Before Risk: From Purity to Safety

Although it is common to locate the origins of federal environmental policy in the early 1970s, there were crucial early precedents. The first federal environmental regulations arose in response to the production and marketing of processed foods in the late nineteenth century. At the time, most Americans viewed packaged foods with well-founded suspicion, as many contained inferior and even dangerous ingredients. It was a young chemist in the Department of Agriculture, Harvey Wiley, who took it upon himself to investigate the ingredients in new, packaged foods. Perhaps best known for his “poison squad” experiments (in which he hired healthy young men to consume vast amounts of preservative-laden food in order to document the ill effects), Wiley emerged as a vociferous and influential crusader for “pure” food. He argued that most food additives were unnecessary and were used by unscrupulous producers simply to mask inferior ingredients and increase profits. These arguments were instrumental in securing the first federal legislation addressing food safety, the Pure Food and Drug Act of 1906. That law banned “poisons” in the food supply and gave to Wiley’s bureau responsibility to determine which chemicals were “injurious to health” and thus illegal. This law assumed that good food was basically pure and that the problems stemmed from a few errant chemicals that could be addressed on a case-by-case basis.^{Footnote 5}

The quest for food purity emerged out of dominant understandings of health at the time. The Food and Drug Act was grounded in medical bacteriology and the new “gospel of germs,” which located the cause of disease in microscopic pathogens that penetrated an otherwise healthy body. Modern medicine constituted the healthy body as a body free from disease-causing agents, and the new food law—like concurrent regulations enacted for water quality—did the same.^{Footnote 6}

But purity was an elusive goal. Court decisions in the interwar period placed the burden of proving harm on the government and with the rapid expansion of the chemical and processed food industries, hundreds of chemicals entered the food supply with little and frequently no testing. Although major revisions to the Pure Food and Drug Act enacted in 1938 reaffirmed the basic ban on “poisons” in the food supply, the rapid unleashing of thousands of new chemicals in all walks of American life after World War II marked an uncontrolled material revolution. In most arenas (housing, consumer products, workplaces), these new chemicals were initially subject to no substantive regulation. While the Food and Drug Administration (FDA) held authority to approve or disapprove the use of particular chemicals in foods, that agency quickly found itself overwhelmed. Before the war, the FDA had dealt with perhaps a dozen or so chemicals, but by the early 1950s it was dealing with nearly one thousand, with dozens more being introduced every year.^{Footnote 7}

What replaced the concept of “purity” within food regulation was the ideal of “safety.” Safety was a concept that had been operationalized in the industrial workplace. By the turn of the century, new legal doctrines placed a burden on employers to provide safe workplaces and safe tools, and the discovery of industrial disease extended that concept to the factory environment.^{Footnote 8} But modern factory environments were in no sense pure. Consequently, as historian Christopher Sellers has documented, the definition of a safe workplace with respect to chemicals would be built around the new science of physiology. Physiology focused on metabolism, conceptualizing the body as a self-regulating system that sought, in Harvard professor Walter Cannon’s term, a condition of “homeostasis,” or equilibrium. The new physiologists argued that when disturbed—by fatigue, by extreme temperatures, or by toxins—the body’s ability to self-regulate would result in a new state of equilibrium, so long as it was not pushed past certain boundaries.^{Footnote 9}

Academic physiologists and chemists worked together during the interwar period, supported by their industrial patrons, to develop the new field of industrial toxicology. The basic assumption underpinning their work was that for any chemical exposure, the body could adapt—up to a point. Put another way, they held that a minimum chemical dosage was required to generate a biological effect: “the dose makes the poison.” These new toxicologists worked to develop quantitative limits for effects by dosing laboratory animals (mice, rats, dogs) with measured amounts of chemicals. The goal was to determine the level of chemical concentration below which no biological effects were observed; this they labeled a “safe” exposure or threshold level. Consequently, for these professionals the concept of environmental purity had no role: that both environments and bodies absorbed some quantity of industrial chemicals was not, in itself, a problem.^{Footnote 10} Viewed from the perspective of the factory manager, the absorption of chemicals, much like industrial accidents or fatigue, was simply part of the bodily trade-off for wages.

The concept of biologic thresholds and safe levels gradually spread across the regulatory bureaucracy and into the consumer realm. Significantly, the FDA moved away from its earlier emphasis on purity and began to adopt toxicological techniques to evaluate drugs and pesticides in the 1930s; the agency then extended these techniques to proposed food additives. This was a pragmatic change within the FDA that occurred without public scrutiny or debate. However, the FDA adopted the approaches of industry warily. Outside the factory, the assumption of a voluntary trade-off for wages did not hold, and the FDA recognized that regulations intended to protect consumers had to take account of sensitive individuals: the sick, the elderly, the very young. There were, moreover, the problems of multiple exposures and unknown synergistic effects. Consequently, FDA pharmacologists relied on their own discretion and scientific judgment in deciding which additives to approve, which to deny, and which tests to require before approving an additive. Additionally, they applied a “safety factor” (typically 100) to laboratory-defined threshold levels. What this meant was that regulators looked at the largest dosage that generated no observable effects in tested animals and then divided that dose by 100 to account for any differences in animal-versus-human susceptibility.^{Footnote 11}

This approach toward ensuring safety in the face of chemicals, and later environmental pollution, would be endorsed legislatively after the fact. In 1950, Congressman James Delaney of New York opened a set of hearings on food additives that would last for more than two years. These hearings made public the fact that the FDA had quietly moved from a regime of “purity” to one of “safety,” relying on the methods of toxicological animal testing. The Food Additives Amendment, passed in 1958, formally endorsed the FDA’s safety approach: it required premarket testing for all chemicals in foods (which the FDA had sought), while excluding from these requirements compounds already in use and those “generally regarded as safe.” However, the hearings had revealed growing concern about possible carcinogens in the food supply and had drawn attention to the claim of several scientists that there was no demonstrably “safe” level for such chemicals because the dose-response relationship was so uncertain. At the insistence of Congressman James Delaney, Congress rejected the safety approach for cancer-causing chemicals. The so-called “Delaney Clause” to the Food Additives Amendment banned outright the presence of any additive known or suspected to cause cancer in laboratory animals or humans. In the particular case of carcinogens, older ideas of bodily purity held sway, and zero tolerance remained the law.^{Footnote 12}

