑related problems." [FN302] INPO officials sift SEE‑IN reports to distinguish harmless disruptions of operations from dangerous ones. [FN303] Thorough analyses of the causes of the dangerous disruptions, and ways of preventing them, are then circulated as Significant Operating Experience Reports, or SOERs. [FN304] Industry Operating Experience Reviews are then conducted periodically to assess the ability of particular plants to make effective use *372 of the information provided by SOERs and other means to improve their own affairs. [FN305] For purposes of this review, a team of specialists in a variety of areas evaluates the plant's troubles since the last INPO inspection, paying particular attention to the plant's own reports on how it has responded to SOERs. [FN306] As many as six of the inspectors operate as experts on loan from their companies. [FN307] The team spends a week preparing at INPO headquarters in Atlanta, and then spends two weeks of twelve‑ hour days doing "'nothing but watch[ing] what is going on at the plant."' [FN308] In particular, as in the case of error detection in firms, the inspectors are "always asking the 'Why?' question." [FN309]
The results of the report are made available to the CEO of the utility operating the power plant and to that utility's board of directors to ensure that criticism is not blunted as it passes up the managerial hierarchy. [FN310] The rankings naturally expose managers to intense peer pressure. In addition, INPO ranks all plants with respect to a number of summary measures of operating safety, and reports these rankings annually to a meeting of the utility CEOs. [FN311] These reviews, finally, are supplemented by continuing analysis of accident data and development of standards. [FN312] Participants in these activities are drawn from utilities, as are some fraction of INPO's full‑time staff. [FN313] In addition to providing information to the plants, results of INPO evaluations are communicated to the NRC, the federal agency responsible for monitoring reactor safety. [FN314]
As measured by two broad substantive indicators, INPO is an unqualified success story. The two measures are the number of "scrams," or rapid reactor shutdowns, and the number of safety system actuations. Both represent a gauge of the frequency of emergencies and are therefore inversely correlated with overall reactor safety. Between 1980 and 1990, the number of scrams per unit decreased by eighty percent, and the number of safety system actuations decreased by sixty percent between 1985 (the first year such measures were taken) and 1990. [FN315]
Still, one persistent criticism of INPO remains. Because it is not itself a government entity, INPO's activities are not subject to all of the usual responsibilities associated with such entities; most importantly, its reports *373 are not directly available to the public. [FN316] Thus, critics note that the public reports produced by the NRC present sanitized versions of INPO's data, thereby shielding the plants from public scrutiny. [FN317] On the other side of the question is the view that some degree of confidentiality facilitates full, open self‑criticism by the plants. In addition, given the public's likely reaction to even extremely small increases in the risk of the very large harm that a nuclear power plant accident would occasion, general release of INPO assessments might be misunderstood and misused by the public. This in turn would divert plant staff from safety improvements and other operations, causing them to pay excessive attention to public relations.
It is difficult to judge the merits of this debate at its most abstract. Surely there are contexts in which confidentiality breeds correctives. Public overreaction to potentially catastrophic harm, no matter how unlikely, has arguably distorted public policy in the past. [FN318] But it is a staple of democracy that in most contexts publicity encourages change for the better. Into which category particular aspects of nuclear power plant safety inspection fall strikes us as just the kind of question that can only be answered by experience and experiment.
But as regards the operation of INPO, now it seems that the criticism of secrecy is misleadingly narrow, while the Institute's secretive handling of information unjustifiably disparages the citizens' ability to deliberate about matters that concern them viscerally. The criticism is misleading because it slights the fact that INPO's reports do reach important watchdog groups, including not only the NRC, but also utilities (often themselves publicly accountable) and public service commissions. [FN319] Fears of potential conspiracy thus come down to the much more limited claim that citizens living near reactors have a need for direct access to all information. To provide anything less is not to flaunt unaccountability. But INPO's reticence seems disproportionate as well, for, as we will see next, public disclosure of environmental hazards comparable to those posed by nuclear power stations has often proved an effective instrument of regulation. [FN320]
4. Innovative Environmental Regulation.‑‑Many of the cases of precocious or emergent experimentalism discussed so far were set in settings *374 that combined bureaucracy and disorganization in ways inhospitable, even inimical, to innovation. If experimentalism has prospects of succeeding in such settings, we argued, it is likely to prosper almost everywhere else as well. The circumstances of environmental regulation, by contrast, combine aspects of locality and generality that invite experimentalist information pooling. [FN321] On the locality side, it is necessary to take full account of local topography, wind conditions, and economic activity to determine the exposure of a particular population to various environmental risks. Effective reduction of such risks requires full attention to the local complexities of industrial production, sewage disposal, power generation, or traffic patterns from which they arise. On the generality side, the epidemiological determination, say, that a substance is toxic or otherwise hazardous, and in what concentrations, requires the pooling of evidence from many different and dispersed sources. The same sort of information pooling is necessary for estimating feasible reductions in particular environmental harms and for ascertaining the most effective methods of achieving such reductions.
