824 F.2d 1146 | D.C. Cir. | 1987
Opinion for the Court filed by
Current scientific knowledge does not permit a finding that there is a completely safe level of human exposure to carcinogenic agents. The Administrator of the Environmental Protection Agency, however, is charged with regulating hazardous pollutants, including carcinogens, under section 112 of the Clean Air Act by setting emission standards “at the level which in his judgment provides an ample margin of safety to protect the public health.” 42 U.S.C. § 7412(b)(1)(B) (1982). We address here the question of the extent of the Administrator’s authority under this delegation in setting emission standards for carcinogenic pollutants.
Petitioner Natural Resources Defense Council (“NRDC”) contends that the Administrator must base a decision under section 112 exclusively on health-related factors and, therefore, that the uncertainty about the effects of carcinogenic agents requires the Administrator to prohibit all emissions. The Administrator argues that in the face of this uncertainty he is authorized to set standards that require emission
I.
Section 112 of the Clean Air Act provides for regulation of hazardous air pollutants, which the statute defines as “air pollutants] to which no ambient air quality standard is applicable and which in the judgment of the Administrator cause[ ], or contribute^ to, air pollution which may reasonably be anticipated to result in an increase in mortality or an increase in serious irreversible, or incapacitating reversible, illness.” 42 U.S.C. § 7412(a)(1) (1982). The statute requires the Administrator to publish a list containing each hazardous pollutant for which he intends to adopt an emission standard, to publish proposed regulations and a notice of public hearing for each such pollutant, and then, within a specified period, either to promulgate an emission standard or to make a finding that the particular agent is not a hazardous air pollutant. See id. § 7412(b)(1)(B). The statute directs the Administrator to set an emission standard promulgated under section 112 “at the level which in his judgment provides an ample margin of safety to protect the public health.” Id.
This case concerns vinyl chloride regulations. Vinyl chloride is a gaseous synthetic chemical used in the manufacture of plastics and is a strong carcinogen. In late 1975, the Administrator issued a notice of proposed rulemaking to establish an emission standard for vinyl chloride. 40 Fed. Reg. 59,532 (1975). In the notice, the EPA asserted that available data linked vinyl chloride to carcinogenic, as well as some noncarcinogenic, disorders and that “[r]easonable extrapolations” from this data suggested “that present ambient levels of vinyl chloride may cause or contribute to ... [such] disorders.” Id. at 59,533. The EPA also noted that vinyl chloride is “an apparent non-threshold pollutant,” which means that it appears to create a risk to health at all non-zero levels of emission. Scientific uncertainty, due to the unavailability of dose-response data and the twenty-year latency period between initial exposure to vinyl chloride and the occurrence of disease, makes it impossible to establish any definite threshold level below which there are no adverse effects to human health. Id. at 59,533-34. The notice also stated the “EPA’s position that for a carcinogen it should be assumed, in the absence of strong evidence to the contrary, that there is no atmospheric concentration that poses absolutely no public health risk.” Id. at 59,534.
Because of this assumption, the EPA concluded that it was faced with two alternative interpretations of its duty under section 112. First, the EPA determined that section 112 might require a complete prohibition of emissions of non-threshold pollutants because a “zero emission limitation would be the only emission standard which would offer absolute safety from ambient exposure.” 40 Fed.Reg. at 59,534. The EPA found this alternative “neither desirable nor necessary” because “[c]omplete prohibition of all emissions could require closure of an entire industry,” a cost the EPA found “extremely high for elimination of a risk to health that is of unknown dimensions.” Id.
The EPA stated the second alternative as follows:
An alternative interpretation of section 112 is- that it authorizes setting emission standards that require emission reduction to the lowest level achievable by use of the best available control technology in cases involving apparent non-threshold pollutants, where complete emission prohibition would result in widespread industry closure and EPA has determined that the cost of such closure would be grossly disproportionate to the benefits of removing the risk that would remain after imposition of the best available control technology.
On October 21, 1976, the EPA promulgated final emission standards for vinyl chloride which were based solely on the level attainable by the best available control technology. 41 Fed.Reg. 46,560 (1976). The EPA determined that this standard would reduce unregulated emissions by 95 percent. Id. With respect to the effect of the standard on health, the EPA stated that it had assessed the risk to health at ambient levels of exposure by extrapolating from dose-response data at higher levels of exposure and then made the following findings:
EPA found that the rate of initiation of liver angiosarcoma among [the 4.6 million] people living around uncontrolled plants is expected to range from less than one to ten cases of liver angiosarcoma per year of exposure to vinyl chloride____ Vinyl chloride is also estimated to produce an equal number of primary cancers at other sites, for a total of somewhere between less than one and twenty cases of cancer per year of exposure among residents around plants. The number of these effects is expected to be reduced at least in proportion to the reduction in the ambient annual average vinyl chloride concentration, which is expected to be 5 percent of the uncontrolled levels after the standard is implemented.
Id. The EPA did not state whether this risk to health is significant or not. Nor did the EPA explain the relationship between this risk to health and its duty to set an emission standard which will provide an “ample margin of safety.”
The Environmental Defense Fund (“EDF”) filed suit challenging the standard on the ground that section 112 requires the Administrator to rely exclusively on health and prohibits consideration of cost and technology. The EDF and the EPA settled the suit, however, upon the EPA’s agreement to propose new and more stringent standards for vinyl chloride and to establish an ultimate goal of zero emissions.
The EPA satisfied its obligations under the settlement agreement by proposing new regulations on June 2, 1977. While the proposal sought to impose more strict regulation by requiring sources subject to a 10 parts per million (“ppm”) limit to reduce emissions to 5 ppm, and by establishing an aspirational goal of zero emissions, the EPA made it clear that it considered its previous regulations valid and reemphasized its view that the inability scientifically to identify a threshold of adverse effects did not require prohibition of all emissions, but rather permitted regulation at the level of best available technology. 42 Fed.Reg. 28,154 (1977). The EPA received comments on the proposal, but took no final action for more than seven years. On January 9, 1985, the EPA withdrew the proposal. Noting that certain aspects of the proposed regulations imposed “unreasonable” costs and that no control technology “has been demonstrated to significantly and consistently reduce emissions to a level below that required by the current standard,” 50 Fed.Reg. 1182, 1184 (1985), the EPA concluded that it should abandon the 1977 proposal and propose in its place only minor revisions to the 1976 regulations.
This petition for review followed.
II.
