Bridging the Data Gap: Balancing the Supply and Demand for Chemical Information
By Applegate, John S
I. Introduction Since the beginning of serious environmental regulation in the 1970s, the United States (and later Europe) has increasingly, and seemingly inexorably, adopted a risk-based approach to the regulation of environmental threats and, in particular, of toxic chemicals and pesticides. In these regulatory systems-the Toxic Substances Control Act’ (TSCA); the Federal Insecticide, Fungicide, and Rodenticide Act2 (FIFRA); and the toxics provisions of general pollution statutes3-the extent of regulation is primarily determined by a highly quantitative evaluation of the toxic potency of a chemical and the degree of human exposure, resulting in quantitative assessment of the likelihood of harm to human health. An aggressive and comprehensive program of chemical regulation is severely restricted by the information demands of this approach, however, because a wide “data gap” exists between the information required to justify regulation and the information that is actually available.
This Article has three objectives and three central Parts. First, it traces the rise of risk as the basis for chemical regulation, distinguishing between two different meanings and functions of risk in chemical regulation. second, it demonstrates how the adoption of particular risk-based strategies can result in a severe imbalance between the demand for and the supply of chemical information.4 Third, building on the different functions and meanings of risk, it offers a fundamentally different approach to chemical regulation, which reduces the demand for chemical information (bridging the data gap, so to speak) rather than attempting the ultimately unattainable goal of completely supplying it (filling the gap). This Article is thus a variation on the Symposium theme of “harnessing the power of information”: It advocates harnessing the demand for information as the most promising way to make the existing and foreseeable supply of chemical information more effective in supporting a robust program of protective regulation.
II. The Rise of Risk
Risk has a lengthy history in the legal response to damage to persons and property, and over time different legal frameworks have emphasized different functions and meanings of risk.5 Before tracing the conceptual development of risk, therefore, it will help to distinguish the functions and meanings. In order to be effective, or even coherent, a regulatory system must describe both the preexisting conditions that call for regulatory attention and the future conditions that regulatory action is supposed to achieve. The “before” and “after” states of the world can be called the trigger and standard, respectively.6 The trigger defines the problem, and the standard describes the appropriate regulatory response. Thus, under the Clean Air Act,7 if the Environmental Protection Agency (EPA) finds that a facility is emitting a hazardous air pollutant (the trigger), then it must apply restrictions equivalent to the maximum achievable control technology (the standard).8 Under the Delaney Clause,9 if the Food and Drug Administration (FDA) finds that a nonpesticide chemical causes cancer in man or animal (trigger), then it must prohibit the use of the chemical as a food additive (standard).10
As we will see, risk has been used for both of these functions in chemical regulation, that is, as the definition of the problem to be solved and as the means of disciplining or constraining the response to the identified risk. Risk is able to perform these dual functions because it has two distinct lexical meanings. According to Merriam- Webster’s, risk means both “possibility of loss or injury” and “the chance of loss or the perils to the subject matter of an insurance contract; also: the degree of probability of such loss.”" The first has an open-ended meaning roughly equivalent to “danger” or “threat.” The second meaning is probabilistic and implies greater precision. It suggests a quantitative approach to describing the extent of the likelihood of a harm occurring.12
A. Defining the Problem
The default or baseline legal response to physical harm is the tort system, which provides a remedy for those who have suffered harm due to someone else’s actions. The tort system is almost entirely retrospective. It reduces the number or severity of future harms only indirectly, if at all, by threatening that “negligent” behavior will be taxed with the cost of the harm it causes.13 Negligence can be measured in several ways, such as reference to “reasonableness” or to industry custom.14 Another technique, which has become the foundation of the law-and-economics approach,15 uses Judge Learned Hand’s famous B < P ? L formula: a defendant is negligent when the burden of taking precautions would have been less than the expected value of the harm (that is, the loss discounted by its probability).16 Hand's formula in effect defines negligence as an unreasonably high risk, and under this approach a defendant who exposes others to such a risk should expect to pay for harm caused thereby.
Environmental legislation can be understood as a response to a number of important limitations of the tort system in the environmental setting.17 First, whereas tort law seeks to repair or compensate a harm that has already occurred, the raison d’etre of environmental regulation is the prevention of harm before it occurs. The court in Ethyl Corp. v. EPA.5 explained, “Regulatory action may be taken before the threatened harm occurs; indeed, the very existence of such precautionary legislation would seem to demand that regulatory action precede, and, optimally, prevent, the perceived threat.”19 Prevention is also preferable to repair because many of the things that tort law compensates-lives, health, suffering, lost time-cannot be restored with money, or at all. Avoidance is more humane and usually less costly than restoration. As the Council for Environmental Quality (CEQ) stated, “We need no longer be limited to repairing damage after it has been done; nor should we allow the general population to be used as a laboratory for discovering adverse health effects.”20
Second, being retrospective and reparative, tort law demands a high standard of proof of causation of the individual plaintiffs harm.21 While this is a serious obstacle to proving toxic causation,22 it is a defensible policy in the great majority of tort cases in which the harm, having already occurred, can be traced back to the defendant’s conduct in a deterministic manner. To achieve a preventive objective, however, regulation must be anticipatory in the sense that the degree of certainty of causation must be relaxed to permit prediction in advance of actual harm. Moreover, “some of the questions involved in the promulgation of [environmental] standards are on the frontiers of scientific knowledge, and consequently as to them insufficient data is presently available to make a fully informed factual determination.”23 The precautionary principle, a nearly ubiquitous feature of recent international environmental agreements, embodies the anticipatory approach by rejecting demands for “full scientific certainty” when faced with “threats of serious or irreversible damage.”24
Third, tort law permits (indeed, encourages) risk creators to reach their own economic decisions about the level of risk that they are willing to impose on others, because the tort mechanism of monetary damages operates as an incentive system rather than as a set of specific commands for specific actions. Potential defendants are free to choose to risk paying damages, and this aspect of the tort system is often praised as permitting efficient decision making by potential defendants.25 However, as Judge Calabresi observed many years ago, there are also situations in which society may legitimately decide that a collective determination of the level of safety (the inverse of risk) is appropriate-situations in which normative concerns for avoiding harm trump economic efficiency.26 Regulatory commands, unlike tort law, do not typically offer the option of choosing to harm.