The prevailing assumption in environmental health regulation at midcentury was that regulations should emerge from biological understandings of health and that, with the notable exception of the Delaney Clause, the government should attempt to ensure individual safety with respect to chemicals but not zero exposure. This assumption would structure FDA’s response to food additives until the late 1970s as well as the federal government’s early environmental pollution legislation; it would also empower those with certain scientific backgrounds as arbitrars of public health.^{Footnote 13} The Clean Air Act (CAA) of 1970, one of the nation’s most far-reaching environmental laws, is a key example. Enacted after years of hearings, that legislation directed the Environmental Protection Agency (EPA) to establish vigorous standards for toxic air pollutants with an “ample margin of safety to protect public health.” In the debate that preceded it, Congress explicitly rejected calls to quantify the benefits to public health, as well as any consideration of costs. That legislation passed both houses of Congress with just one negative vote.^{Footnote 14}

When EPA issued health-based air-quality standards under the CAA, industries immediately challenged the strict limits. Among the most important challenges was a case brought by Lead Industries, a trade organization whose seventy-eight members included most of the country’s producers and commercial consumers of lead. That lawsuit explicitly challenged the agency’s failure to take cost, or “economic feasibility,” into account when promulgating air-quality standards. Yet the D.C. Circuit Court strongly endorsed the EPA’s approach, stating that when promulgating air-quality standards the EPA was neither required nor permitted by Congress to consider economic or technological feasibility. Furthermore, had the Administrator considered such factors, it would have amounted to a “trespass beyond the bounds of [the Administrator’s] statutory authority.” Congress took a similar approach in the passage of amendments to the Federal Water Pollution Control Act two years later.^{Footnote 15}

The adoption of pollution standards and goals based on toxicological studies—and without regard to cost—is indicative of a popular and broadly held understanding that a healthy environment was a human right. While acknowledging that modern environments were in no sense pure, early environmental legislation assumed that anthropogenic pollution should not be allowed to materially affect health. It was this popular understanding that made it impossible for industry supporters to overturn the Delaney Clause (forbidding the use of suspected carcinogens in food) despite repeated attempts. Moreover, it was this understanding to which Rachel Carson appealed in her best-selling book, Silent Spring. Although the book posed a radical challenge to contemporary pesticide policy, it did so by appealing to classic American values: democracy, freedom, government transparency, and individual rights. For Carson, those rights included the right to enjoy an unpolluted natural world, the right to know about environmental harms, and the right to bodily integrity—as she put it in her testimony before President Kennedy’s Science Advisory Committee, “the right of the citizen to be secure in his own home against the intrusion of poisons applied by another person.” The committee would endorse that view just a few years later, explicitly recognizing “individual rights to quality of living, as expressed by the absence of pollution,” and flatly declaring that “there is no right to pollute.” These core beliefs crossed the American political spectrum, finding expression on the right, for example, in vehement opposition to the fluoridation of public water supplies.^{Footnote 16} Yet less than two decades after the publication of Silent Spring, environmental policy would be grounded on a very different basis.

Nuclear Weapons and Systems Analysis

The seeds of an alternative approach already existed and had for quite some time in a key area of government policy: nuclear weapons production. The U.S. government had pursued the production and testing of atomic weapons since the early 1940s, despite unavoidable effects on worker and public safety. Scientists had known since the mid-1920s that radiation exposure caused genetic effects across generations with no identifiable threshold, or safe, level; by the early 1950s, most believed the same was true for radiation-induced cancer, particularly leukemia. Yet, in the prosecution of World War II and the Cold War, American leaders opted to pursue nuclear technologies, relying on prominent scientists to both manage and justify the hazards.^{Footnote 17} In the minds of those managing the production of the first atomic bombs, wartime success clearly outweighed most risks to worker and public health. Although radiation standards for workers were adopted, they were selected so as to ensure that production goals were not seriously compromised.

The decision to continue nuclear weapons production after the war and to extend nuclear technologies into the civilian realm forced the U.S. government to acknowledge the dangers of radiation and to publicize radiation-exposure standards. Yet in the process, officials at the Atomic Energy Commission also sought to obscure and minimize the known dangers. In the case of genetic effects, as well as for many cancers with long latency periods, it is impossible to say whether a particular effect is due to radiation exposure or whether it would have occurred anyway. Moreover, risks that are substantial across a population may be slight for an individual. This paradox made it possible for government officials to downplay known and suspected threats of exposure by focusing on only the relatively small likelihood of effects at the individual level, while glossing the statistical likelihood of tens of thousands of mutations across the population.

The shifting language in which radiation health standards were couched is telling. During the war, biologists working within the Manhattan Project had developed a “tolerance dose” for nuclear workers to limit illness and injury. Immediately after the war, the U.S. standard-setting body, the National Committee on Radiation Protection (NCRP), adopted the radiation-exposure standards developed within the weapons program, despite the discomfort of some of the scientists involved. The concept of a tolerance dose, derived from chemical toxicology, implied a level below which no effects occurred. Scientists recognized that there was no minimum level that would prevent genetic effects on the population, and some even doubted whether such a level existed for somatic effects on the individual. As early as 1947, key members of the NCRP had advocated both renaming and rethinking the standards. In a report issued in 1950, the International Committee for Radiation Protection, a nongovernmental body on which key members of the NCRP served, embraced the term “maximum permissible exposure,” set at a level so as to “involve a risk that is small compared to the other hazards of life.” In 1956, the NCRP formally renamed its own standard: “tolerance dose” became “maximum permissible dose.” The new standard allowed for an “acceptable risk” of harm defined as a “risk so small that it is readily acceptable to the average individual; that is, . . . essentially the same as is present in ordinary occupations involving no exposure to radiation.”^{Footnote 18} The NCRP introduced both the language of risk into U.S. health regulation and the concept of risk as a unit for comparison.