We must address at the outset two procedural challenges to the NRDC’s petition for review. First, an industry intervenor, the Vinyl Institute, argues that the petition for review is not timely filed. Second, the EPA argues that the NRDC has failed to exhaust its administrative remedies and that we must, therefore, dismiss the petition for review.
A.
The Vinyl Institute argues that this court does not have jurisdiction because the statute provides that “[a]ny petition for review ... shall be filed within sixty days from the date notice of [the] promulgation, approval, or action appears in the Federal Register, except that if such petition is based solely on grounds arising after such
The contention that this case is a “backdoor” challenge to the 1976 regulations is refuted by the substance of the petitioner’s brief and the relief requested. The petitioner states that “[i]n withdrawing the proposed amendments the EPA violated the law by employing cost-benefit and technological feasibility tests that are prohibited by the Clean Air Act.” Brief for NRDC at 3. Indeed, the brief makes explicit that the petitioner is specifically challenging the EPA’s reliance on cost and technological feasibility in its withdrawal of the proposed amendments. Id. at 11-13. Additionally, the petitioner does not ask the court to overturn the 1976 standards, but rather asks the court to vacate the EPA’s decision to withdraw the amendments. See Brief for NRDC at 36-37. We think it clear, therefore, that the NRDC has challenged the 1985 withdrawal of the proposed amendments. The petition for review is timely.
B.
The EPA argues that the petitioner has failed to exhaust available administrative remedies because it failed to participate in the proceedings below. Congress included in section 307(d) of the Clean Air Act a statutory requirement of exhaustion, which provides that “[o]nly an objection to a rule or procedure which was raised with reasonable specificity during the period for public comment (including any public hearing) may be raised during judicial review.” 42 U.S.C. § 7607(d)(7)(B) (1982). This statutory requirement of exhaustion, however, does not apply here. The statute also provides that “[t]he requirements of ... subsection [307(d) of the Act] shall take effect with respect to any rule the proposal of which occurs after ninety days after August 7, 1977.” 42 U.S.C. § 7607(d)(ll) (1982). The withdrawal of the proposed amendments is the final EPA action on a notice of proposed rulemaking that issued on June 2, 1977. See 42 Fed.Reg. 28,154 (1977). Thus, even if we assume that the action of withdrawing a proposed rule amounts to a “rule” for purposes of section 307(d), the proposal withdrawn here was issued before section 307(d) took effect. Accordingly, we must look to the common law doctrine of exhaustion of remedies. See Safir v. Kreps, 551 F.2d 447, 452 (D.C. Cir.), cert. denied, 434 U.S. 820, 98 S.Ct. 60, 54 L.Ed.2d 76 (1977). The result, however, is the same.
Courts have long required a party seeking review of agency action to exhaust its administrative remedies before seeking judicial review. See, e.g., Myers v. Bethlehem Shipbuilding Corp., 303 U.S. 41, 50-51, 58 S.Ct. 459, 463, 82 L.Ed. 638 (1938). In this case, the administrative remedy was participation in the rulemaking proceedings during the comment period. Indeed, this court generally requires such participation as a prerequisite to a petition for direct review of the resulting regulations. See Environmental Defense Fund v. EPA, 598 F.2d 62, 91 (D.C.Cir.1978).
The NRDC did not participate in the rule-making proceedings in this case, but ar
In this case, the issue of whether the EPA may set a standard under section 112 on the basis of cost and technological feasibility was raised before the agency. First, the 1977 proposed amendments were the product of the settlement of a lawsuit challenging the previous vinyl chloride standards on the ground that the EPA impermissibly considered these factors. The EPA, therefore, had notice of this issue and could, or should have, taken it into account in reaching a final decision on the proposed amendments. Indeed, in its notice of proposed rulemaking, the EPA remarked that “[t]he [1976] vinyl chloride standard has been criticized for allegedly placing unwarranted emphasis on technological rather than health considerations.” 42 Fed.Reg. 28,154 (1977). The notice then continued by discussing the “ample margin of safety” language, the potential problem under this standard of shutting down an entire industry that produces a nonthreshold pollutant, and the way the proposed amendments resolved the problem by moving toward zero emissions without banning vinyl chloride. Id. Thus, it is clear that the EPA actually did consider the issue raised by the NRDC in its petition for review.
Moreover, the EDF explicitly raised the issue before the EPA in its comments on the proposed amendments. In this respect, the EDF stated:
The proposed amendments represent a true compromise between what EDF could have pressed for in court and the existing standard. Section 112 of the Clean Air Act requires that emission standards for hazardous air pollutants ... be set “at a level which in the judgment of the Administrator provides an ample margin of safety to protect the public health from such hazardous air pollutants.” It clearly requires a health-linked, not a technology-based standard. Yet, inconsistent with the statutory requirement, the original standards were based on what EPA believed industry could accomplish with best available technology____ EPA recognized that vinyl chloride is “an apparent non-threshold pollutant” which creates a risk to public health at all levels. Had the case gone to trial, EDF would have taken the position that § 112 required a zero emission standard, the only standard adequate to provide the required margin of safety for a non-threshold pollutant. Instead, EDF settled for a compromise which establishes a goal of zero emissions and requires industry to move one step closer to that goal.
J.A. at 72-73. The EDF’s comments contain other similar references, such as the assertion, in response to cost arguments raised by the industry, that “the statute EPA operates under requires regulations based on protection of health and not cost and technology concerns.” Id. at 83. Thus, the EPA had before it the question of whether the statute permits considerations of cost and technology in setting standards, and it had the opportunity to consider that question in deciding to withdraw the proposed amendments.
The EPA also suggests, however, that we should be “especially” loath to allow this petition for review because the “NRDC
III.
The NRDC’s challenge to the EPA’s withdrawal of the 1977 amendments is simple: because the statute adopts an exclusive focus on considerations of health, the Administrator must set a zero level of emissions when he cannot determine that there is a level below which no harm will occur.