Risk came to define the regulatory problematique because it elegantly defines all three departures from the tort paradigm. A preventive system can be founded on risk because risk is the forward- looking description of as yet inchoate harm. It is no coincidence that the CEQ followed its description of the need for prevention, above, with advocacy of a statute that prevented “unreasonable risk,”27 nor that the Ethyl Corp. court spoke of the need to “assess risks” as the antithesis of “a high quantum of factual proof, proof of actual harm rather than of a ‘significant risk of harm.’”28 Likewise, risk is anticipatory in that it does not require certainty of actual outcomes in a deterministic sense. Last, risk reflects a legislative judgment that dangerous activities are, if not entirely unacceptable, at least subject to collective control.
This use of risk functions as a trigger for regulatory action because it defines the universe of governmental concern and tells regulators what requires their attention. In performing this particular function, risk does not establish standards for determining how much danger is desirable or acceptable. Instead it is a binary concept in which a substance or activity is either safe or unsafe. This meaning of risk is embodied in, for example, the requirement to set emissions levels for toxic air pollutants that are safe with “an ample margin of safety to protect public health,”29 which is a coherent standard only if safety is an on-or- off, yes-or-no proposition. This view of risk has been dubbed “hazard,”30 and it reflects the first definition of risk quoted above. B. Disciplining the Response
Risk-as-hazard is a very open-ended criterion for regulation, because most chemicals have some harm-causing potential. In fact, our general experience of the industrial world is that hazards abound, and we navigate the world with the aid of a rough sense of the degrees of danger. The rise of the negligence standard in the nineteenth century was itself an acknowledgement that a risk-free industrial society is unobtainable.31 That is certainly the lesson that EPA and other federal agencies drew as their experience with binary statutory commands increased.
In addition to the everyday sense that a risk-free world is impossible, emerging understanding of the mechanisms of cancer produced a parallel conclusion for toxic substances.32 Environmentally induced cancer was found to be a probabilistic rather than deterministic phenomenon, caused by-to simplify greatly- the interactions of individual molecules of certain substances within individual cells.33 One cannot know in advance whether a particular exposure will cause cancer in a particular individual, but one can say that the greater the exposure, the greater the odds of cancer occurring. Moreover, under the conservative “one-hit” model of carcinogenesis, carcinogens have no threshold level of exposure below which the probability of injury drops to zero.34 The only exposure that can definitively be called safe under this model is no exposure at all. Under these conditions, as an EPA deputy administrator put it, “Many EPA professionals no longer used bi- modal terms such as safe or unsafe, but rather began to think and talk in probabilistic terms.”35 The question was no longer the existence of a risk, but its extent-and ultimately its acceptability- along a spectrum of likelihood of harm that ranges from zero, to small and acceptable levels of risk, to unacceptably high levels of risk, to certainty of harm.
Because risk-as-hazard is open-ended, it implies relatively draconian regulatory responses to the existence of a risk: if a risk exists, it should be eliminated. Not surprisingly, this use of risk engenders enormous counterpressure to constrain its application. Substantive discipline is exercised, in part, through the establishment of a postregulation goal or standard that tells an agency how to respond to the universe of problems that has been defined for it.36 Before 1990, [section] 112 of the Clean Air Act established a trigger for regulating hazardous air pollutants: “may reasonably be anticipated to result in an increase in mortality or an increase in serious irreversible, or incapacitating reversible, illness.”37 For any air pollutant so identified, EPA was required to set an emission standard that “provides an ample margin of safety to protect the public health.”38 Since true protection with an ample margin of safety could only be a ban on emissions (or something very close to it),39 EPA issued a mere seven air-pollution standards in the twenty years of the pre-1990 section’s existence.40 Congress, in other words, defined the problem broadly but failed to discipline the response sufficiently to make regulation politically feasible for the regulatory agency.41
For chemical regulation, the probabilistic meaning of risk provided the perfect conceptual framework for reorienting risk to discipline the response. As probabilistic thinking took hold, risk was available to express not simply the potential for harm, but also the extent of the potential harm. This development took place initially when FDA tried to cope with the Delaney Clause. It is the apotheosis of the risk-as-hazard paradigm, because it flatly bans as a food additive any substance that causes cancer in humans or animals at any level of potency or exposure.42 While FDA never fully succeeded in escaping the binary nature of the Delaney framework,43 it worked with other agencies to develop a new probabilistic mechanism for measuring the spectrum of risk: quantitative risk assessment.44
In the process described in the 1983 National Research Council report, Risk Assessment in the Federal Government: Managing the Progress45 (universally known as the Red Book, for its cover), the first step of quantitative risk assessment is the determination of risk-as-hazard-”hazard identification.”46 The Red Book procedure goes on to estimate the potency of the substance (“dose-response assessment”)47 and the level of actual exposure (“exposure assessment”).48 The resulting product of dose-response and exposure (“risk characterization”) tells the regulator quantitatively where on the spectrum of danger a particular chemical or activity lies.49 Furthermore, the Red Book separates the foregoing steps, known collectively as risk assessment, which is regarded as an objective and scientific process, from risk management, which is a judgmental and essentially political process.50 The assessment-management distinction underscores the ideas that risk can be a relative concept, and that the regulatory response to risk can and should be informed and constrained by economic and political concerns.51 Thus, the need to discipline risk created the need for an appropriate analytical tool,52 and the development of the tool reciprocally encouraged the use of risk to discipline regulation.53 Risk became risk assessment.54
The Red Book structure demonstrates that risk can serve two regulatory functions: risk-as-hazard acts as the trigger, and risk- as-probability acts as the standard. Risk-as-hazard creates a potential for extremely wide-ranging agency action to control inchoate harms, and risk-as-probability applies constraint by making room for political and policy considerations. Risk-ashazard defines the problem, and risk-as-probability disciplines the response. This framework is summarized in the following table:
Table 1: Regulatory Functions of Risk
C. Definition and Discipline in the Courts
To put the foregoing framework into more concrete terms, the shift from risk-as-hazard defining the toxics problem to risk-as- probability disciplining the regulatory response is evident in the transformation of the judicial approach to toxic harm.
