concurring in part and dissenting in part:
Our legal system assigns to the courts the task of saying what the law is, even in those cases where administrative agencies have arrived at interpretations of their own. See Marbury v. Madison, 5 U.S. (1 Cranch) 137, 177 (1803); Chevron U.S.A. v. Natural Resources Defense Council, 467 U.S. 837, 842-43, 104 S.Ct. 2778, 2781-82, 81 L.Ed.2d 694 (1984); Motor Vehicle Mfrs. Ass’n v. State Farm Mut. Ins. Co., 463 U.S. 29, 43,103 S.Ct. 2856, 2866, 77 L.Ed.2d 443 (1983). Although the Supreme Court has made clear that the federal courts must give agency interpretations substantial deference, Chevron, 467 U.S. at 844, 104 S.Ct. at 2782, agencies must follow the clear intent of Congress, and the courts must enforce it. The unsupported assertion that a congressional command is in some respect ambiguous is not sufficient, and should not be sufficient, to block all significant judicial review. In the case before the court today, the Environmental Protection Agency has withdrawn a proposed regulation under § 112 of the Clean Air Act on the basis of technological and economic factors Congress clearly removed from the scope of proper deliberation. See 42 U.S.C. § 7412(b)(1) (1982); 50 Fed.Reg. at 1184 (Jan. 9, 1985), reproduced in Joint Appendix (JA) 48. It is the responsibility *728of this court, despite EPA’s interpretation, to uphold Congress’ clear decision that § 112 emission standards should reflect public health considerations alone.
I concur in Judge Bork’s succinct rejection of the statute of limitations and exhaustion arguments made by the intervenor and the agency respectively. But I must respectfully dissent from that portion of the majority opinion that would defer to EPA’s consideration of economic feasibility and available technology in its rationale for the withdrawal.
The provision under review here is fairly simple. Section 112 of the Clean Air Act creates an administrative procedure for regulation of “hazardous air pollutants.” 42 U.S.C. § 7412. The central language of the provision directs the EPA Administrator to establish an emission standard “at the level which in his judgment provides an ample margin of safety to protect the public health from such hazardous air pollutant.” Id. § 7412(b)(1)(B). The provision contains no language authorizing the Administrator to consider technological and economic factors. On the contrary, the language on its face clearly makes the Administrator’s decision dependent only on health considerations. The majority argues that that the term “judgment” implies that the Administrator has discretion to consider feasibility. But the statute is clear: “judgment” applies only to an “ample margin of safety to protect the public health,” and does not make any reference to feasibility considerations, either technological or economic.
Despite this, EPA withdrew its 1977 proposed vinyl chloride standard on feasibility grounds. 50 Fed.Reg. at 1185, JA 49. The new standard is “based on judgments concerning the costs and benefits of the standard to society. The [new] standard is not designed to eliminate V[inyl] C[hloride] exposure risk entirely. Rather, it strikes a balance between public health protection and the cost of that protection.” Id. at 1183, JA 47. This approach is impermissible for several reasons, each of which I will address in turn.
First, though it may seem reasonable to consider feasibility here, the face of the statute unambiguously precludes it. In § 112 Congress had the public health first, foremost, and apparently exclusively in mind. See 116 Cong.Rec. 42385 (1970) (comments of Sen. Muskie). Second, the legislative history of the Act is not so ambiguous as to obscure the intent of Congress in § 112, the majority’s claim to the contrary notwithstanding. The history of § 112 supports the clear meaning of its language. Finally, EPA’s rationale for the withdrawal of the 1977 proposals is not as modest as the majority claims. According to the majority, EPA evaluated economic and technological feasibility in this case only as a last resort, when faced with chronic scientific uncertainty as to the extent of harmful effects of a particular hazardous pollutant. The majority would uphold this use of feasibility considerations. But EPA itself does not limit its consideration of technological and economic feasibility to such an extent, and argues that § 112 gives the Administrator discretion to consider feasibility in all cases. See Brief for Respondents at 12-14; 50 Fed.Reg. at 1185, JA 49. By withdrawing the 1977 proposed regulations as infeasible, EPA has contravened the mandate of the Clean Air Act. The proposed vinyl chloride standard should be remanded to the agency for reconsideration and reexamination consistent with the intent of Congress.