We must determine whether the EPA’s actions are arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law. 42 U.S.C. § 7607(d)(9)(A) (1982). Review begins with the question of whether “Congress has directly spoken to the precise question at issue” and has expressed a clear intent as to its resolution. Chevron, U.S.A., Inc. v. NRDC, 467 U.S. 837, 842-43, 104 S.Ct. 2778, 2781, 81 L.Ed.2d 694 (1984). If so, “that intention is the law and must be given effect.” Id. at 843 n. 9, 104 S.Ct. at 2781 n. 9. “[I]f the statute is silent or ambiguous with respect to the specific issue,” we must accept an agency interpretation if it is reasonable in light of the language, legislative history, and underlying policies of the statute. Id. at 843, 104 S.Ct. at 2782; NRDC v. Thomas, 805 F.2d 410, 420 (D.C.Cir.1986). We find no support in the text or legislative history for the proposition that Congress intended to require a complete prohibition of emissions whenever the EPA cannot determine a threshold level for a hazardous pollutant. Instead, there is strong evidence that Congress considered such a requirement and rejected it.
Section 112 commands the Administrator to set an “emission standard” for a particular “hazardous air pollutant” which in his “judgment” will provide an “ample margin of safety.” Congress’ use of the term “ample margin of safety” is inconsistent with the NRDC’s position that the Administrator has no discretion in the face of uncertainty. The statute nowhere defines “ample margin of safety.” The Senate Report, however, in discussing a similar requirement in the context of setting ambient air standards under section 109 of the Act, explained the purpose of the “margin of safety” standard as one of affording “a reasonable degree of protection ... against hazards which research has not yet identified.” S.Rep. No. 1196, 91st Cong., 2d Sess. 10 (1970) (emphasis added). This view comports with the historical use of the term in engineering as “a safety factor ... meant to compensate for uncertainties and variabilities.” See Hall, The Control of Toxic Pollutants Under the Federal Water Pollution Control Act Amendments of 1972, 63 Iowa L.Rev. 609, 629 (1978). Furthermore, in a discussion of the use of identical language in the Federal Water Pollution Control Act, this court has recognized that, in discharging the responsibility to assure “an ample margin of safety,” the Administrator faces “a difficult task, in
Congress’ use of the word “safety,” moreover, is significant evidence that it did not intend to require the Administrator to prohibit all emissions of non-threshold pollutants. As the Supreme Court has recently held, “safe” does not mean “risk-free.” Industrial Union Dep’t, AFL-CIO v. American Petroleum Inst., 448 U.S. 607, 642, 100 S.Ct. 2844, 2864, 65 L.Ed.2d 1010 (1980). Instead, something is “unsafe” only when it threatens humans with “a significant risk of harm.” Id.
Thus, the terms of section 112 provide little support for the NRDC’s position. The uncertainty about the effects of a particular carcinogenic pollutant invokes the Administrator’s discretion under section 112. In contrast, the NRDC’s position would eliminate any discretion and would render the standard “ample margin of safety” meaningless as applied to carcinogenic pollutants.
The NRDC also argues that the legislative history supports its position. To the contrary, that history strongly suggests that Congress did not require the Administrator to prohibit emissions of all non-threshold pollutants; Congress considered and rejected the option of requiring the Administrator to prohibit all emissions.
The Senate bill would have required the Administrator to prohibit any emission of a hazardous pollutant, threshold or non-threshold, unless he found, after a hearing, that a preponderance of the evidence demonstrated “that such agent is not hazardous to the health of persons” or that “departure from ... prohibition for [a] stationary source will not be hazardous to the health of persons.” S. 4358, 91st Cong., 2d Sess. § 6(b), 116 Cong.Rec. 32,375 (1970). The definition of hazardous agent included any pollutant “whose presence ... in trace concentrations in the ambient air ... causes or will cause, or contribute to, an increase in serious irreversible or incapacitating reversible damage to health.” Id. Presumably, this provision would have required the complete prohibition of emissions of carcinogenic agents because the Administrator cannot demonstrate by “a preponderance of the evidence” that trace concentrations of these agents will not cause harm. The final version of section 112, however, omits any reference to a
The only arguable support for the NRDC’s position is a passage in the summary of the provisions of the conference agreement attached to Senator Muskie’s statement during the post-conference debate on the Clean Air Act:
The standards must be set to provide an ample margin of safety to protect the public health. This could mean, effectively, that a plant could be required to close because of the absence of control techniques. It could include emission standards which allow for no measurable emissions.
Senate Consideration of the Report of the Conference Committee, Exhibit 1 to Statement of Sen. Muskie, Congressional Research Service of the Library of Congress, 93d Cong., 2d Sess., 1 A Legislative History of the Clean Air Amendments of 1970 at 133 (Comm.Print 1974). This statement does not, as the NRDC supposes, mean that the Administrator must set a zero-emission level for all non-threshold pollutants. On its face, the statement means only that, in certain conditions, there may be plant closings and sometimes zero emissions may be required. Senator Muskie did not say this would invariably be so when scientific uncertainty existed. His statement confirms that the Administrator is permitted to set a zero-emission level for some pollutants; it does not hold that the Administrator is invariably required to do so whenever there is some scientific uncertainty.
It is also significant that this is the only reference after the Conference Committee compromise to the possibility of plant closure as a result of the Administrator’s actions under section 112. To accept the petitioner’s contention that section 112 requires the Administrator to prohibit all emissions of non-threshold pollutants, we would have to conclude that, without even discussing the matter, Congress mandated massive economic and social dislocations by shutting down entire industries. That is not a reasonable way to read the legislative history. The EPA has determined that a zero-emissions standard for non-threshold pollutants would result in the elimination of such activities as “the generation of electricity from either coal-burning or nuclear energy; the manufacturing of steel; the mining, smelting, or refining of virtually any mineral (e.g., copper, iron, lead, zinc, and limestone); the manufacture of synthetic organic chemicals; and the refining, storage, or dispensing of any petroleum product.” National Emission Standards for Hazardous Air Pollutants; Policy and Procedures for Identifying, Assessing and Regulating Airborne Substances Posing a Risk of Cancer, 44 Fed.Reg. 58,642, 58,660 (1979). It is simply not possible that Congress intended such havoc in the American economy and not a single representative or senator mentioned the fact. Cf. Industrial Union Dep’t, 448 U.S. at 645, 100 S.Ct. at 2865 (“In the absence of a clear mandate in the Act, it is unreasonable to assume that Congress intended to give the Secretary [this] unprecedented power over American industry.”). Thus, we find no support for the NRDC’s extreme position in the language or legislative history of the Act.
IV.
We turn now to the question whether the Administrator’s chosen method for setting emission levels above zero is consistent with congressional intent. The Administrator’s position is that he may set an emission level for non-threshold pollutants at the lowest level achievable by best available control technology when that level is anywhere below the level of demonstrated harm and the cost of setting a lower level is grossly disproportionate to the benefits of removing the remaining risk. The NRDC argues that this standard is arbitrary and capricious because the EPA is never permitted to consider cost and technological feasibility under section 112 but
A.