1. Reserve Mining and Ethyl Corp.-The earliest cases on the regulation of toxic substances emphasized the change wrought by the antipollution statutes. The congressional response to the inadequacies of tort law was “precautionary” legislation, embodied in the key term “endangering,” which the courts held was used “in a precautionary or preventive sense, and, therefore, evidence of potential harm as well as actual harm comes within the purview of that term.”55 The courts specifically rejected adoption of the tort standard of “probable” harm.56 Instead, they embraced risk-as- hazard: “Danger is a risk, and so can only be decided by assessment of risks.”57 The context makes clear that “assessment of risks” in these cases does not imply quantitative risk assessment, a term that had not yet been coined:
Where a statute is precautionary in nature, the evidence difficult to come by, uncertain, or conflicting because it is on the frontiers of scientific knowledge, the regulations designed to protect the public health, and the decision that of an expert administrator, we will not demand rigorous step-by-step proof of cause and effect.58
The cases contrast “assessment of risks” with deterministic, retrospective proof of harm.59
The existence of a danger in these cases is an open-ended trigger for regulatory action, and there is no inherent limiting principle. In Reserve Mining,60 the court ordered the elimination of all asbestos discharges into Lake Superior because the mere presence of asbestos in drinking water posed a risk-as-hazard.61 In Ethyl Corp., the court clearly recognized that the appropriateness of agency action (to phase out lead additives in gasoline) would depend on “a lesser risk of a greater harm [or] a greater risk of a lesser harm,”62 but offered nothing to define the risk level further. Effective discipline, in other words, had to come from other sources. In Reserve Mining, the open-endedness was constrained by the court’s own ultimate control of the injunctive remedy: “We are fortified in this view [of the broad meaning of 'endangering'] by the flexible provisions for injunctive relief which permit a court ‘to enter such judgment and orders enforcing such judgment as the public interest and the equities of the case may require.’”63 In Ethyl Corp., the court relied on the congressional investiture of regulatory authority in expert agencies as the appropriate constraint on the governmental response,64 ultimately supervised by the collaborative relationship of courts and agencies espoused by the D.C. Circuit in this period.65
2. Benzene.-The Benzene case66 is universally acknowledged to be the turning point in the judicial approach to the regulation of toxic substances.67 The Supreme Court was squarely faced with the difficulties of regulating a nonthreshold carcinogen under a statute whose language implied a binary model of safety.68 The Court’s plurality declared that “‘safe’ is not the equivalent of ‘risk- free,’”69 and that the Occupational Safety and Health Administration (OSHA) must predicate its regulations on a well-supported finding of the existence of a “significant risk.”70 The Benzene decision specifically rejected the agency’s case for regulation based solely on evidence of benzene’s carcinogenicity, i.e., risk-as-hazard.71 Instead, the plurality insisted that the finding of a “significant risk” be based on preexisting exposure levels,72 which implicates the dose-response and exposure assessment elements of risk assessment. Not surprisingly, despite the plurality’s protestations to the contrary, the opinion is universally understood-based in part on the plurality’s own use of numerical risk to circumscribe the meaning of “significant”73-to require agencies to develop a system of quantitative risk assessment.74 For the Benzene plurality, the open-ended nature of risk-as-hazard represented a danger of governmental overreaching rather than an opportunity to better protect the public. While Ethyl Corp. emphasized the expansiveness of agency authority to deal with novel problems,75 the Benzene plurality was plainly convinced that OSHA had used its powers improvidently and needed to be reined in.76 The opinion pointedly (and otherwise irrelevantly) observed, “As presently formulated, the benzene standard is an expensive way of providing some additional protection for a relatively small number of employees.”77 Quantitative risk assessment, then, served as the plurality’s method for disciplining the agency response. The concurring justices went even further: Justice Powell argued for the application of a cost- benefit test to constrain the agency,78 and then-Justice Rehnquist concluded that the congressional risk-as-hazard standard was so open- ended as to be unconstitutional under the nondelegation doctrine.79
3. Gulf South and Corrosion Proof Fittings.-The logical extreme of using risk-as-probability to discipline the agency response was reached in a pair of Fifth Circuit cases that subjected agency decisions on toxic substances to withering judicial scrutiny. In Gulf South*0 the court rejected the Consumer Product Safety Commission’s ban on urea-formaldehyde foam insulation.81 The court criticized the design of the agency’s main study, the study’s applicability to the exposure scenario, and the default assumptions underlying the resulting risk assessment.82 The judges concluded that the agency’s use of the evidence before it was “not good science.”83 Moreover, even if the studies were usable, they did not meet the standard of proof that the court expected of the agency: “To make precise estimates, precise data are required.”84
Corrosion Proof Fittings*5 takes the Gulf South approach still further in the context of a proposed EPA ban on asbestos.86 The court found in TSCA’s “least burdensome” language a principle that the more stringent the regulation is, the greater the degree of proof required to justify it.87 Within this already challenging structure, the court found fault with “the manner in which the EPA conducted some of its analysis” and “some of the methodology employed by the EPA in making various of the calculations that it did perform,” criticizing among other things the extent of reliance on “unquantified benefits” and the degree of reliance on population exposure.88 The court remanded the asbestos ban, in effect killing it.
Gulf South and Corrosion Proof Fittings are unquestionably outliers in the aggressiveness of their use of risk-as-probability to discipline agency action.89 However, they do have their defenders,90 and they have unquestionably shut down the regulation of toxic chemicals under the Consumer Product Safety Act and TSCA.91 They exemplify both the transition from an open-ended, risk-as- hazard approach to the problem of chemical regulation, in which risk broadly defines the problem and acts as the trigger for regulatory action, to a far more constrained approach that is characterized by risk-as-probability serving as the standard that disciplines the agency’s response.92
III. Risk and the Data Gap
The transformation of the legal approach to toxics from an open- ended risk-as-hazard trigger to a constrained risk-as-probability standard has major consequences for the information needs of the regulatory system for chemicals. Risk-as-probability requires that regulatory action be supported by extensive amounts of toxicity and exposure information-information that simply does not exist. This data gap must be remedied if a program of chemical regulation is to be effective.