However, neither the NCRP nor the Atomic Energy Commission, the federal agency charged with nuclear weapons and energy production in the postwar period, tried to explain or rationalize these novel approaches to public health. The government’s deliberations over radiation safety—and much of its data—were secret, its levels of “acceptable risk” a compromise between limiting health effects and maintaining a nuclear weapons program. However, by assuming the equivalence of multiple different kinds of radiation exposure while casting risk in individual terms, officials could put existing and familiar environments (“background radiation”) into the same frame as highly polluted ones—and thus normalize the unavoidable public health effects and mortality associated with the nation’s nuclear ambitions.^{Footnote 19}

At the same moment, weapons and defense planning had generated a rethinking of “safety” in other realms. Recent developments in the fields of operations research, game theory, and statistics offered new mathematical techniques for planning and evaluating complex weapons systems. No institution was more central to this reorientation than the RAND Corporation, a southern California think tank that began in 1946 as a consultant to the U.S. Air Force. Viewing the defense policy and decision process as chaotic and liable to produce dangerously poor choices, analysts at RAND aspired to make it scientific. The problem, as they saw it, was how to elicit the best decisions in the face of vast uncertainties, terrible potential consequences, and decidedly limited human capabilities. RAND analysts emphasized the fallibility of human decision-makers within complex systems; consequently, their goal, in the words of historian Hunter Heyck, was “the design of systems that would generate rational choices automatically, whether the humans involved were rational creatures or not.” Their approach was the logical identification of all policy choices, followed by the mathematical evaluation of alternatives according to objective criteria. RAND came to label their method “systems analysis,” which was generally understood to mean the reduction of a large problem or choice into manageable parts for quantification, systematic evaluation, and “strategic” decision making.^{Footnote 20}

Although RAND’s research focused primarily on planning and policy choices, it emerged out of and in conversation with weapons engineering, and it relied upon the same methods being used to assess hardware systems. RAND itself had begun as a subdivision within the Douglas Aircraft Company and initially it retained many physical scientists and engineers on its staff. The question faced by weapons and aircraft engineers was how to assess reliability in highly complex systems. With the expedited integration of nuclear weapons into military hardware systems, it was no longer feasible to build and test prototypes, and yet the consequences of failure were potentially catastrophic. In the detached assessment of one engineer, it was “the increased lethality of high performance/high yield weapons systems” that “demanded a ‘system’ approach to predict risk.” The Air Force’s Intercontinental Ballistic Missile (ICBM) Program rapidly pushed forward the application of new mathematical methods in aerospace engineering during the 1950s and 1960s. Rather than trying to assure the safety of plant or missile design through costly redundancies, engineers began to evaluate a system’s overall reliability by estimating the probability of failure of various components. Within the ICBM program, system safety was (re)defined as “the optimum freedom form abnormal or out of sequence events which would jeopardize personnel or equipment, consistent with the operational requirement.”^{Footnote 21} To measure the risk of failure and locate key vulnerabilities, engineers disaggregated the system into parts, assigning probabilities to each possible failure event within the system; they then interrelated these subsystem-failure probabilities with Boolean logic gates. When engineers had no data on which to base the probability of failure, they assigned a subjective probability based on expert judgment. The result was a branching diagram of the overall system that allowed for a quantitative estimate of the likelihood of catastrophic failure. “Fault-tree analysis,” as this technique was labeled, was quickly adopted throughout the aviation and aerospace industry and had entered the curricula of leading engineering schools by the early 1960s. In essence, the shift from “deterministic” to “probabilistic” design in engineering marked a shift from safety to risk.^{Footnote 22}

Within the cloistered world of weapons manufacturing and defense planning, this language and its underlying assumptions were widely shared. Engineers and systems analysts worked for the same client and shared a commitment to maintaining American nuclear supremacy. They utilized the same mathematics and theorems, drew the same type of diagrams, and frequently cited one another’s work. Individual mathematicians, engineers, and physicists moved among RAND, leading aerospace companies, and major engineering schools. Within this world, unprecedented danger was something to be managed, not avoided, and the key to good management lay in formal techniques and improved quantification. The question was not whether to pursue an activity, but how to minimize the chance of failure and the extent of collateral health effects “consistent with the operational requirement.”^{Footnote 23}

Going Public with Risk

The assumptions about safety and risk that had emerged in nuclear weapons production, along with methodologies for addressing them, would enter the public sphere as a result of the push to develop commercial applications for nuclear power. As the U.S. government committed itself to the expansion of commercial nuclear power and “atoms for peace” in the early 1950s, much of the public was growing increasingly concerned over the effects of radioactive fallout. Concerns were heightened by massive nuclear fallout episodes in the United States (1953) and the Marshall Islands (1954) and subsequently the Windscale nuclear plant disaster (1957) in Great Britain. The government was forced to directly confront these concerns in order to obtain congressional approval of the Price-Anderson Act—a piece of legislation that would shield utility companies from most liability in the case of a catastrophic nuclear accident. In response, the Atomic Energy Commission published the first nuclear reactor safety study. Although that study argued that the likelihood of a large-scale nuclear accident was very small, it nonetheless laid out what the authors’ considered a worst-case accident scenario: 3,400 deaths, 37,000 injuries, and $7 billion in property damage. Although Congress passed Price-Anderson, the debate heightened public concerns, and a full-fledged antinuclear movement had emerged by the early 1960s. Nuclear insiders expressed frustration, even chagrin, at what they felt to be the unfair and irrational targeting of their industry as somehow excessively dangerous. They began to discuss new ways to justify nuclear plant siting decisions.^{Footnote 24}