On its face, section 112 does not indicate that Congress intended to preclude consideration of any factor. Though the phrase “to protect the public health” evinces an intent to make health the primary consideration, there is no indication of the factors the Administrator may or may not consider in determining, in his “judgment,” what level of emissions will provide “an ample margin of safety.” Instead, the language used, and the absence of any specific limitation, gives the clear impression that the Administrator has some discretion in determining what, if any, additional factors he will consider in setting an emission standard.
B.
The petitioner argues that the legislative history makes clear Congress’ intent to foreclose reliance on non-health-based considerations in setting standards under section 112. We find, however, that the legislative history can be characterized only as ambiguous.
The NRDC directs us to the hazardous air pollutants provision of the House bill, which states that “[i]f ... emissions [from any class of new stationary sources] are extremely hazardous to health, no new source of such emissions shall be constructed or operated, except where (and subject to such conditions as he deems necessary and appropriate) the [Administrator] makes a specific exemption with respect to such construction or operation.” See H.R. 17255, 91st Cong., 2d Sess. § 5(a), 116 Cong.Rec. 19,226 (1970). Thus, as to extremely hazardous emissions, the House bill granted the Administrator a rather open-ended power to exempt a source from the regulation imposed, a power that the petitioner presumes, probably correctly, to have allowed for exemptions on the basis of non-health considerations.
Unlike the House bill, the Senate version dealt only with hazardous air pollutants and did so with respect to all stationary sources. The bill proposed a relatively narrow definition of hazardous agents, restricting this category to pollutants “whose presence, chronically or intermittently, in trace concentrations in the ambient air, either alone or in combination with other agents, causes or will cause, or contribute to, an increase in mortality or an increase in serious irreversible or incapacitating reversible damage to health.” S. 4358, 91st Cong., 2d Sess. § 6(b), 116 Cong.Rec. 32,-375 (1970). Under the scheme set up by the bill, the Administrator was to publish a list of hazardous agents, and follow it by a “proposed prohibition of emissions of each such agent or combination of agents from any stationary source.” Id. The bill then provided for a hearing, after which the Administrator was required to promulgate the prohibition unless a preponderance of evidence demonstrated either “that such agent is not hazardous to the health of persons” or “that departure from prohibition for [a] stationary source will not be hazardous to the health of persons.” Id. If the Administrator found either such condition to exist, he would then implement an emission standard in lieu of a prohibition. Id.
Given these starting points, the inference the petitioner draws from the changes made at conference appears tenuous at best. The final version defines a “hazardous air pollutant” as
an air pollutant to which no ambient air quality standard is applicable and which in the judgment of the Administrator may cause, or contribute to, air pollution which may reasonably be anticipated to result in an increase in mortality or an increase in serious irreversible, or incapacitating reversible, illness.
Clean Air Amendments of 1970, Pub.L. No. 91-604, § 4(a), 84 Stat. 1676, 1685.
C.
The petitioner argues next that a finding that section 112 does not preclude consideration of cost and technological feasibility would render the Clean Air Act structurally incoherent and would be inconsistent with the Supreme Court’s interpretation of section 110 of the Act, see Union Electric Co. v. EPA, 427 U.S. 246, 96 S.Ct. 2518, 49 L.Ed.2d 474 (1976), and this court’s interpretation of section 109 of the Act, see Lead Indus. Ass’n v. EPA, 647 F.2d 1130 (D.C.Cir.), cert. denied, 449 U.S. 1042, 101 S.Ct. 621, 66 L.Ed.2d 503 (1980), as precluding consideration of these factors. We do not believe that our decision here is inconsistent with either the holding or the statutory interpretation in either case.
First, as discussed below, the court in each case rejected an argument that the EPA must consider cost and technological feasibility as factors equal in importance to health. We reject the same argument here. See infra pp. 1163-1166. In this case, however, we must also address the question of whether the Administrator may consider these factors if necessary to further protect the public health. This issue was not addressed in either Union Electric or Lead Industries.
Second, these decisions do not provide precedential support for the petitioner’s position that, as a matter of statutory interpretation, cost and technological feasibility may never be considered under the Clean Air Act unless Congress expressly so provides. In each case there was some indication in the language, structure, or legislative history of the specific provision at issue that Congress intended to preclude consideration of cost and technological feasibility. As discussed above, we find no such indication with respect to section 112.
A similar analysis distinguishes this court’s reasoning in Lead Industries. In Lead Industries, we held that the Administrator is not required to consider cost and technology under the mandate in section 109 of the Clean Air Act to promulgate primary air quality standards which “allow[ ] an adequate margin of safety ... to protect the public health.” 42 U.S.C. § 7409(b)(1) (1982).
The Lead Industries court did note that the statute on its face does not allow con
In Lead Industries, moreover, the relevant Senate Report stated flatly that “existing sources of pollutants either should meet the standard of the law or be closed down.” 647 F.2d at 1149. This is a far clearer statement than anything in the present case that Congress considered the alternatives and chose to close down sources or even industries rather than to allow risks to health.
The substantive standard imposed under the hazardous air pollutants provisions of section 112, by contrast with sections 109 and 110, is not based on criteria that enumerate specific factors to consider and pointedly exclude feasibility. Section 112(b)(l)’s command “to provide an ample margin of safety to protect the public health” is self-contained, and the absence of enumerated criteria may well evince a congressional intent for the Administrator to supply reasonable ones. Further, section 112, in marked contrast to the regime of ambient air standards, operates through nationally enforced standards; the state plans are permissive and may not interfere with national enforcement of any hazardous pollutant standard. 42 U.S.C. § 7412(d) (1982). No detailed provisions preclusive of technological and economic considerations govern the state plans allowed under section 112; indeed, the Administrator must delegate enforcement and implementation authority to the state (subject to his continuing ability to enforce national standards) if he finds the state plan “adequate.” Id. Thus, nothing in the scheme of state implementation plans under section 112 demonstrates disfavor for feasibility considerations, and this further distinguishes section 112 from the Lead Industries court’s interpretation of section 109.
Thus, in Lead Industries, the court found clear evidence that Congress intended to limit the factors the Administrator is permitted to consider in setting a “margin of safety” under section 109. The “margin of safety” standard in section 112 is not so adorned. For that reason, Lead Industries does not control this case.