I. Congressional Intent and Section 112 The Supreme Court’s Chevron opinion, of course, guides our decision here. In that case the Supreme Court upheld an EPA interpretation of the Clean Air Act, 467 U.S. at 839-40, 104 S.Ct. at 2780, and found that, in the absence of clear congressional intent on a particular issue, the courts may only reverse an agency’s interpretation of a provision if it is “impermissible” or “unreasonable.” Id. at 843-44, 104 S.Ct. at 2782. But the Court stressed that, if “the intent of Congress is clear, that is the end of the matter; for the court, as well as the agency, must give effect to the *729unambiguously expressed intent of Congress.” Id. at 842-43,104 S.Ct. at 2781-82. In the case before this court, the language of the Clean Air Act and its structure clearly show that Congress intended to make § 112 a “health-based” rather than a “technology-based” provision. The majority’s discovery of ambiguity in the provision is not borne out by a review of precedent on closely related portions of the Clean Air Act, or by examination of the structure of the Act.
The Clean Air Act of 1970 was largely the result of public concern over the growing destruction of our air quality. See Schoenbrod, Goals Statutes or Rules Statutes: The Case of the Clean Air Act, 30 U.C.L.A.L.Rev. 740, 744 (1983); 116 Cong. Rec. 42381-82 (remarks of Sen. Muskie). The Air Quality Act of 1967 had failed to improve air quality to any significant extent, Schoenbrod, supra, at 745, and in 1970 Congress responded to the problem by enacting a series of substantial amendments to the Act. Id. See Clean Air Act Amendments of 1970, Pub.L. No. 91-604, 84 Stat. 1676, codified at 42 U.S.C. §§ 7401-7626 (1982). In that legislation Congress attempted to stiffen the spines of state clean air administrators, see Schoenbrod, supra, at 744-47, and to institute a policy of “technology-forcing.” The latter policy imposes air quality standards that are unattainable under present technology in order to force industry to produce the equipment necessary for effective control of airborne pollutants. Train v. Natural Resources Defense Council, Inc., 421 U.S. 60, 91, 95 S.Ct. 1470, 1487, 43 L.Ed.2d 731 (1974); see also Note, Forcing Technology: The Clean Air Act Experience, 88 Yale L.J. 1713, 1713-15 (1979). Although “technology forcing” necessarily involves imposition of standards that at present seem unreasonable, Congress clearly made the “hard choice” to follow this policy in search of cleaner air. Union Electric Co. v. EPA, 427 U.S. 246, 257, 96 S.Ct. 2518, 2525, 49 L.Ed.2d 474 (1976); 116 Cong.Rec. 42381 (comments of Sen Muskie).
The statute implements the “technology-forcing” policy by refusing to allow the agency to consider technological and economic feasibility in the establishment of most emissions and ambient air standards. Section 108(a)(2), which sets out the grounds for establishment of air quality standards, does not mention technological or economic factors. 42 U.S.C. § 7408(a)(2) (1982). Section 109(b)(1), which directs the EPA Administrator to establish ambient air quality standards “the attainment and maintenance of which in the judgment of the Administrator, based on [criteria not including feasibility] and allowing an adequate margin of safety, are requisite to protect the public health,” does not mention technological or economic feasibility. 42 U.S.C. § 7409(b)(1) (1982). This court has held that § 109(b)(1) not only does not require the Administrator to consider technological and economic feasibility, but actually bars him from doing so. Lead Industries Ass’n, Inc. v. EPA, 647 F.2d 1130, 1150 (D.C.Cir.1980), cert. denied, 449 U.S. 1042, 101 S.Ct. 621, 66 L.Ed.2d 503 (1980).1 Section 110(a)(2), which directs the EPA Administrator to review state ambient air quality standards, does not mention technological or economic feasibility either. 42 U.S.C. § 7410(a)(2) (1982). The Supreme Court itself has held that § 110(a)(2) does not allow the EPA Administrator to consider economic and technological feasibility when reviewing state ambient air quality standards. Union Electric, 427 U.S. at 257-58, 266, 96 S.Ct. at 2525-26, 2529.