A. Risk-Based Regulation and Information Demands
As we have seen, the Red Book divides the risk-analysis process into the objective, scientific “risk assessment” phase, and the judgmental, policy-oriented “risk management” phase.93 Risk assessment consists of two main elements: the toxic potency (dose response) of a chemical that has been identified as causing harm, and the level of exposure to the chemical.94 Both elements are capable of quantitative expression, and quantification is expected. While the basic risk equation (toxicity x exposure = risk) is simple, its components are extremely complex phenomena.95 The data needs of the first step, hazard identification, are by far the least burdensome because they are limited to determining whether the substance has the capacity to cause a particular disease, typically cancer.96 Dose-response testing requires far more extensive and expensive studies,97 and the uncertainties are, in fact, magnified in such studies because of the need to extrapolate detailed toxicological information from animal test results.
The uncertainty encountered on the toxicity side of the risk equation has been described as “knowledge uncertainty,” because it is mainly based on the lack of a thorough understanding of the underlying mechanisms of cancer.98 The uncertainties associated with the exposure side represent “information uncertainty,” because the problem is less one of understanding than of the sheer extent of the data needs.” Exposure assessment requires the risk assessor to measure or estimate the concentrations of a chemical from the source or sources from which it is initially introduced into the environment through the relevant pathway or pathways (including air, water, groundwater, soil, sediment), to the target organism through the relevant route or routes of exposure (including ingestion, inhalation, dermal absorption).100 The problem of filling the data needs of the elements of quantitative risk assessment is exacerbated by important qualitative aspects of the analysis.101 Given the expectation of quantification that is implicit in Benzene and explicit in the Red Book, the scientific norms of certainty, precision, and comprehensiveness are woven into the fabric of the inquiry. More generally, science has an undeniable cultural and political attraction in justifying expensive regulation.102 Thus, the Gulf South court stated, “To make precise estimates, precise data are required”103-even though there is no inherent reason why we should value precision or pinpoint accuracy in the regulation of public health and the environment. The Ethyl Corp. court more realistically observed: “[C]ertainty in the complexities of environmental medicine may be achievable only after the fact, when scientists have the opportunity for leisurely and isolated scrutiny of an entire mechanism. Awaiting certainty will often allow for only reactive, not preventive, regulation.”104 In fact, certainty and precision are simply not available from the risk assessment techniques or the current state of the sciences behind them.105
To make informational matters worse, the risk-as-probability approach was encouraged by, and in turn facilitated, the approach to regulation that Professor McGarity has called “comprehensive analytical rationality.”106 Comprehensive rationality and all so- called rational-choice methodologies aspire to gather and systematically analyze all relevant data about a problem and its potential solutions in an effort to determine the optimal regulatory response.107 Thus, in addition to the full battery of chemical information already described, the risk-as-probability model serves the enterprise of determining all of the costs, benefits, and other consequences of a proposed regulatory initiative and of alternative measures. These efforts not only have their own information demands and uncertainties, but they place additional back-end pressure on the scientific data to be complete and precise, because the data provide the quantitative foundation for the benefits side of cost- benefit analysis.108
In sum, the use of risk-as-probability to discipline regulatory responses demands the acquisition of highly sophisticated scientific information that is neither readily available nor easily and cheaply obtained; some is at the frontiers of scientific knowledge, and some is at the outer limits of the practical ability to collect adequate amounts.109 Risk-as-probability commits the regulator to an extraordinarily expensive and time-consuming analytical effort to justify restrictions on chemicals.
B. Documenting the Data Gap
It seems intuitive that a substantial data gap would follow from the foregoing analysis because it is hard to imagine how the regulatory system could supply all of the information that the risk- as-probability approach demands. Indeed, lack of chemical information was one of the principal reasons for enactment of TSCA in 1976;110 the problem had been identified by the CEQ as early as 1971.’” The data gap has been verified in numerous empirical studies.”2
In 1984, the National Research Council systematically examined the available data in a report, Toxicity Testing.^13 Sampling the universe of commercial chemicals over a range of uses and production volumes, the panel found that no toxicity testing was available for more than 80% of all toxic substances in commerce, and that only 22% of high production volume (HPV) chemicals had even a minimum data set available.114 As this chart summarizing the Toxicity Testing results shows, there is a consistent data gap across all categories of industrial chemicals, regardless of production volume.115
Table 2: Summary of Toxicity Testing Results The situation has changed little since 1984. Using the baseline Screening Information Data Set (SIDS) for chemicals,”6 the Environmental Defense Fund (now Environmental Defense) published a study in 1997, Toxic Ignorance,111 which found that full SIDS data were publicly available for only 29% of the 100 HPV chemicals (greater than 1,000,000 Ibs/year) in their sample.”8 Conversely, none or only part of the SIDS mammalian-toxicity data set was publicly available for 71% of the sample.”9 In response, the Chemical Manufacturers Association (CMA, now American Chemistry Council) undertook its own study. The similarities in results are more striking than the differences. CMA concluded that complete SIDS data exist for 47% of chemicals,120 though others have read the study to show that only 6% of HPV chemicals have publicly available data for all SIDS elements.121 In an effort to sort this out, EPA undertook its own study of approximately 3,000 HPV chemicals. The agency summarized its findings: “[N]o basic toxicity information … is publicly available for 43% of the high volume chemicals manufactured in the US and a full set of basic toxicity information is available for only 7% of these chemicals.”122 A more recent study of EPA’s own Integrated Risk Information System (IRIS) by the Center for Progressive Reform123 found major gaps and outdated information in this database on a wide range of chemicals regulated under EPA’s statutes.124 And some areas, such as children’s health, are in even worse shape.125
The European Commission sponsored several studies of the data gap in preparing its legislative proposal for REACH, a complete overhaul of the European Union’s chemical regulation system.126 The absence of chemical information was a major motivation for the overhaul,127 just as it had been thirty years earlier with TSCA (plus ca change, plus c’est la meme chose).]2S One such study concluded that publicly available base data existed for only 14% of the HPV chemicals studied, less than a base set existed for 65%, and no data existed for 21%.129 Another European study found a similar pattern (17-22%) across a range of production amounts, from 10 to over 1000 metric tons per year,130 and yet another reached conclusions similar to EPA’s.131
It is particularly noteworthy that virtually all of the foregoing results actually understate the data gap in three crucial ways. First, most rely on the SIDS data, which are limited and focus on hazard identification (risk-ashazard). Second, the SIDS data leave the exposure side almost entirely unaddressed. The SIDS data are thus by no means sufficient for a defensible quantitative risk assessment that would form the basis for a determination of risk-as- probability. Third, the studies focus on HPV chemicals, which are not only a small portion of all chemicals, but are also the ones about which one might expect the greatest amount of data to exist. The existence and severity of the chemical information data gap must therefore be taken as firmly established.