On the other side of this controversy, the EPA argues that the 1977 amendments of the Clean Air Act, in light of Congress’ awareness of the 1976 vinyl chloride regulations, amounts to a ratification of the use of cost and technological feasibility considerations in setting standards under section 112. We think this overstates the significance of the legislative history leading up to the Clean Air Act Amendments of 1977. To understand why this is the case, and to appreciate what significance, if any, the 1977 amendments have, we turn to an examination of the history of those amendments.
In 1976, both houses of Congress passed bills purporting to amend the Clean Air Act. The first section of the House bill sought to spur the EPA to take action with respect to specified unregulated pollutants, including vinyl chloride. Within one year of the enactment of the amendments, unless the Administrator found after notice and a hearing that the enumerated “substance [would] not cause or contribute to air pollution which [could] reasonably be anticipated to endanger public health,” he was to include such substance on the list of pollutants subject to regulation under an ambient air standard pursuant to sections 108 through 110 or under the hazardous air pollutant provisions of section 112, or to
In addressing this section of the bill, the Report of the Committee on Interstate and Foreign Commerce discussed the vinyl chloride problem in some detail, emphasizing the dangerous nature of the substance. H.R.Rep. No. 1175, 94th Cong., 2d Sess. 23 (1976). During the development of the bill, the Report noted, the EPA had “proposed emission standards for vinyl chlorides under section 112 of the Act for major sources in the plastics industry,” but the Committee retained vinyl chlorides in the proposed legislation to underscore “the Committee’s concern that the standards be promulgated without delay and that standards be promulgated for any other significant sources of vinyl chlorides which may exist.” Id. at 23-24. The Committee also noted, however, that it “did not intend to specify the degree of emission reduction which should be required,” but that “the Administrator should apply the appropriate means and extent of regulation under the existing statutory criteria.” Id. at 26. The House passed the bill, leaving section 101 intact.
The Senate bill contained nothing similar. The Conference Committee, however, decided to adopt the House provision regarding unregulated pollutants in relevant part. See H.R.Rep. No. 1742, 94th Cong., 2d Sess. 25-26 (1976). The threat of a total filibuster, however, prevented the Senate from voting on the recommendations of the conferees, and the bill died.
In 1977, both houses reintroduced legislation to amend the Clean Air Act. In the interim between the abandonment of the 1976 amendments and the introduction of the new legislation, the EPA promulgated emission standards for vinyl chloride under section 112 of the Act. See 41 Fed.Reg. 46,560 (1976). In so doing, the EPA clearly articulated that the regulations adopted reduced vinyl chloride emissions “to the level attainable with best available technology” and that, while “section 112 does not explicitly provide for consideration of costs,” the agency believed it could take them into account for the limited purpose of "as-surfing] that the costs of control technology are not grossly disproportionate to the amount of emission reduction achieved.” Id. at 46,560, 46,562. The 1977 House bill contained a provision “nearly identical” to section 101 of the 1976 House bill, differing primarily in its inclusion of radioactive materials and deletion of vinyl chloride from the unregulated pollutants specified. H.R. Rep. No. 294, 95th Cong., 1st Sess. 3 (1977), U.S.Code Cong. & Admin.News 1977, pp. 1077, 1080-81. The Committee explained the deletion of vinyl chloride on the ground that “[d]uring the past year the Administrator [had] promulgated final regulations for the control of vinyl chloride emissions.” Id.
The bill passed the House with the unregulated pollutants provision intact. Once again, the Senate bill had no such provision, and, at conference, the House’s provision was adopted in relevant part. See H.R.Rep. No. 564, 95th Cong., 1st Sess. 141-42 (1977), U.S.Code Cong. & Admin. News 1977, p. 1077, 1521-23. The conference recommendations regarding unregulated pollutants passed both houses intact, and President Carter signed the Clean Air Act Amendments into law on August 7, 1977. See Pub.L. No. 95-95, 91 Stat. 685 (1977).
The 1977 legislation comprehensively amended the Clean Air Act, and, in fact, amended the very section that is the subject of this lawsuit. Indeed, that amendment added a further subsection employing substantially the language that the EPA had construed in the vinyl chloride regulations to allow consideration of economic and technological feasibility. The relevant amendment to section 112 empowered the Administrator under certain circumstances to forgo use of an emission standard and instead to “promulgate a design, equipment, work practice, or operational standard, or combination thereof, which in his judgment is adequate to protect the public health ... with an ample margin of safe
The House knew of the 1976 regulations,
V.
Since we cannot discern clear congressional intent to preclude consideration of cost and technological feasibility in setting emission standards under section 112, we necessarily find that the Administrator may consider these factors. We must next determine whether the Administrator’s use of these factors in this case is “based on a permissible construction of the statute.” Chevron, 467 U.S. at 843,104 S.Ct. at 2782. We must uphold the Administrator’s construction if it represents “a reasonable policy choice for the agency to make.” Id. We cannot, however, affirm an agency interpretation found to be “arbitrary, capricious, or manifestly contrary to the statute.” Id. at 844, 104 S.Ct. at 2782. Nor can we affirm if “it appears from the statute or its legislative history that the accommodation [chosen] is not one that Congress would have sanctioned.” United States v. Shimer, 367 U.S. 374, 383, 81 S.Ct. 1554, 1560, 6 L.Ed.2d 908 (1961).
Our role on review of an action taken pursuant to section 112 is generally a limited one. Because the regulation of carcinogenic agents raises questions “on the frontiers of scientific knowledge,” Industrial Union Dep’t, AFL-CIO v. Hodgson, 499 F.2d 467, 474 (D.C.Cir.1974), we have recognized that the Administrator’s decision in this area “will depend to a greater extent upon policy judgments” to which we must accord considerable deference. Id.; Lead Indus., 647 F.2d at 1146-47; Environmental Defense Fund, 598 F.2d at 82; Hercules, Inc. v. EPA, 598 F.2d 91, 106 (D.C.Cir.1978). We have also acknowledged that “EPA, not the court, has the technical expertise to decide what inferences may be drawn from the characteristics of ... substances and to formulate policy with respect to what risks are acceptable,” Environmental Defense Fund, 598 F.2d at 83-84, and we will not second-guess a determination based on that expertise. American Petroleum Inst. v. Costle, 665 F.2d 1176, 1184 (D.C.Cir.1981); Lead Indus., 647 F.2d at 1146. Our only role is to determine whether “ ‘the agency has exercised a reasoned discretion, with reasons that do not deviate from or ignore the ascertainable legislative intent.'" Lead Indus., 647 F.2d at 1145 (quoting Greater Boston Television Corp. v. FCC, 444 F.2d 841, 850 (D.C.Cir.1970), cert. denied, 403 U.S. 923, 91 S.Ct. 2229, 29 L.Ed.2d 701 (1971)). Despite this deferential standard, we find that the Administrator has ventured into a zone of impermissible action. The Administrator has not exercised his expertise to determine an acceptable risk to health. To the contrary, in the face of uncertainty about risks to health, he has simply substituted technological feasibility for health as the primary consideration under Section 112. Because this action is contrary to clearly discernible congressional intent, we grant the petition for review.