Throughout most of the 1960s, that discussion occurred in technical journals and conferences of the industry. Writing in Nuclear Safety, a journal begun in 1958 at the Oak Ridge National Laboratory in part to foster these discussions among insiders, nuclear engineer Chauncey Starr argued in 1964 that what was needed was a quantitative comparison of the risk of nuclear power to those from fossil fuel plants. Starr was a veteran of the Manhattan Project and at the time a vice president of North American Aviation—an aerospace company in southern California with close ties to RAND and an early manufacturer of commercial nuclear power reactors. According to Starr’s estimates, the health risks of fossil fuel plants were much higher due to their contributions to air pollution deaths. The following year, Frank Sowby, a member of the International Committee on Radiation Protection, published a paper in the journal Health Physics titled simply “Radiation and Other Risks.” It suggested that the risks of nuclear power might be profitably compared to other, more common risks of everyday life: car accidents, heart disease, lung cancer, and so on. However, while Sowby marshaled accident and occupational data to estimate the more familiar risks, he lacked data from which to calculate a similar estimate for a nuclear reactor accident. There was no clear way to quantify the risk of an activity without a substantial history. Two years later, a UK nuclear physicist showed an audience of nuclear experts how the results of a probabilistic systems safety analysis might be combined with estimates of excess deaths due to radiation exposure to quantify the overall risks to a population from a given nuclear plant. These estimates might then be compared to the risks of everyday life along the lines suggested by Sowby; the initial comparisons appeared quite favorable to nuclear power. His paper was well received.^{Footnote 25}

Then, in 1969, Science magazine published a landmark article by Chauncey Starr, “Social Benefit versus Technological Risk.” Starr, who was now dean of the UCLA School of Engineering, generalized for a broad audience the framework that nuclear scientists had been discussing for years. Arguing that the public assessment of many technologies was “irrational,” Starr presented quantitative estimates of the “benefits” of various technologies and activities (a number derived from either dollars spent on the activity or the estimated additional income generated by the technology) and their “risks” (quantified as the individual risk of fatality per hour of exposure). Plotting these logarithmically, he used this data to infer the public’s utility function for “risk”; his conclusion was that the public routinely accepted risks that were “crudely proportional to the third power of the benefits.” While admitting that the data on which his analysis relied were spotty, Starr concluded that nuclear power easily met this criterion.^{Footnote 26}

Starr’s paper generated immense interest and considerable controversy. Those in the nuclear field already endorsed this approach, however, and in the wake of the Science paper several people began to refine the analyses and publicly advocate their adoption. Shortly thereafter, Starr organized a conference on “risk-benefit assessment” under the auspices of the National Academy of Engineering. Meanwhile, the Atomic Energy Commission had undertaken a massive study of nuclear power safety in anticipation of the renewal of the Price-Anderson Act. The draft study was released in 1974. Prepared under the direction of MIT engineer Norman Rasmussen, the Reactor Safety Study brought the methods of fault-tree analysis, subjective probabilities, and risk comparisons before Congress and to a broad public audience for the first time. Building upon the papers of Sowby, Farmer, and Starr, the authors compared the risks of commercial nuclear power to those of driving a car, falling down, and being struck by lightning. They famously concluded that the risk of dying from a nuclear power accident was roughly equivalent to that of being hit by a meteorite.^{Footnote 27}

That report met with withering criticism, from academics, antinuclear activists, and many members of Congress. But most within the nuclear community remained defenders of the report’s methods and redoubled their efforts to extend and explain them to the public. In this context, William Rowe, a high-level administrator at the EPA with a Ph.D. in Operations Research (and who had been involved in reviewing the Rasmussen report), published An Anatomy of Risk in 1977. Drawing on recent research in systems analysis, Rowe’s book refined and extended the methods used in the Rasmussen report. Anatomy of Risk would become an essential citation in the emerging social science literature on risk, and Rowe himself would become a successful consultant to government and industry.^{Footnote 28}

At the same moment, the National Academy of Sciences—particularly its Committee on Science and Public Policy (COSPUP)—emerged as a key location for the support of quantitative risk assessment and risk comparison. COSPUP was a highly influential committee that had from its inception been dominated by nuclear scientists. Its first two chairmen, George Kistiakowsky and Harvey Brooks, were veterans of the Manhattan Project and had served as presidential science advisers, while Brooks had been involved in reactor safety debates since the 1950s. In 1975, the committee undertook a review of the Rasmussen report, concluding that the overall risks of nuclear power production, though uncertain, were not alarming and needed to be weighed against the risks of other forms of energy use. At the same time, COSPUP hosted a recent biochemistry Ph.D. as a postdoctoral fellow. William Lowrance found himself working for the committee’s panel on Science and the Determination of Safety, whose senior members were none other than Kistiakowsky and Brooks. While at COSPUP, Lowrance produced a popular though not especially insightful book on safety and public policy, the most important part of which was its definition of “safety” from which the book took its title: “of acceptable risk.” Whether Lowrance knew it or not, the term “acceptable risk” had been in common parlance within the nuclear science community for more than two decades. In addition to reintroducing the term, the book addressed the assessment of safety across a range of issues (food additives, pesticides, nuclear energy, transportation, and consumer products), signaling the migration of risk beyond nuclear policy and into a host of new policy domains. Lowrance’s book, along with Rowe’s, became a staple of the emerging literature.^{Footnote 29}

“Policy Analysis”

Intellectuals and academic institutions played a key role in migrating the techniques and assumptions of systems analysis from defense planning and nuclear engineering to other fields, and, in addition to the NAS, both RAND and Harvard’s Kennedy School of Government were crucial hubs. Systems analysis had formally entered the federal government when President Kennedy selected Robert McNamara to head the Department of Defense in 1961. McNamara brought key RAND staff into the Pentagon and introduced RAND’s “Policy-Program-Budgeting System” (PPBS), which attempted to integrate policy goals with agency budgets through the use of systems analysis. PPBS sought to optimize defined policy outcomes per unit of cost. Advocates of this approach argued that it improved decision making by presenting agency leaders with a limited range of viable options for achieving a specified objective; those options could then be quantitatively compared using an objective criterion (cost). The best choice was the most cost-effective means of achieving a defined goal. These methods spread quickly to other agencies following an Executive Order issued by President Johnson in 1965 that directed most federal agencies to implement PPBS. As domestic budgets grew under Johnson’s Great Society programs, the implementation of funding optimization techniques helped to quell some of the criticism. Increasingly, funding optimization and “strategic” resource allocation became managerial values throughout American government, and the influence of quantitative analysts grew.^{Footnote 30}