C. Managing the Data Gap
1. Supply and Demand.-As the foregoing discussion demonstrates, different regulatory triggers and standards create different levels of demand for chemical information. Regardless of who has the burden of generating the information, risk-as-probability clearly demands a far broader range and detail of chemical data than risk-as- hazard.132 Systems that abjure risk entirely-technology-based regulation or pollution taxes, for examplecreate different and less onerous sets of demands.
Conversely, a regulatory system operates in an environment that has a preexisting supply of chemical information. Some information is naturally created through the obvious incentive for the manufacturer of an industrial chemical to have some basic knowledge of its useful and acutely hazardous properties. However, the natural incentive only takes one so far in supplying chemical information. Hazard information has negative liability, regulatory, and economic consequences,133 while remaining in ignorance has few negative consequences because long latency, nonsignature health effects, and diffuseness and rarity of effect make it difficult to trace health effects to their sources.134 A major function of regulation, therefore, is to increase the supply of information, and the informational requirements of a regulatory system have a profound effect on supply.
Ideally, the demand and supply of information are more or less in equilibrium. A system that is generating more information than it needs to make decisions is probably inefficient, unless there are secondary uses for the information. A system that generates less than it needs will not be able to effectively perform its protective function.135 The former is not a common problem; the latter is. As Professor Karkkainen explains in his contribution to this Symposium, information demands can be debilitating to agencies, resulting in underproduction of regulation, too-late regulation, and outdated regulations.136 By adopting risk-as-probability as the regulatory standard for disciplining the response, the chemical regulatory system has created a high demand for chemical information. Professor Houck’s description cannot be improved upon: risk-based standards “eat up heroic amounts of money, remain information-starved, feature shameless manipulation of the data, face crippling political pressure, and produce little abatement.”137 The data gap documented above is clear evidence of a substantial and system-wide disequilibrium in which the demands of regulation for chemical information far outstrip the supply.
2. Bridging and Filling.-Perhaps a more vivid way to picture the difference between supply and demand is to translate it into the data gap metaphor.138 A gap can be filled or it can be bridged. A filling strategy increases supply. It acknowledges the lack of adequate data, but insists on the need to acquire such data before regulating. It is, in this sense, the logical consequence of the risk-as-probability approach to regulation, and in particular, of the quantitative character ofthat approach. A bridging strategy takes the opposite tack. Acknowledging the lack of adequate data for a full analysis of (especially) risk-as-probability, it adopts regulatory techniquestriggers, standards, and procedures-that do not require as much information. Instead of increasing supply, bridging reduces demand.
IV. Harnessing the Demand for Information
Deliberately or not, regulatory systems deploy techniques that either redress or exacerbate the imbalance of demand and supply of chemical information. The default strategy for chemical regulation, as one might expect, is to fill in missing data and adopt regulatory techniques accordingly. This Part of the Article begins with a discussion of the baseline filling strategy. We next consider more effective filling techniques. Even with significant improvements, however, the filling strategy has inherent limitations. The Article therefore concludes by describing several alternative techniques that are based instead on a bridging strategy.
A. Filling Strategies: The Supply-Side Baseline
A rational regulatory system is built on information about the external world, and accordingly the natural impulse of any such system is to supply missing data. TSCA epitomizes the filling approach to regulation that is associated with risk-as- probability.139 TSCA’s statement of policy commits both regulators and the regulated industry to a comprehensive examination of the risks and costs of the use of a chemical140 and to the development of “adequate data. . . with respect to the effect of chemical substances and mixtures on health and the environment.”141
The demand side of TSCA’s balance is substantial in several respects. Congress did not reject all risk, but only “unreasonable” risk.142 Unreasonable risk was expressly understood to be a level of risk above zero,143 to be set by consideration of the probability of harm and the costs of regulation.144 Other requirements, such as adopting the “least burdensome”145 restrictions, have been read to demand a detailed costbenefit analysis of numerous potential levels of regulation, with the most severe restrictions requiring the strongest justification.146 Procedurally, TSCA requires that a rule be accompanied by “a statement with respect to” the human and environmental risks of the chemical, the benefits and available substitutes for the chemical, and “the reasonably ascertainable economic consequences of the rule.”147 More importantly, the adoption of the “substantial evidence” standard of judicial review was intended by Congress to signal a more skeptical and searching review,148 which intensifies the agency’s need to generate large amounts of data in support of its actions. The extreme case is Corrosion Proof Fittings, in which the court deemed EPA’s ten years of study and over one hundred studies of the effects of asbestos to be insufficient to support the restrictions that the agency had sought to impose.149
While TSCA’s statement of policy indicates that Congress sought to balance the demand and supply of chemical information,150 in fact, the techniques for generating “adequate data” are woefully inadequate to the task.151 For new chemicals, TSCA adopted the relatively toothless premanufacture-notice (PMN) procedure.152 PMN requires just thatnotification-and gives EPA only a brief window to request more information.153 According to the Agency itself, “There is no defined base data set required before PMN, and TSCA does not require prior testing of new chemicals. Consequently, less than half of the PMNs submitted include toxicological data.”154 The Governmental Accountability Office (GAO) concluded that “EPA lacks sufficient data to ensure that potential health and environmental risks of new chemicals are identified.”155
For existing chemicals, TSCA’s data-gathering system is equally limited. EPA can require manufacturers to submit a variety of existing environmental and health-effects data, but it encounters several procedural and definitional barriers in the statute itself.156 EPA also has the authority to require manufacturers to test existing chemicals,157 but in each case EPA must first make several formal findings that are subject to judicial review under the demanding “substantial evidence” standard.158 GAO concludes, “EPA does not routinely assess existing chemicals, has limited information on their health and environmental risks, and has issued few regulations controlling such chemicals.”159 In fact, TSCA represents not only the supply-side baseline, but also the worst case. Not only does it adopt a filling strategy, it places the burden of proof squarely on the government to prove unsafely, that is, to do the filling. Thus, it creates almost no new incentives, other than the purely hortatory, for the private sector to develop chemical information beyond preexisting liability considerations.160 The effect of this allocation of the burden is intensified by the requirements that EPA make certain specific findings on toxicity, the adequacy of other federal laws, and alternative regulatory approaches161-all of which are subject to “substantial evidence” judicial review.162 This burden has been read aggressively by the courts that have reviewed TSCA cases.163 Moreover, EPA must go through an elaborate hearing process that includes oral testimony by interested parties and even cross-examination.164
Professors McCubbins, Noll, and Weingast have observed, “Because industries possess much of the information relevant to regulatory decisions, elaborate processes give them more power by increasing the importance of that information.”165 As they predicted, EPA has never taken a great deal of mandatory action under TSCA.166 The agency has promulgated only five chemical bans since 1976 (thirty- two years ago),167 and Corrosion Proof Fittings effectively put an end even to that in 1991.168 As long as the burden remains on the government, regulation will be limited by governmental resources and will be subject to demands for ever more information in judicial (Corrosion Proof Fittings) or political (“sound science” rhetoric169) forums. The result is a regulatory system that only addresses easy cases and relies primarily on voluntary compliance.170
B. Better Filling Strategies: Supply-Side Improvements
The problem with the TSCA filling strategy is familiar and widely, if not universally, acknowledged. There is a considerable literature that seeks to address the problem by attending to the supply side, that is, by developing better filling strategies.