Given the foregoing analysis of the language and legislative history of section 112, it seems to us beyond dispute that Congress was primarily concerned with health in promulgating section 112. Every action by the Administrator in setting an emission standard is to be taken “to protect the public health.” In setting an emission standard for vinyl chloride, however, the Administrator has made no finding with respect to the effect of the chosen level of emissions on health. Nor has the Administrator stated that a certain level of emissions is “safe” or that the chosen level will provide an “ample margin of safety.” Instead, the Administrator has substituted “best available technology” for a finding of the risk to health.
In the decision withdrawing the proposed 1977 amendments, the Administrator mentioned the risks to health, see 50 Fed.Reg. 1182 (1985), but based his decision solely on the finding that “there is no improved or new control technology that has been demonstrated to significantly and consistently reduce emissions to a level below that required by the current standard.” Id. at 1184. Nowhere in the decision did the Administrator state that the 1976 emission standards provide an “ample margin of
In the 1977 proposal to decrease the level of emissions, the Administrator did not determine the risk to health under the then existing standard or under the proposed new standard. Nor did the Administrator explain why one standard was “safe” and the other was not. See 42 Fed.Reg. 28,-154-157 (1977).
The absence of any finding regarding the relationship between the risk to health at a certain level of emissions and the “ample margin of safety” standard is also evident in the Administrator’s decision adopting the 1976 standards. Again, the Administrator mentioned the risks to health before and after regulation, see 41 Fed.Reg. 46,-560 (1976), but did not provide any explanation as to whether the risk was significant, or whether the chosen standard provided an “ample margin of safety.”
In the three decisions regarding emission standards for vinyl chloride, the Administrator has made one finding regarding the duty to set emission standards that will provide an “ample margin of safety.” The Administrator has determined that he is not required to determine on a case-by-case basis the risk to health at a particular level of emissions or to determine the relationship between that risk and “safety.” Instead, the Administrator has adopted a generic rule, which when met, will always result in an “ample margin of safety.” The Administrator has determined that this standard is met whenever he sets “emission standards that require emission reduction to the lowest level achievable by use of the best available control technology in cases involving apparent non-threshold pollutants where complete emission prohibition would result in widespread industry closure and EPA has determined that the cost of such closure would be grossly disproportionate to the benefits of removing the risk that would remain after imposition of the best available control technology.” 40 Fed.Reg. 59,532, 59,534 (1975).
Thus, in setting emission standards for carcinogenic pollutants, the Administrator has decided to determine first the level of emissions attainable by best available control technology. He will then determine the costs of setting the standard below that level and balance those costs against the risk to health below the level of feasibility. If the costs are greater than the reduction in risk, then he will set the standard at the level of feasibility. This exercise, in the Administrator’s view, will always produce an “ample margin of safety.”
If there was any doubt that the Administrator has substituted technological feasibility for health as the primary consideration in setting emission standards under section 112, that doubt was dispelled by counsel for the EPA at oral argument. In response to a question from the court regarding a carcinogenic pollutant known to cause certain harm at 100 ppm, counsel stated that the Administrator could set an emission level at 99 ppm if that was the lowest feasible level and the costs of reducing the level below 99 ppm would be grossly disproportionate to the reduction in risk to health. Given the strong inference that harm would also certainly result at 99 ppm, the Administrator appears to have concluded that the “ample margin of safety” standard does not require any finding that a level of emissions is “safe.” Instead, the Administrator need only find that the costs of control are greater than the reduction in risk to health. We disagree.
We find that the congressional mandate to provide “an ample margin of safety” “to protect the public health” requires the Administrator to make an initial determination of what is “safe.” This determination must be based exclusively upon the Administrator’s determination of the risk to health at a particular emission level. Because the Administrator in this case did not make any finding of the risk to health, the question of how that determination is to be made is not before us. We do wish to note, however, that the Administrator’s decision does not require a finding that “safe” means “risk-free,” see Industrial Union Dep't, 448 U.S. at 642, 100 S.Ct. at 2864, or a finding that the determination is free from uncertainty. Instead, we find only that the Administrator’s decision must be
Congress, however, recognized in section 112 that the determination of what is “safe” will always be marked by scientific uncertainty and thus exhorted the Administrator to set emission standards that will provide an “ample margin” of safety. This language permits the Administrator to take into account scientific uncertainty and to use expert discretion to determine what action should be taken in light of that uncertainty. See Environmental Defense Fund, 598 F.2d at 83 (“by requiring EPA to set standards providing an ‘ample margin of safety,’ Congress authorized and, indeed, required EPA to protect against dangers before their extent is conclusively ascertained”); Hercules, 598 F.2d at 104 (“Under the ‘ample margin of safety’ directive, EPA’s standards must protect against incompletely understood dangers to public health and the environment, in addition to well-known risks.”). In determining what is an “ample margin” the Administrator may, and perhaps must, take into account the inherent limitations of risk assessment and the limited scientific knowledge of the effects of exposure to carcinogens at various levels, and may therefore decide to set the level below that previously determined to be “safe.” This is especially true when a straight line extrapolation from known risks is used to estimate risks to health at levels of exposure for which no data is available. This method, which is based upon the results of exposure at fairly high levels of the hazardous pollutants, will show some risk at every level because of the rules of arithmetic rather than because of any knowledge. In fact the risk at a certain point on the extrapolated line may have no relationship to reality; there is no particular reason to think that the actual line of the incidence of harm is represented by a straight line. Thus, by its nature the finding of risk is uncertain and the Administrator must use his discretion to meet the statutory mandate. It is only at this point of the regulatory process that the Administrator may set the emission standard at the lowest level that is technologically feasible. In fact, this is, we believe, precisely the type of policy choice that Congress envisioned when it directed the Administrator to provide an “ample margin of safety.” Once “safety” is assured, the Administrator should be free to diminish as much of the statistically determined risk as possible by setting the standard at the lowest feasible level. Because consideration of these factors at this stage is clearly intended “to protect the public health,” it is fully consistent with the Administrator’s mandate under section 112.