Systems analysis and PPBS had never lacked for critics; strong critiques of RAND’s approach had emerged in its earliest work for the Air Force, and those critiques grew only more trenchant as systems analysts moved into other areas of policy where the problems and constraints were less well defined and far more difficult to mathematize. Skeptics pointed to the narrow framework that systems analysis adopted, its inability to address political, behavioral, and moral issues (including issues of distributional justice), and its tendency to centralize political power among small groups of like-minded experts who lacked practical knowledge and experience of the issues. Moreover, despite its claims to “objectivity,” even some sympathetic critics noted that systems analysis inevitably embedded the analyst’s own value judgments within its quantifications. The disastrous unfolding of the Vietnam War emboldened critics of RAND’s “whiz kids” and their role in forging America’s military policies; by the end of the 1960s, RAND analysts found themselves on the defensive. Yet even as systems analysis and PPBS lost some of their luster, the methods were in many cases already entrenched. While systems analysts lost much of their influence in defense planning, partisans of the method moved into new areas while carving out a seemingly more modest role for their techniques. In 1967, RAND initiated a department of “System Sciences” focused on domestic policymaking and began to seek new clients among state and local governments. Acknowledging that quantitative methods could not capture all key elements of a problem, they argued nonetheless that systems analysis offered a superior means of defining a policy choice and identifying what additional information was needed. As a RAND researcher put it: “Systems analysis is a way of addressing strategic decision problems that emphasizes explicitness, quantification wherever appropriate, recognition of uncertainties, and sharpening—rather than replacing—the judgments of the decision-makers.”^{Footnote 31} Who could quarrel with that?

Howard Raiffa, a mathematician who held appointments in Harvard’s Business School and Economics Department, is both an exemplary and important figure in the extension of new quantitative techniques to environmental policy problems. Raiffa obtained his Ph.D. in mathematics from the University of Michigan and was influenced early in his career by developments in operations research and game theory. In 1957, Raiffa co-authored the book that would introduce game theory to social scientists. After moving to Harvard, Raiffa generalized fault-tree analysis to business management decisions, while developing techniques to incorporate subjective probabilities into the analysis where data was lacking; the goal was to maximize “expected value” in the lack of complete information. Raiffa labeled his method “decision analysis.” In the mid-1960s, Raiffa was invited to spend time at RAND, a period in which that organization was expanding into new areas of domestic policy. While at RAND, Raiffa began to extend his methods beyond business decisions to broader problems of public policy, such as transportation planning and medicine.^{Footnote 32}

In the late 1960s, this style of systems analysis—more modest in some of its claims and flexible in its methods, and yet more ambitious in terms of its potential applicability—became the basis for a new kind of graduate professional training in the social sciences. As Susan Amadee has observed, “the basic set of decision-theoretic tools developed at RAND in the 1950s were the conceptual foundation for the new field of ‘public policy’ in the late 1960s” and a new discipline of “policy analysis” that leaned heavily on economics, statistics, and game theory. Many former RAND people moved into key academic positions, while others wrote academic papers and textbooks. Raiffa, along with colleagues in economics, statistics, and political science, helped to found Harvard’s Kennedy School of Government and its new professional degree in “public policy.” With generous funding from the Ford Foundation, similar programs soon emerged at Berkeley, Duke, Penn, at Rand itself, and other leading schools. The curricula were remarkably similar, emphasizing quantitative methods, microeconomics, statistics, and organizational analysis—methods that were assumed to be broadly applicable across all areas of policymaking. (Raiffa’s 1968 textbook, Introductory Lectures in Decision Analysis, became a key text.) As the Kennedy School founders emphasized, their intent was not to train students in any particular area of public policy, but to train them to work anywhere in the federal government. Existing programs in public administration and political science were increasingly seen as old-fashioned and unscientific, while the Kennedy School emerged as “the pre-eminent institution for the study of decision making, ultimately overshadowing RAND itself.”^{Footnote 33}

Advocates of quantitative policy analysis saw the goal of their work as nothing less than the transformation of governance through the infusion of “rationality” into decision making. The new policy analysts viewed bureaucracies as inherently obstructive and agency professionals as caught up inefficient behavior; hence the need to centralize planning and decision making at the top among those trained in rigorous, disciplined thinking. But theirs was a very particular kind of rationality. Rather than assuming the reasoned exercise of judgment in context, this “Cold War rationality” assumed that, given human limitations, good decisions emerged from quantification, rule-based procedures, and the numerical evaluation of alternative outcomes. Good decision makers were those able to dispassionately assess the relevant variables and then apply rigorous scientific methods to arrive at a solution.^{Footnote 34}

Policy analysts were clear about the need to subordinate technical experts to those better versed in the decision sciences; that had been the very basis of McNamara’s defense department reforms in the 1960s. Although subject experts had a role to play in assembling the necessary data, the exercise of judgment was reserved for those appropriately trained and able to detach themselves from parochial interests. “Specialists,” Raiffa wrote, should “stick to their craft.” Hence “policy analysis” advocated the compartmentalization of technical assessment; in the process, authority was transferred into the hands of a new breed of agency managers. The complexity and quantitative nature of the techniques lent prestige to their practitioners while establishing a common language among analysts working on a diverse array of issues—a new lingua franca of the upper levels of the bureaucracy.^{Footnote 35}