1. Licensing: FIFRA and REACH.-The most effective way to increase the supply of chemical data is to place the burden of proving safety on the proponent of a product or activity, as opposed to placing the burden of proving unsafety on the regulator. Licensing is a relatively unusual technique in environmental law, but it is deployed in FIFRA,171 the federal pesticide statute.172 FIFRA requires manufacturers of pesticides to demonstrate that a pesticide “will perform its intended function without unreasonable adverse effects on the environment,”173 a standard that, like TSCA’s “unreasonable risk” formulation, takes into account risk and “economic, social, and environmental costs and benefits.”174 Not only does pesticide registration require registrants to provide a battery of safety and use data to support the claims made for it, but FIFRA also creates a powerful legal device, the “data call-in,” for demanding data on existing products without any prerequisites or even recourse.175 The Toxicity Testing study demonstrates the effectiveness of this approach: there is a marked difference between the data available for chemicals subject to licensing (such as pesticides, foods, drugs, and cosmetics) and other industrial chemicals.176 Licensing does not by any means reduce the demand for chemical information-FIFRA is just as much a filling statute as TSCA- but it does a better job of supplying the information.
More recently, the European Union adopted a new chemical regulatory regime, known as REACH.177 Judging TSCA to be a cautionary tale of how not to regulate chemicals effectively, the EU designed REACH to be in many ways the anti-TSCA.178 The most dramatic divergence from TSCA is the aggressive commitment to generating data on existing chemicals,179 and the primary means for doing this is a registration procedure.180 REACH’S objective is to move from “no data, no problem” to “no data, no market.”181 Registration is primarily an information-provision process, with the addition of a chemical-safety assessment for chemicals produced in quantities over ten metric tons.182 REACH sets out very deliberately to eliminate the common difference between the information supporting new and existing chemicals by implementing “a step by step process to address the ‘burden of the past’ and develop adequate knowledge for existing substances that industry wants to continue marketing.”183 Finally, in order to avoid a massive increase in animal testing, REACH encourages the development of nonvertebrate testing mechanisms (such as structure-activity relationships between chemicals) to be an adequate basis for regulatory action.184
The differences between REACH and TSCA can be overstated.185 Both REACH and TSCA address chemical regulation in a comprehensive, rationalist manner. They seek to grapple with the entire problem and to assemble all of the relevant information. Scientific information is highly privileged in this approach-REACH requires a “sound scientific basis” for restrictions on chemicals186-and the ultimate standard for acceptability is risk-as-probability.187 Both the authorization and restriction phases of REACH regulation require economic analysis and the justification of residual hazards by benefits.188 Fundamentally then, REACH also adopts a filling strategy, but it deploys techniques like licensing and the use of structureactivity relationships that should increase the supply of data well above the TSCA baseline.
2. More and Better Use of Information.-The supply of information can be increased by better funding for chemical testing by the government or government grantees, or by the use of new information technologies to analyze and disseminate existing data. Adequate funding for toxicity research is a chronic problem, and the federal contribution to chemical research and development has been in a steady decline for years.189 Several innovative solutions have been suggested, such as Professor Lyndon’s “superstudy” funding proposal,190 but in the meantime, EPA’s overall budget for research, which includes many other kinds of research, declined substantially between 1976 and 2005.191 Government has a particularly important role to play in generating information in areas of structural market failure, such as “orphan” chemicals, discontinued (but persistent) chemicals, foundational research, and information that requires widespread inquiry (such as epidemiology), particularly expensive equipment or new technologies.192
Existing information can also be made more useful, and in that sense, more plentiful, by the deployment of new technologies for data collection, analysis, and dissemination. Professor Esty’s projections of the impact of new technologies on the data gap provides an excellent, if optimistic, survey of “our ability to fill the information gaps” in environmental regulation.193 Institutionally, government agencies are well positioned to assemble and organize large amounts of information from studies published in scattered publications, submitted through applications and reporting requirements, and collected from monitoring and inspection.194 Government is also uniquely situated to provide uniform standards and formats for data, to define key data sets (like SIDS), and to develop quality-control policies and practices. All of these are activities that the U.S. and European governments currently engage in to some degree, and they could usefully be expanded.