The incentives for mutual learning and monitoring are particularly strong in environmental matters, moreover, given the poorly confined or even unconfined nature of hazards such as air or water pollution. For example, coal‑burning plants in the Midwest produce acid rain in the Northeast and Canada. In the worst case, one locale's indifference to its environment puts all the others at terrible risk. The existing regulatory regime does more to obstruct than to further such learning. It typically sets fixed emissions standards for particular processes, and divides responsibility for controlling pollution in the air, water, and other media to different regulatory authorities. Obtaining permits for the processes in use at any production site therefore requires a laborious round of agencies, and abatement of pollution at the prompting of authorities in monitoring one medium can easily lead to additional problems in areas where there is less vigilance. [FN322] If public and private actors were not drawn to forms of democratic experimentalism to reconcile the needs of local and global learning under these conditions, they would be unlikely to give them much consideration anywhere else.
We conclude this subsection, therefore, by noting the recent spread of forms of self‑, state, and federal environmental regulation whose experimentalist features may be cohering into a system of learning by monitoring by a peer inspectorate along the lines we saw in our discussion of INPO. The prospect of such an outcome is especially striking because it arises in part from systematic public disclosure of alarming, potentially catastrophic hazards: the same kind of information, we just saw, sometimes *375 said to obstruct reform by private actors when revealed to the public. We begin with a review of the innovative, if barebones, national system of benchmarking self‑ reported releases of toxic substances‑‑the Toxic Release Inventory (TRI)‑‑and the efforts at self‑regulation it has provoked on the part of the chemical manufacturers. Then we consider the Massachusetts Toxics Use Reduction Act of 1989 [FN323] (TURA), which, as the most developed of several similar state regimes, requires not only that firms report their use of certain toxic substances, but also that they formulate and periodically revise plans to reduce that use in consultation with a peer inspectorate created for this purpose.
Finally, we examine the Program for Regulatory Excellence (XL) and the Common Sense Initiative (CSI) of the federal Environmental Protection Agency (EPA). XL is a pilot project that waives many current permitting requirements. It allows firms greater flexibility in defining the precise means and ends of their regulatory performance in exchange for commitments to improve on that performance and commitments to provide sufficient information to facilitate monitoring their actual behavior. CSI invites proposals for regulatory reform of this general kind‑‑some eventually requiring waivers, others not‑‑from teams of stakeholders in six industrial sectors. But the limits of XL's authority to authorize experimentation are vague, as are the conditions under which proposals for reform formulated under CSI can actually be tested in practice. Neither project makes provision for benchmarking or any other form of information pooling by which the EPA, together with the affected actors, might devise standards for judging the suitability of revisions of current rules. These deficiencies explain why, so far, XL counts many more failed negotiations than successful ones, while CSI promises reforms without realizing them. Together TRI, TURA, and the federal programs suggest additional, complementary lessons about the possibilities and limits of the incremental introduction of experimentalism in our federalist polity.
a. TRI.‑‑TRI was created under the Emergency Planning and Community Right‑to‑ Know Act of 1986. [FN324] The Act and TRI were, like the creation of INPO after Three Mile Island, a response to catastrophe: in this case, the explosion in 1984 of a Union Carbide facility in Bhopal, India that killed many thousands of persons, mostly through release of toxic methyl isocyanate gas. [FN325] Unlike the familiar regulatory regimes defined by the Clean Air Act [FN326] or the Clean Water Act, [FN327] the Right‑to‑Know legislation neither fixes targets for the reduction of aggregate levels of pollution, nor requires specific pollution‑abatement efforts by particular *376 classes of polluters. Rather, in establishing TRI, the Act requires only that private and government‑run facilities meeting statutory size requirements report estimates (calculated by EPA methods) of the amounts of some 650 chemicals they transfer off‑site, or routinely or accidentally release. [FN328] Since passage of the Pollution Prevention Act of 1990, facilities must also report transfers within the plant and efforts at pollution reduction and recycling. [FN329] These reports are then made publicly available in both raw form and as tables comparing amounts released by substance, facility, industry, and location. [FN330] In addition, since 1989, the EPA has published an annual summary of emissions, with a comparison to previous years. Failure to file a report as required by the Act may result in penalties, but inaccuracies in reporting do not. In fact, the EPA does little to verify the accuracy of emissions reports, and has no inspection or other enforcement authority directly related to TRI. [FN331] Citizens, however, may sue firms for failure to comply with TRI's disclosure provisions, [FN332] and the data obtained can then be used to establish violations of other, substantive statutory obligations, or as a lever by which to apply public pressure for improvements. TRI is thus environmental "regulation" in the minimal sense of formally requiring disclosure of a body of comparative information from which environmental rules and standards, fixed or rolling, might eventually be fashioned or enforced. [FN333] Its operation therefore constitutes a rough test, under admittedly favorable circumstances, of whether benchmarking in general‑‑and benchmarking of "alarming" information in particular‑‑can play the central role that we *377 have attributed to it in guiding and synchronizing performance‑improving efforts and rules to encourage these efforts.