For the foregoing reasons, the petition for review is granted, the decision withdrawing the 1977 proposed rule is vacated, and this case is hereby remanded for timely reconsideration of the 1977 proposed rule consistent with this opinion.
It is so ordered.
. With the exception of mercury, every pollutant the Administrator has listed or intends to list under § 112 is a non-threshold carcinogen. See 40 C.F.R. § 61.01(a) (1986) (listing asbestos, benzene, beryllium, coke oven emissions, inorganic arsenic, radionuclides, and vinyl chloride); 50 Fed.Reg. 24,317 (June 10, 1985) (chromium); 50 Fed.Reg. 32,621 (Aug. 13, 1985) (carbon tetrachloride); 50 Fed.Reg. 39,626 (Sept. 27, 1985) (chloroform); 50 Fed.Reg. 40,286 (Oct. 2, 1985) (ethylene oxide); 50 Fed.Reg. 41,466 (Oct. 10, 1985) (1,3-butadiene); 50 Fed.Reg. 41,994 (Oct. 16, 1985) (ethylene dichloride); 50 Fed. Reg. 42,000 (Oct. 16, 1985) (cadmium); 50 Fed. Reg. 52,422 (Dec. 23, 1985) (trichloroethylene); 50 Fed.Reg. 52,880 (Dec. 26, 1985) (perchioroethylene).
. The bill prohibited new sources if and because they emitted extremely hazardous pollutants. Allowing exemptions to such prohibitions without specifying the permissible bases for exemption seems to invite consideration of non-health factors, for it would be strange indeed to construct a scheme under which both a prohibition and an exemption from prohibition were available on the basis of the same criterion. To be sure, the Senate bill did set up a system under which the agency was to propose a prohibition of hazardous pollutants, and could refuse to promulgate that prohibition only for health-based reasons, see infra pp. 1156-57, but this does not make any more plausible the notion that the House bill set up a prohibition and exemption on the sole basis of health. The Senate provision did not set up a system of prohibitions and exemptions, but rather a procedural system to guide the agency. See S. 4358, 91st Cong., 2d Sess. § 6(b), 116 Cong.Rec. 32,375 (1970). The agency was to publish a list of hazardous pollutants on the basis of "available material evidence." This list would amount to a proposed prohibition of emissions of the substances on the list. The agency was then to provide notice and a public hearing for each agent or combination of agents included on the list and could refuse to promulgate the prohibition only if a preponderance of the evidence at the hearing refuted the initial basis for the substance’s inclusion. Under the House bill, the exemption applied to something still considered "extremely hazardous.” We do not see how such an exemption could rest on exclusively health-based considerations.
. Congress in the Clean Air Act Amendments of 1977 made a minor alteration in the definition of a "hazardous air pollutant," replacing "may cause, or contribute to” with "causes, or contributes to." Pub.L. No. 95-95, § 401(a), 91 Stat. 685, 791 (1977). This change does not affect the outcome of this case.
. The NRDC also argues that the structure of § 112 itself supports its contention; Congress expressed a clear intent to preclude consideration of cost and technological feasibility in setting an emission standard under § 112(b)(1) by specifically directing the EPA to consider these factors in three other subsections of § 112. These provisions, the NRDC contends, would be superfluous if the EPA could consider cost and technological feasibility in setting an emission standard under § 112(b)(1).
The NRDC's argument fails because the cited provisions continue to have significance if the Administrator is permitted to consider cost and technological feasibility under § 112(b)(1). Section 112(c)(l)(B)(ii) authorizes the EPA to grant an existing source a waiver from an emission standard for up to two years if "necessary for the installation of controls.” 42 U.S.C. § 7412(c)(l)(B)(ii) (1982). This provision could be utilized to grant a waiver to a source that is not able to comply with a standard which was based upon cost and technological feasibility because it does not have the appropriate control technology.
Section 112(c)(2) allows the President to exempt any stationary source from emission standards "if he finds that the technology to implement such standards is not available and the operation of such source is required for reasons of national security.” 42 U.S.C. § 7412(c)(2) (1982). This provision would be necessary if the Administrator considered cost and technological feasibility in setting an emission level for non-threshold pollutants and then set the level below that achievable by the best available control technology because the balance favored the elimination of the risk. This provision would also be necessary when the known threshold level for a hazardous pollutant is below the level that current technology can attain.
Finally, § 112(e) authorizes the EPA to set a "design, equipment, work practice, or operational standard” if in the Administrator's judgment "it is not feasible to prescribe or enforce an emission standard for control of a hazardous air pollutant.” 42 U.S.C. § 7412(e)(1) (1982). For the purpose of this subsection, the term "feasible," however, relates only to the ability to measure emissions. See id. § 7412(e)(2). Thus, this subsection has no relevance to the Administrator's ability to consider cost and technological feasibility in setting an emission standard.
. The statutory scheme involved in Lead Industries regulates air pollutants the "emissions of which, in [the Administrator’s] judgment, cause or contribute to air pollution which may reasonably be anticipated to endanger public health or welfare ... [and] the presence of which in the ambient air results from numerous or diverse mobile or stationary sources.” 42 U.S.C. § 7408(a)(1) (1982). The Administrator must then publish “air quality criteria” for the pollutants thus listed, id. § 7408(a)(2), and prescribe primary and secondary ambient air standards based on these criteria. See id. § 7409. Finally, states must adopt state implementation plans to meet these ambient air standards and must submit their plans to the EPA for approval. See id. § 7410.