The rise of “policy analysis” in academic centers coincided with the rise of environmental issues in American politics; thus it was all but inevitable that academics in the field would promote their own methods as part, and often the major part, of the answer. Raiffa actively promoted his methods, offering summer workshops for academics from different institutions and disciplines and playing a key role in the founding of a multilateral institute to address global problems through systems analysis. Nuclear scientists had already embraced these methods (Rasmussen’s Reactor Safety Study had relied on Raiffa’s decision trees), and this group quickly gravitated to the emerging field of environmental policy analysis, while policy analysts gravitated to the study of environmental policy with the belief that their overarching analytical frameworks could easily accommodate the range of environmental problems. Harvey Brooks, former COSPUP chair and a member of Harvard’s Applied Physics Department, would migrate to the Kennedy School.^{Footnote 36} And by the early 1970s, students of Raiffa’s were applying their methods to questions of air pollution and emphasizing the broad applicability of decision analysis to complex environmental problems. In 1976, Raiffa and several others with close ties to RAND would be tapped by the Ford Foundation to serve on an important nuclear energy policy committee.^{Footnote 37}

The following year Raiffa was invited to join a National Research Council committee charged with examining the EPA’s use of scientific and technical information. As part of that process, Raiffa and two colleagues prepared an eighty-page background paper enthusiastically advocating the use of quantitative decision methodologies in health and environmental policy. That paper argued for the necessity of quantifying all aspects of a problem (including the value of life), and concluded that rational decisions required that decision makers not allow their perspective to be clouded by emotional concerns that might arise from their own accountability for bad outcomes (e.g., increased mortality). The authors assumed, moreover, that economic and environmental goals were “competing objectives” whose risks and benefits needed to be compared. Several members of the committee were partisans of Raiffa’s approach, and his influence is evident in several places; yet overall, the final report incorporated Raiffa’s ideas only selectively. But Raiffa did not give up. Just two years later, he was invited to chair another NAS committee. Under the auspices of COSPUP, the Committee on Risk Assessment and Decision-Making in Government was strongly supportive of the concepts of quantitative risk analysis from the outset. The staff leader of the study was James Vaupel, a former student of Raiffa’s and then on the faculty of Duke’s new public policy program. Vaupel and Raiffa selected the committee members; most were academics, and several were their close colleagues and committed systems analysts. Raiffa embraced the committee’s task, and over the next eighteen months he and committee-staffer John Graham, then a public policy student at Duke University, produced a draft report of seven hundred pages, demonstrating how quantitative decision analysis could be applied to a wide variety of health and safety problems in order to mathematize trade-offs and yield more “rational” decisions. That document, however, met strong resistance during the peer-review process. The committee ended up publishing a mere fifty-page summary, and even that short document intimated at substantive disagreements within the committee and failed to achieve consensus. Raiffa later recalled the experience as a major disappointment.^{Footnote 38} However, in the report’s summary, Raiffa underscored the need to clearly separate the assessment of risks from their management, and to assess risks “objectively.” Equally important, a key participant on that committee was William Ruckelshaus, the first administrator of the EPA under President Richard Nixon and at the time a vice president for Weyerhaeuser Corporation—though Ruckelshaus would soon return to the EPA. As Raiffa recalled years later, Ruckelshaus “was most enthusiastic about the original draft report.”^{Footnote 39}

“Societal risk assessment,” as Raiffa and others had come to call it, was not exactly systems analysis, but it shared and extended several aspects of the RAND approach: belief in the ability to separate “objective” from “subjective” analysis (and the elevation of the former), the need to break a problem down into discrete steps to facilitate quantification and analysis, the emphasis on formal rule-based procedures, and quantification to the maximum extent possible—even where objective data might be lacking or when quantification might be perceived as inappropriate or impossible. Even though its partisans refrained publicly from trying to quantify the costs of higher levels of risk, the need to place some sort of value on human health and human life was understood and was implicit in the methods. From the perspective of serious policy analysts, the valuation of life was not something that could or should be avoided; whatever resistance the public might hold to doing so was understood as irrational.^{Footnote 40}

Crucially, the appeal of this approach to decision making crossed classic political lines, at least at the upper levels of the bureaucracy. For many liberals and conservatives alike, these methods offered the ability to assess policy in a “neutrally competent” manner. While introduced into the federal government by democratic presidents, modified forms of systems analysis found support in the Nixon and Ford administrations, and many of its precepts, especially the extension of cost-benefit analysis into new areas, would be championed by President Ronald Reagan. While William Ruckelshaus, a Republican appointee with close ties to industry, pushed formalized risk assessment at the EPA, it was Donald Kennedy, a Carter appointee with strong academic and environmental credentials, who introduced it to the FDA. In the early 1970s, Howard Raiffa, with the help of NAS president Philip Handler, would help found the International Institute for Applied Systems Analysis to address international environmental problems with the belief that these supposedly value-neutral, technical methods could bridge even the East-West divide.^{Footnote 41}

The Chemical Industry Discovers Risk

It is perhaps not obvious that the formalized approach of quantitative risk analysis and decision making would appeal to the leaders of regulated industries; it was not initially obvious to many industry leaders. Since the onset of federal environmental regulation, industries had fought hard to oppose or weaken proposals by arguing, often vehemently, about cost. But these were neither popular nor persuasive arguments during the 1960s and 1970s. Where arguments about cost had fallen short, industries had challenged agency procedures along with the scientific basis for specific contaminant limits; but, as several scholars have shown, courts initially accorded considerable deference to agency experts in establishing appropriate levels of pollution and safety. Meanwhile, industry, particularly the chemical industry, had seen its reputation suffer badly. By the mid-1970s, at least some leaders in the chemical industry recognized the need to respond to regulations proactively and to do so in a persuasive and much more politically astute way.^{Footnote 42}