Three results suggest that it can. First, the collection and publication of TRI data immediately discipline polluting private actors: Public comparisons of polluters compiled by journalists or community activists from TRI data lead to significant declines in the share value of publicly traded firms that show poorly. [FN334] Some of this decline might simply reflect investors' fears that bad publicity, however unfounded, always means costly trouble. But it is at least as likely to reflect a cold‑eyed calculation that facilities listed among the worst polluters of their locale or type are not well controlled by managers, and may therefore suffer from problems not directly connected to pollution as well. Once it is clear that a poor pollution ranking leads to costly penalties in financial markets, and from there to clean‑up expenses, managers have strong incentives to avoid the costs either by reducing environmental burdens in advance of the disclosure or (given negligible penalties for inaccurate reporting) shading their estimates of toxic releases to obscure their true situation. Investors might reasonably conclude that management that did neither was simply unaware of the extent of the problem in comparison with other facilities. This conclusion might just as reasonably prompt the more general concern that management might be ignoring problems unrelated to pollution as well. In any event, details of the connection between finance‑market discipline and TRI data aside, commentators agree that "public release of information about discharge of toxic chemicals has by itself spurred competition to reduce releases, quite independently of government regulation." [FN335]
Beyond this immediate discipline, Bhopal and TRI have had second‑order effects on the behavior of the actors by inducing changes in industry associations and the information they pass among firms. Where Three Mile Island led to INPO, Bhopal and TRI have led to the creation, within the Chemical Manufacturers Association (CMA) of a "Responsible Care" program, launched, as the head of the Association confessed, because "the industry had no choice." [FN336] This program encourages firms: 1) to link pollution‑ prevention efforts to the core disciplines of error‑detection and elimination which they apply to managing their production processes; 2) to involve suppliers and distributors in these expanded efforts at continuous improvement; 3) to set target dates for installing *378 these new disciplines; 4) to monitor progress towards their goals; and 5) to document and disseminate best practices. [FN337]
In theory, then, Responsible Care could in time become the armature of monitored self‑monitoring with a family resemblance to the INPO system. Assessment of the gap between this theoretical possibility and current practice is, however, impossible based on currently available evidence. On the one hand, certain associations, such as the Society of Organic Chemical Manufacturers, make compliance with Responsible Care disciplines within fixed time limits a condition of membership, [FN338] and at least some companies‑‑ Union Carbide, to take a chastened and chastening example [FN339]‑‑publicly report progress towards the Responsible Care standards as well as performance by TRI measures. On the other hand, the trade associations do not seem to be organizing the kinds of practices on which INPO depends: systematic inspections of facilities, comprehensive reporting on hazardous disturbances, and analysis of countermeasures taken by the facilities. Instead, the CMA, for example, has thus far emphasized self‑monitoring by firms and minimized the need for any systematic review of these results, except by such amorphous entities as company‑appointed Citizen Advisory Panels drawn from the communities in which participating members operate. Environmentalists plausibly suspect that these panels may be denied relevant information and dominated by company interests. [FN340] Peer pressure, public opinion, or the discovery of further benefits to pollution‑control measures could of course lead to more rigorous benchmarking of problems and countermeasures, but this outcome is no more automatic than was the adoption of the Responsible Care program in the first place.
Third and concurrently, as the EPA itself has noted, in making possible comparisons across regions and facilities, the release of TRI information has allowed federal, state, and local governments to cooperate with the public and industry to "evaluate existing environmental programs, establish regulatory priorities, and track pollution control and waste reduction progress." [FN341] At the state level above all, the boldest of this collaborative redirection of regulatory activity has gone to the creation of just the kinds of review and information‑pooling services which even the *379 largest trade associations have not (yet) undertaken to provide, and which may always be beyond the reach of industry associations in sectors, such as injection molding or metal plating, characterized by small, dispersed facilities. Partly by complementing, partly by competing with (and thus forcing emulation by) industry associations in the provision of services, programs in states such as Massachusetts, Oregon, New Jersey, Washington, and Minnesota are likely to shape the emergent regime of environmental regulation and accentuate its experimentalist features. [FN342]
b. TURA.‑‑One of the most established, comprehensive, and influential of such programs was created by the Massachusetts TURA of 1989. [FN343] Because of the scope and integrity of this founding Act, the relation among the parts of the Massachusetts regulatory system and their connection to federal benchmarking are particularly clear. We turn briefly to TURA and its operation, therefore, for a first impression of one variant of a more developed experimentalist regime of environmental regulation.