. The petitioner also asserts that our decision here is inconsistent with the Supreme Court’s decision in American Textile Mfrs. Inst. v. Donovan, 452 U.S. 490, 101 S.Ct. 2478, 69 L.Ed.2d 185 (1981), and this court’s decision in Hercules, Inc. v. EPA, 598 F.2d 91 (D.C.Cir.1978). We do not agree. The relevance of the Court's decision in American Textile to this case is not clear. The issue in American Textile was whether the Occupational Safety and Health Administration ("OSHA’’) was required to employ cost-benefit analysis, which was defined as a determination of "whether the reduction in risk of material health impairment is significant in light of the costs of attaining that reduction," 452 U.S. at 506, 101 S.Ct. at 2489, in setting occupational health standards under a statute which directed OSHA to consider feasibility. OSHA argued that it was not required to employ cost-benefit analysis, but instead was only required "to promulgate standards that eliminate or reduce such risks to the extent such protection is technologically and economically feasible.” Id. at 507, 101 S.Ct. at 2489. The Court found that
The holding in American Textile would seem to be limited to the finding that when Congress directs an agency to consider feasibility, the agency is not required to employ cost-benefit analysis. That issue is not before us here. The Administrator has not argued that he intends to weigh the marginal gain against the marginal cost of each increment of further regulation and then to set the level of regulation at the point at which the latter exceeds the former. Instead, he intends to set the level at the lowest level that is feasible. Thus the issue in this case involves an authority that OSHA concededly had in American Textile and, therefore, the case does not affect our decision here.
The petitioner argues that this court’s decision in Hercules supports its claim that the "ample margin of safety" language prohibits consideration of cost and technological feasibility. Hercules involved § 307(a)(4) of the Federal Water Pollution Control Act, which directs the EPA to set standards for toxic water pollutants which provide "an ample margin of safety." See 33 U.S.C. § 1317(a)(4) (1982). In relevant part, the decision dealt with an industry petitioner's claim that certain regulations promulgated by the EPA under § 307(a) failed adequately to take feasibility into account. The EPA responded that § 307(a) does not require consideration of any such factor.
The court agreed with the EPA, principally on the ground that § 307(a)(2) enumerated six specific factors to take into account in setting standards for toxic water pollutants, and none involved economic or technological criteria. 598 F,2d at 111. Reinforcing this interpretation was the fact that “[s]ection 307(a)(4) directs EPA to set standards providing 'an ample margin of safety’ without any mention of feasibility criteria.” Id. This, however, does not support the NRDC’s position in this case, for Hercules merely stands for the proposition that the unadorned appearance of "ample margin of safety” does not require economic and technological considerations; the case says nothing about what such language may permit.
Nor do we find persuasive the dicta from Hercules that the NRDC cites on the subject of § 112 of the Clean Air Act. The Hercules court noted similarities between the Federal Water Pollution Control Act Amendments and the Clean Air Amendments of 1970. 598 F.2d at 112. The court then discerned a distinction applicable to both statutes positing "health-based” regulation for toxic water and hazardous air pollutants and "technology-based” regulation for other water and air pollutants. See id. The court also noted that "Congress enacted section 112 ... without provision for considerations of feasibility." Id.
First, an interpretation of this dicta as providing that an agency is authorized to consider cost and technology only under "technology-based” statutes would render the court's decision inconsistent with our mandate as expressed in Chevron. Our role at this stage is only to determine whether Congress has expressed a clear intent to preclude consideration of cost and technological feasibility under § 112. If we do not discern such an intent, we cannot simply impose our views as to whether this is a "safety-based" or "technology-based” statute and then limit the Administrator’s discretion solely on the basis of that view. See Chevron, 467 U.S. at 842-45, 104 S.Ct. at 2781-83.
Nor do we believe that the Hercules court's casual observation that § 112 makes no provision for feasibility changes the analysis. There is also no indication in § 112 that Congress intended to preclude consideration of feasibility.
. The Administrator could employ this alternative kind of regulation when an emission standard proved infeasible because "(A) a hazardous pollutant or pollutants cannot be emitted through a conveyance designed and constructed to emit or capture such pollutant, or ... any requirement for, or use of, such a conveyance would be inconsistent with any Federal, State, or local law, or (B) the application of measurement methodology to a particular class of sources is not practicable due to technological or economic limitations.” 42 U.S.C. § 7412(e)(2) (1982); Pub.L. No. 95-95, § 110, 91 Stat. 685, 703 (1977).
. The NRDC argues that because the EPA had promulgated its 1977 Notice of Proposed Rule-making to amend the vinyl chloride standard, it becomes unclear whether the EPA’s cost and technological feasibility interpretation was well established at the time of the 1977 amendments. We think it was. Although embracing a zero emission goal, the notice explicitly disagreed with criticism that the 1976 regulations placed unwarranted emphasis on technological, rather than health-based goals. 42 Fed.Reg. 28,154 (1977). Of more importance, the proposed rules mandated only "more efficient use of existing control technology at existing plants and more effective controls at new plants." Id. The proposal encouraged new technology but refused to ban vinyl chloride because of the drastic implications such a measure would hold for the industry. Id. Thus, while more stringent, the proposed regulations did not by any means abandon the EPA's earlier position.
. The EPA also argues that asbestos regulations based on feasibility considerations had been promulgated pursuant to § 112 prior to the 1977 amendments. We give this no weight because the EPA has given us no indication that Congress had any awareness of the existence, let alone the content, of those regulations.
. A court must only sparingly accept arguments based on acquiescence or ratification, for "the Framers of our Constitution deliberately made the passage of legislation difficult — more difficult, for instance, than in parliamentary democracies — [and] Congress simply cannot be obliged affirmatively to correct subsequent administrative interpretations inconsistent with original legislative intent; that is the responsibility of the courts." Coalition to Preserve the Integrity of Am. Trademarks v. United States, 790 F.2d 903, 917 (D.C.Cir.1986). Congressional inaction on a particular point may "betoken[ ] unawareness, preoccupation, or paralysis,” rather than tacit assent. See Zuber v. Allen, 396 U.S. 168, 185-86 n. 21, 90 S.Ct. 314, 324 n. 21, 24 L.Ed.2d 345 (1969). Of course, in the case before us there was more than mere congressional silence; there was affirmative evidence that members of the House knew what the EPA was doing and failed' to object and that the House Committee employed language that showed a desire to achieve a reduction, but not necessarily the total elimination, of emissions. This is something more than mere failure to enact corrective legislation.
. In response to the facts presented in this case we have analyzed this issue by using a two-step process. We do not mean to indicate that the Administrator is bound to employ this two-step process in setting every emission standard under § 112. If the Administrator finds that some statistical methodology removes sufficiently the scientific uncertainty present in this case, then the Administrator could conceivably find that a certain statistically determined level of emissions will provide an ample margin of safety. If the Administrator uses this methodology, he cannot consider cost and technological feasibility: these factors are no longer relevant because the Administrator has found another method to provide an "ample margin” of safety.