3. Voluntary Efforts.-EPA can also seek to obtain chemical information through the voluntary efforts of the enterprises that produce and use chemicals. Professors Coglianese, Zeckhauser, and Parson identify “two basic strategies that regulators employ to secure information: exploit asymmetries of interests across or within firms, and create incentives for disclosure.”195 They consider a variety of techniques that apply these strategies- mandated disclosure, rewards and recognition, formal interaction between regulators and regulated entities-and they ultimately settle on extensive informal interactions between the regulators and regulated as the technique most likely to encourage voluntary information sharing.196 EPA’s Office of Pollution Prevention and Toxic Substances (OPPTS) is firmly committed to cooperative strategies (styled “partnerships”).197 The High Production Volume (HPV) Challenge Program is a particularly relevant example of such partnerships. A collaboration between Environmental Defense, the American Chemistry Council, and EPA,198 the program calls on the manufacturers of HPV chemicals to voluntarily make health and environmental-impact data publicly available through sponsorship of individual chemicals.199 According to EPA, 1400 HPV chemicals have been sponsored through the program (and more through international programs) as of June 2007.200 While EPA characterizes data collection under the program as “nearing its conclusion,”201 Environmental Defense has been extremely critical of the actual results of the program.202
Several other voluntary programs exist (for example, the FYI program under TSCA [section] 8(e)), and they do generate data.203 However, the structural flaw in all of these programs is that, being voluntary, they give EPA no new authority to demand that gaps actually be filled.204 So, when cooperation is not forthcoming or voluntary commitments are not honored, EPA can only resort to the inadequate filling techniques that we have already seen. EPA implicitly acknowledges this problem by “redefinfing] success” under its chemical-testing programs to include voluntary approaches in its totals.205 While this may be fair enough as far as it goes, the width of the data gap demonstrated above suggests that it is an extremely optimistic position. Moreover, a recent Report by the EPA’s Office of Inspector General was highly critical of EPA’s many voluntary programs. While EPA makes much of their successes, the Report found that they lacked the kinds of measures and internal controls that are needed to determine whether the programs are in fact successful.206 4. Competition-Based Regulation.-Professor Wagner has recently offered the intriguing suggestion that more chemical information could be generated by a regulatory mechanism that would (to use her phrase) “divide and conquer” the manufacturers of chemicals by placing them in competition with one another to demonstrate the superior safety of their products.207 By creating a “market” for chemical-safety information where none currently exist-the markets already perform well for utility and price information, of course-competition among manufacturers would create a strong incentive to generate (or reveal) the data that would demonstrate safety, lest a competitor either demonstrate its own superior safety or another’s greater danger.208 “A company who receives an ‘inferior’ designation would, at the very least, be required to label their product.”209 Liability consequences (good or bad) would also be possible.
Wagner’s proposal aims to improve on a filling strategy. Its objective is “to dredge up more comprehensive and accurate information on chemical risks and safer substitutes.”210 Moreover, it envisions a process of adversarial hearings before EPA to establish superiority and assign appropriate labels to products.2″ These would presumably be highly information-intensive procedures- deliberately so, because their objective is to reveal and evaluate as much chemical risk information as possible. As adjudicator, EPA would not be in the position of generating the information (in this respect, it is like a licensing scheme), so the amount of information would not be limited by EPA’s resources.212
5. Limitations on Filling Strategies.-The foregoing ideas are useful and even important improvements in the supply-side strategy of filling the data gap. Information costs can be lowered, and more effective incentive systems can be established. But no matter who is required to generate the information, how much efficiency is gained, or how great the incentives are to generate data, the demands of the risk-as-probability approach are extensive, expensive, and time- consuming. The REACH schedule-which is intended to be extremely aggressive and is probably overly optimistic-contemplates eleven years to fill the data gap for existing chemicals.213
The fundamental limitation on filling strategies is that a quantified, comprehensive approach to regulatory standards is an unquenchable thirst. It is the nature of scientific inquiry that there are always more questions than answers, and answers beget more questions. This approach to knowledge has served science well, but- as the Ethyl Corp. court and other commentators have observed-there is a basic mismatch between the needs and norms of scientific inquiry and the needs of regulatory agencies to act preventively to address environmental threats.214 Moreover, in the highly contested world of regulatory action, a study that demonstrates one thing will always call forth a new study to contradict it, creating at least apparent uncertainty.215 The response of the chemical industry to the Environmental Defense Fund’s Toxic Ignorance study is a case in point. The legal216 and political217 rewards of manufacturing uncertainty are obvious and practically irresistible. The data gap is, from this perspective, a bottomless pit of uncertain and contestable information, with never enough resources to fill it. Therefore, while supply-side improvements may be very appropriate for the limited demands of defining the problem as risk-as-hazard, they are not adequate to support disciplining the response using risk-as-probability.
C. Bridging Strategies: Demand-Side Reform
The limitations of the supply-side strategy call for serious consideration of a demand-side approach. Techniques that limit demand have some obvious advantages for supporting an active program of chemical regulation.218 A demand-side strategy reduces the amount of information required in order to take regulatory action, permitting regulatory agencies to spread their limited resources over a wider front-or in the case of stalled programs like TSCA, to take mandatory action at all. A demand-side strategy also avoids the bottomless-pit problem because, to continue the metaphor, a bridge can span a gap of any depth. It bears emphasis, though, that none of the bridging techniques are data-free. At a minimum, there must be a prior determination that a particular substance poses a threat. Thus, all of these bridging techniques require a regulatory trigger- risk-as-hazard-that defines the universe of chemicals requiring attention; the difference from the supply-side baseline is that the regulatory standard is not risk-asprobability. Professors Shapiro and Schroeder, in their article on reorienting cost-benefit analysis, recommend precisely this rethinking of the relationship between trigger and standard.219 By clearly distinguishing between them, the regulatory system can exploit the different meanings of risk to reduce information demands and bring them into equilibrium with supply.