As an initial matter, TURA broadens and extends TRI. It broadens TRI by requiring firms to report not only toxic releases, but also use or generation of toxics in any stage of production. [FN344] TURA further requires that the reports on use of toxics be connected to plans for usage reduction. [FN345] Thus, under the Act, firms meeting statutory requirements regarding size and line of business must annually file a Toxic or Hazardous Substance Report listing the amounts (in excess of certain minima) of designated toxics used as inputs to processes, generated as byproducts, or shipped as end products. In addition, the Report must estimate changes in the amount of toxic byproduct and emissions generated per unit of product as compared with the preceding year, and specify whether the changes were due, for example, to altered inputs, new production processes, improved operations, reformulation of the product, recycling or other extension of the usefulness of toxic substances, or other modifications of the manufacturing setup. [FN346] These reports then form the starting points for biannual Toxics Use Reduction Plans centered on "a comprehensive economic and technical evaluation of appropriate technologies, procedures and training programs for potentially achieving toxics use reduction for each covered toxic or hazardous substance." [*380 FN347] On the basis of this benchmarking survey of possibilities, firms specify in the Plan particular measures to be adopted, the schedule for implementing them, and two‑ and five‑year targets for toxic use reduction. [FN348] Although TURA establishes the general goal of reducing use of toxics in Massachusetts by fifty percent by 1997, [FN349] and penalizes "willful" violations of the requirements to report toxics use and plan for reduction, [FN350] the Act sets no more specific performance standards nor does it penalize failure to act on reduction plans. Thus, rather than fix objectives, and compel their attainment, TURA furthers the TRI strategy of using the obligation for self‑monitoring to induce firms and citizens to acquire information that reveals problems and possibilities for their solution. TURA provides an additional possibility for citizen involvement by providing a right of citizens to sue to have firms comply with the Act's largely procedural requirements. [FN351]
At the same time, TURA extends and helps formalize industry efforts at improved environmental performance both by creating a peer inspectorate to review the usage reduction plans, and by enabling provision of ancillary technical consulting services to aid firms in their formulation. [FN352] The peer inspectorate is created by requiring that the accuracy of Plans be certified by a toxics use reduction planner, then by making completion of various training programs a condition for initial or renewed certification as a planner. Two possibilities are anticipated. The first is that the prospective planner complete satisfactorily a comprehensive program in toxics use reduction planning organized by the Commonwealth. In that case, the planner may certify the reduction plan of any firm submitting one. The second option is for the planner to demonstrate, for purposes of initial certification, two years of practical experience in toxics use reduction in a particular firm. In that case, the planner can only certify plans of his or her employer. The training requirement of the two paths, and the experiences of the two types of planners, converge, however, as participation in continuing‑education classes in the subject is a condition for re‑certification after two years in both cases. [FN353] The Act accordingly *381 establishes a Toxics Use Reduction Institute (TURI) at the University of Massachusetts at Lowell to develop the curricula and provide the courses required for certification or re‑ certification as a planner, or generally to inform industry or the public of developments in this area, and to conduct research necessary to these activities. [FN354] It also establishes an Office of Technical Assistance (OTA) to assist firms (particularly small, first‑time filers) in meeting their TURA obligations, and to help coordinate the provision of relevant services by the public and private sectors. Confidential information obtained by the OTA in the course of consultations with a firm is not reported to the Commonwealth Department of Environmental Protection unless the firm agrees, or unless the information concerns an imminent threat to public health. [FN355] The training of planners, submission of plans, and provision of technical services, furthermore, is coordinated with existing reporting and inspection regimes to minimize the burden of regulatory compliance. [FN356] Taken together, plans, planners, TURI, and OTA create an inspection regime in which current conditions in individual firms or industrial segments can be compared with each other and with academic understanding of best practices, even as that understanding is corrected by scrutiny of innovation in firms.
Finally, applying the pragmatist principles of adjustment of means and ends to the institutions created by the Act itself, TURA provides a high‑level governance structure that periodically suggests modifications of the new state services and reporting requirements in the light of its evaluation of progress towards the Act's original reduction target. [FN357] An Administrative Council, consisting of state government officials with responsibilities in the environmental area, presents an annual review of progress towards the overall goal and suggests possibilities for improvements in, and better coordination [FN358] of, the programs concerned with toxics use reduction. [FN359] An Advisory Board composed of two officials and thirteen representatives of environmental, business, and other interested groups, creates a forum for users of the program to comment directly on its operation and to create, if necessary, ad hoc committees to recommend changes. [FN360] The Act is self‑ financing in the sense that the costs of training inspectors and providing other services are to be paid by the proceeds of a "toxics use fee" on the large toxics users subject to it, [FN361] and the governance bodies can easily monitor the relation between costs and revenues.