A key moment came in the mid-1970s, in the debate over regulating chemical carcinogens. In 1977, the Occupational Safety and Health Administration (OSHA) promulgated its so-called generic policy on carcinogens in the workplace; at the same time, OSHA issued a strict emergency standard for benzene. For industry, OSHA’s proposed policy was alarming in two respects: (1) it sought to regulate chemicals by class rather than individually; and (2) it relied on forthrightly conservative assumptions about chemical health effects, which resulted in low, sometimes very low, allowable levels—as in the case of benzene. At the same time, a federal interagency group formed to address the problem of carcinogens in a consistent manner (the Interagency Regulatory Liaison Group, or IRLG) was developing guidelines that called for conservative assumptions where data were lacking in order to avoid underestimating public health risk.^{Footnote 43}

In response, chemical industry leaders began to organize a very broad coalition of industries under the newly formed American Industrial Health Council (AIHC). The AIHC immediately challenged the OSHA policy. Importantly, its response did not raise the issue of cost. Instead, learning from the example of the tobacco industry, the AIHC, along with the Chemical Manufacturers Association, decided to publicly embrace science while trying to curtail the influence of scientists in the regulatory agencies. The AIHC also initiated a public relations campaign. Specifically, the AIHC argued that “determination of carcinogenicity is a scientific, not a regulatory, question” and called for an independent committee of scientists to assess potentially carcinogenic chemicals. They attacked both the OSHA and IRLG guidelines for “mixing regulatory considerations into a scientific function.”^{Footnote 44}

The belief that it was both possible and necessary to separate technical and scientific questions from those of politics and economics was the essence of RAND-inspired policy analysis, as was the “honest” (rather than conservative) assessment of threats. “Probabilistic reports about adverse consequences to health are very often slanted to be conservative,” observed Howard Raiffa, but “prudence should appropriately be represented in the evaluation process—not in the assessment process.”^{Footnote 45}

Although it was academic policy analysts who first pressed for this distinction, industry advocates quickly recognized its utility. The appeal lay in removing technical decisions from agency staff who typically took (and were sometimes statutorily required to take) a conservative stance vis-à-vis health effects. When relying on quantitative risk assessment amid so many unknowns, the devil was in the details. Small changes in assumptions could have large impacts on results—determining, for instance, whether or not a given chemical could meet the government’s risk standard. But while government scientists insisted that judgments should be conservative with respect to health (e.g., to account for sensitive populations, or unknown synergistic effects), both quantitative policy analysts and industry representatives argued that the insistence on conservatism “mixed regulatory functions into the scientific function.” In contrast, the separation of “risk assessment” from “risk management” allowed for the use of less conservative assumptions on the basis that quantitative estimates should not be “biased” by political concerns. Conservatism, if warranted, could then be implemented later in the process. For industry, the key was to make the numerical—and supposedly “objective”—calculation of risks less conservative. It would then be far easier to trivialize those risks and to argue that the costs to industry were not warranted for only relatively small and uncertain improvements in health.^{Footnote 46}

Not surprisingly, the AIHC found support for its new position among Harvard’s policy analysts. Richard Zeckhauser, a former student and now a colleague of Raiffa’s at the Kennedy School, testified against the OSHA proposal, urging a more “rational approach, as did Richard Wilson, a Harvard nuclear physicist who had also gravitated to policy analysis. Howard Raiffa and James Vaupel testified before Congress, along with AIHC and the National Association of Manufacturers, in favor of a federal mandate that agencies conduct comparative risk assessments as part of their decision-making process. AIHC also employed William Rowe, now the head of a new Institute on Risk housed in the business school at American University. There would be many more such collaborations. Having discovered the academic study of risk and decision making, chemical industry representatives quickly recognized it as a potentially useful science. And much like the toxicological research of the interwar period, emerging centers for risk analysis would be university-based but largely corporate-funded.^{Footnote 47}

The rapidity with which these ideas were accepted surprised even industry representatives. As the AIHC Chairman reported to the membership in 1980, “In a relatively short time, the concept of separating scientific decision making from regulatory policy-setting has gained favor.” Funding for a NAS study was included in that year’s appropriation for the FDA. Moreover, AIHC was “being heard—and believed—in the right places, by the right people.”^{Footnote 48}

The Triumph of Risk: The Red Book

The AIHC had reason to favor a broad study of risk and environmental policy under the auspices of the National Academy. Not only had COSPUP endorsed the separation of risk assessment from risk management and the quantitative comparison of risks, but industry leaders and their supporters viewed the NAS and the Office of Science and Technology Policy as counterweights to the regulatory agencies and a possible locus for generating regulatory “reform.”^{Footnote 49} The chair of the fourteen-member committee on Risk and Decision-Making in the Federal Government was Reuben Stallones, an epidemiologist from the University of Texas whose research focused on the role of diet and lifestyle in cancer. Stallones was skeptical of the role of chemicals in cancer causation; he had, in fact, recently been employed by AIHC to critique a landmark government on that very topic. Several others on the committee were already committed to the distinction between quantitative assessment and management.^{Footnote 50}

The committee’s starting point was the work of Raiffa’s panel, although this time the committee’s task was much narrower: it addressed risk assessment only with respect to toxic substances and public health. The final report, Risk Assessment in the Federal Government, offered little that was new and stopped short of endorsing AIHC’s science panel proposal. Nonetheless, it became “one of the most influential reports in the 140-year history of the National Academy of Sciences and its affiliated institutions”; by 1996, it was in its thirteenth printing. The most significant legacies of the report were its endorsement of the concept of risk, the codification of a four-stage risk-assessment process, and, most important, its recommendation that regulatory agencies “establish and maintain a clear conceptual distinction between assessment of risks and consideration of risk management alternatives.”^{Footnote 51}

When read carefully, however, the “Red Book,” as it came to be called, is a more interesting and conflictual document. As participants later noted, like any consensus document, it reflected “compromises.” Thus, although the executive summary clearly distinguished between risk assessment and management, objective and subjective tasks, that distinction is challenged and qualified throughout the nearly 200-page report. The committee, for instance, acknowledged “the inherent mixture of science and policy in risk assessment” while distinguishing its own meaning of risk assessment from a narrower “quantitative risk assessment.” Far more important than what the Red Book said, however, was how it was received; in fact, several committee members later referred to it as “the mis-read book.” That reception was shaped largely by the publicity campaigns of the AIHC, which continued to push for an independent “science panel,” as well as by a key individual: William Ruckelshaus.^{Footnote 52}