1. Technology-Based and Incremental Standards.-The information advantages of technology-based standards have been thoroughly reviewed by others.220 The capability of a “best available technology” is far more readily and definitively determined than risk-as-probability; the former is largely a matter of establishing proper categories of existing facilities and measuring their performance.221 While a technology-based standard is by no means a trivial inquiry, its information demands “pale in comparison” with risk-asprobability standards.222 On the risk side, technology-based standards only require basic knowledge of the toxic properties of the chemical (risk-as-hazard) for the purpose of initially identifying chemicals subject to regulation. Thus, in the 1990 Clean Air Act Amendments, Congress adopted a technology-based standard (maximum achievable control technology (MACT)),223 in an effort to breathe life into a mostly dead (or slightly alive) air-toxics program. The bulk of the risk reduction work is accomplished by the primary MACT standard. However, if after application of MACT there remains a residual risk above a “lifetime excess cancer risk[] to the individual most exposed” of one in one million or more, then EPA must establish a second set of standards that “provide an ample margin of safety to protect public health” within eight years of the promulgation of the initial standards.224 By focusing first on the technology-based standard and only secondarily on the risk-as- probability standard, EPA has been able to complete most of its primary work on the congressional list of air pollutants (not on time, of course, but no one expected that), and it has recently moved on to the residual risk determinations.225
The Clean Air Act Amendments exemplify not only a technology- based standard, but also a more information-efficient, incremental approach to setting regulatory standards. The weakness of technology- based standards has always been their use of what is essentially a surrogate criterion for environmental protection. Obviously, the point is not to require adoption of any particular technology for its own sake, but rather to reduce hazards to health and the environment. Even the best technology will not necessarily achieve that goal,226 and so technology-based standards are often used as an interim measure on the way to solving the problem on its own, health- based terms.227 The MACT standard is supposed to reduce the bulk of toxic emissions relatively quickly, and it is followed by a risk-as- probability standard to attain Congress’s ultimate health-based objective.228 Likewise, while the later phases of the Clean Water Act are unfortunately more rhetorical than real, the original intention of the statute was that the primary technology-based standards be replaced in the fullness of time with standards that would attain “fishable-swimmable”229 waters, and finally, that “the discharge of pollutants into the navigable waters be eliminated.”230
An incremental approach, which makes decisions through several relatively small steps, contrasts with the comprehensive rationality that riskas-probability and cost-benefit analyses represent.231 Canada’s chemicals program, for example, operates on the basis of triage, focusing not on absolute and quantified risk, but on relative hazard.232 Following this path, Canada has made great strides in establishing standards for the chemicals of greatest concern.233 Incrementalism, as its advocates have noted, is dynamic and flexible; it adjusts to new circumstances, handles uncertainty well, and also permits learning from experience.234 It has distinct weaknesses, too,235 but one undeniable advantage is the lower demand for new information. Smaller steps make better use of existing knowledge and do not make the ravenous information demands that comprehensive solutions do.
2. Market-Mimicking Techniques.-Traditional “command-and- control” regulation calls on government to do most of the work of assembling relevant information and choosing an appropriate response. Because governmental action is subject to strict legal requirements of procedural regularity and substantive political and judicial review, such regulation demands a great deal of supporting information, regardless of the techniques employed. Market-based techniques place the burden of information gathering and even, within limits, decision making, primarily on private actors, giving them an ability and incentive to make decisions more information- efficiently. a. Pollution Taxes.-Market-based, or market-like, regulatory techniques are frequently advocated as efficient improvements to the traditional command-and-control regulatory system because they give individual regulated entities a high degree of flexibility in deciding how best to reduce the risks they create.236 The deterrent effect of tort law operates in this way, inasmuch as individual actors can choose to modify their riskcreating conduct-or not-in response to the prospect of future liability.237 This approach, however, allows the polluter to make its own risk decisions without collective input, which is an especially troublesome problem in the context of toxic risks, where not all costs are internalized.238 The polluter can impose a greater level of risk than the public might otherwise accept, or distribute the risk in ways that are economically efficient but otherwise unfair or unjust. Pollution-tax systems suffer a similar defect. They enjoy widespread theoretical support for their efficiency,239 but they have been rarely adopted.240
On the other hand, tax systems have a substantial information advantage over traditional regulation. Unless the tax rate is assigned in order to “fine tune” risk levels-in which case it would make far more sense just to offer flexibility within a traditional system, like “bubbles”241-setting the rate should not require the kind of detailed risk assessment that is associated with risk-as- probability.242 Indeed, given the flexibility permitted by a tax, the exposure side of the risk equation could not even be estimated for such purposes. A tax or charge system, in other words, raises many issues which may make them unsuitable for chemical regulation, but they would have the advantage of reducing data demands.
b. Penalty Default Rules.-In an important article on contract law, Professors Ayers and Gertner introduced the idea of “penalty default” provisions as legal techniques for incentivizing the production of relevant information.243 A penalty default specifies a harsh provision for one party unless that party can demonstrate, by providing additional information in the bargaining process, that a different provision is more appropriate.244 As explained by Professors Ayers and Gertner:
Penalty defaults are designed to give at least one party to the contract an incentive to contract around the default rule and therefore to choose affirmatively the contract provision they prefer. In contrast to the received wisdom, penalty defaults are purposefully set at what the parties would not want-in order to encourage the parties to reveal information to each other or to third parties (especially the courts).245
Professor Karkkainen and others have recognized the possibilities that penalty defaults offer in the environmental setting, noting that several implicit penalty-default provisions already exist.246 California’s Proposition 65247 bans both the discharge of carcinogens into drinking water248 and the failure to notify individuals of exposure to carcinogens, unless the discharger can demonstrate an acceptably low level of risk.249 The penalty default is the ban on discharge or exposure, and it creates an incentive for the discharger not only to come up with information, but also to support regulation, since a regulation establishing an acceptable risk level acts as a safe harbor against liability under the law. Proposition 65 has been immensely successful in creating hundreds of carcinogen standards.250
At one level, Proposition 65 is a straightforward burden- shifting device for supplying more chemical information, like the licensing statutes we have seen. However, it also reduces information demand in two ways. First, the trigger for action is a finding of risk-as-hazard (“a chemical known to the state to cause cancer or reproductive toxicity”)251 and the initial standard is simple-a ban on discharges or unwarned exposures. Second, only the secondary, optional standard (a discharger can choose to pursue one or not) is based on risk-as-probability, that is, “significant risk.”252 And even this risk-as-probability standard is less demanding of information because industry has a huge incentive to accept regulation based on a scientific record that, in other circumstances, it would have an incentive to challenge as insufficient or too highly contested.253
Penalty-default rules are a subset of a larger category of regulatory techniques that Professors Shapiro and Glicksman call “back-end adjustment.”254 These techniques establish a strict primary standard and then permit deviation from that standard for good cause. Examples are legion: deadline extensions, waivers and exemptions, enforcement discretion, periodic review, and others.255 Back-end adjustment permits generalized regulation while recognizing the inevitability of unforeseen facts and factors. It can account for changed circumstances; it makes a strict regulatory baseline more politically and economically acceptable by providing a safety valve; and it can supply more information by casting the burden of proof upon the seeker of the exemption.256 For our purposes, the major advantage is that back-end adjustment, like incrementalism generally, limits the amount of information needed to address a problem, requiring fine tuning only as needed.257
c. Environmental Performance Bonds.-The idea of an environmental- performance bond is that the proponent of a hazardous product, activity, or technology (the technique is most suited to the “macro” setting of ecological harm, such as road building and timber sales) must obtain in advance a bond sufficient to cover the worst-case consequences of its actions.258 The objective is more to influen