The Red Book was released in February. Ruckelshaus returned to the EPA that April, taking over in the wake of the bureaucratic and political crisis created by the Reagan administration. Shortly after his reappointment, Administrator Ruckelshaus gave a talk at the National Academy of Sciences, highlighting the report and setting out what would become his major initiative during his second term at the EPA: strengthening risk assessment. Ruckelshaus, a veteran of Raiffa’s earlier committee, was hardly new to the idea. Moreover, he was quite familiar with industry’s perspective. During his hiatus from government, Ruckelshaus formed a law firm that provided counsel to both the Society of Plastics Industries and the AIHC; subsequently, he served as a vice president at the Weyerhaeuser Corporation. All these groups were facing potentially crippling regulations on carcinogenic chemicals (vinyl chloride, benzene, and formaldehyde).^{Footnote 53} In his Academy speech, Ruckelshaus glossed over the Red Book’s qualifications and caveats, telling his audience that “the NAS report recommends that these two functions-risk assessment and risk management-be separated as much as possible within a regulatory agency.” In various speeches and articles, Ruckelshaus would continue in this vein. As one committee participant later reflected, Ruckelshaus “encouraged misinterpretation of what the Red Book actually said.”^{Footnote 54} Although the Red Book called into question the validity of any simple distinction between risk assessment and risk management, this distinction became the accepted, though heavily criticized, way of structuring the regulatory process.

Long after the initial enthusiasm for PPBS and formal systems analysis began to wane, and the siting of commercial nuclear power reactors had ground to a halt, the EPA and other agencies embraced the numeric comparison of risks and the distinction between “scientific” assessment and “political” management. In 1987, the EPA published a seminal report that boldly asserted that the agency’s “fundamental mission” was to “reduce risks” and then went on to rank the relative risks posed by various environmental pollutants. Subsequently, Administrator William Reilly announced that the EPA would use risk analysis not only to establish standards but also to set internal priorities; he required the EPA offices to submit risk-based budget proposals.^{Footnote 55} Although RAND’s PPBS system had been officially discontinued by the Nixon administration in 1970, a decade and a half later Reilly had rejuvenated it by another name.

The turn to risk analysis within the federal bureaucracy, even when it is criticized, has typically been portrayed as a logical, even inevitable, outgrowth of advances in science and technology. Chemicals had become more visible and more ubiquitous, rendering earlier notions of safety and human rights unworkable.^{Footnote 56} But the ideas and basic methods that underlie risk assessment arose not in response to environmental policy problems, but elsewhere—in the planning and operation of modern warfare. The language of risk and the methods of risk analysis migrated into domestic policy along with nuclear power and its advocates; although these methods met with strong public resistance and intellectual critique, they were extended to chemical safety when policy academics advocated the application of their own quantitative approaches. Key institutions—the RAND Corporation, the NAS, the Kennedy School—fostered the work of individuals with similar intellectual backgrounds during the 1970s and 1980s and encouraged the elaboration of decision-theoretic approaches and the marginalization of critics. The advocacy and teaching of these approaches within leading schools of public policy underscored their intellectual credibility. In other words, the turn toward risk in environmental and health policy owed much to a particular group of policy intellectuals steeped in economic and mathematical approaches to decision making, and who were determined to spread their methods.

In an influential book, historian Theodore Porter has argued that the turn toward quantification in government policy often stemmed from the weakness of American bureaucrats; they turned to numbers in order to bolster their authority. However, this particular turn toward quantification did not unambiguously bolster the authority of bureaucrats; rather, within the realm of environmental policy, it worked to bolster the authority of certain bureaucrats and disciplines at the expense of others.^{Footnote 57} Unlike the earlier ideas of purity and safety, the paradigm of “risk” had no basis in a biological model of health; its intellectual origins lay in statistics, mathematics, and game theory, and these were the disciplines that fostered its experts. In contrast, turn-of-the-century (c. 1900) food regulations derived from medical bacteriology and assumed that healthy environments were pure and unpolluted, and it lodged authority with chemists. The turn toward “safety” acknowledged the unavoidable presence of some chemical hazards in the modern world; those hazards, however, were finite and subject to control; moreover, it was the responsibility of toxicologists and pharmacologists to ensure that chemical exposures did not harm human health. Environments might not be pure, but if pollutants were held below a given level, they would be biologically safe. True to its roots in nuclear weapons planning, the paradigm of risk imagined the world as always and unavoidably hazardous, while never asking why. The question was how best to navigate the terrain in order to maximize the expected payoff. Issues of human rights and biology, while never fully erased, became secondary to optimizing scarce resources.

There is another, related difference. The idea of risk implies that there is a potential gain as well as a potential loss; and, moreover, that gains can only be realized if risks are accepted. To refuse risk is to invite stagnation, or, worse, the prospect of even greater risk.^{Footnote 58} Although the gains to a manufacturer of producing chemicals are clear and measurable in terms of profits, the potential gains to exposed individuals are often minimal or nonexistent and not so easily measured. In the case of chemicals, the assessment of benefits for the purposes of regulation would be left to the industries themselves, and those benefits would often be glossed as collective gains to the national economy.^{Footnote 59} In this way, the regulatory discourse of risk came to accept and institutionalize the opposition between society’s economic welfare, on the one hand, and health and environmental protection, on the other. Put another away, the binary logic of the Cold War was inscribed onto American environmental policy. Built into the regulatory framework of risk assessment and much of late twentieth-century “environmental policy analysis” is the assumption that improvements in economic well-being must typically be traded off against environmental and health protection. The only question worth asking, then, is, What is the optimal trade-off? In this way, the seemingly irrational perspective that improvements in human well-being must come at the cost of human health and the natural environments on which the human species depends emerged as a wholly rational proposition.