Although the intent of Congress to bar consideration of feasibility in the establish*730ment of air quality standards is clear in §§ 108, 109 and 110, none of the three sections does so explicitly. Neither does the provision under review here, § 112. 42 U.S.C. § 7412. Section 112, addresses the problem of particularly hazardous pollutants, and merely directs the Administrator to “establish any such standard 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). Congress did not mention feasibility considerations.
The absence of language that embraces feasibility factors in any of these sections, however, is enormously significant when juxtaposed with other provisions of the Act that explicitly allow evaluation of technological and economic feasibility. When Congress wanted feasibility to be a factor in the Administrator’s decisions, it knew how to say so. Section 111, for example, explicitly directs the Administrator to base standards of performance for new stationary pollution sources on the best technology available, and fully envisions consideration of technological and economic feasibility in making this determination. 42 U.S.C. § 7411 (1982); Lead Industries, 647 F.2d at 1149 m 37.
In fact, § 112 itself contains two subsections that allow for consideration of technological feasibility under certain special circumstances. Section 112(e) gives the Administrator discretion to promulgate a “design, equipment, work practice, or operational standard, or combination thereof” when “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).2 More importantly, § 112(c)(2) allows the President to grant stationary sources two-year waivers from enforcement of a hazardous pollutant standard “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) (emphasis added). The existence of these two subsections makes little sense unless the section as a whole does not allow the Administrator to consider feasibility. The two provisions set out narrow exceptions to the otherwise strict command of § 112 that the Administrator may only consider health factors in setting standards for hazardous pollutants.
Despite the clarity of this scheme, the agency has gone far beyond either of the two § 112 exceptions in its treatment of the vinyl chloride standard. In its 1985 notice withdrawing the proposed 1977 regulations the agency repeatedly states that, because available technology will not suffice to implement such a strict standard, the proposed standard must be withdrawn entirely. 50 Fed.Reg. at 1184 (1st col., 3d col.), 1185 (1st col., 3d col.), 1186 (1st col.), JA 48, 49, 50. The agency is not merely granting a two-year waiver of enforcement to a source critical to national security that is unable to achieve the emission standard due to lack of technology. See § 112(c)(2). Nor has the agency found that “it is not feasible to prescribe or enforce” an emission standard for vinyl chloride. See § 112(e)(1). The lack of available technology is used to justify the waiver of a strict standard itself, across the board.
This action robs § 112(c)(2) of any meaning whatever and sidesteps the narrow provisions of § 112(e)(1). If best available technology could always be considered, as EPA asserts, Congress would have had little reason to place a special “best available technology” exemption in the statute for national security industries. See W. Rodgers, Environmental Law 277 (1977) (“The occasion for existing sources to argue economics and technology [under § 112] is in a waiver request to the Administrator.”). Similarly, Congress must have believed feasibility was not a permissible consideration when it enacted § 112(e)(1) as an amendment to the Act in 1977. If feasibility were generally a permissible consideration, congressional concerns over the feasibility of *731standards for some types of pollutants would have been unnecessary. See 42 U.S.C. § 7412(e)(1); Pub.L. No. 95-95, 91 Stat. 685, 703 (1977). In short, the structure of the Act and of § 112 clearly deny the Administrator discretion to consider technological and economic feasibility in establishing hazardous pollutant emission standards.
But the structural evidence against the agency’s interpretation does not end even with these strong indications of contrary legislative intent. Congress established national ambient air quality standards under §§ 108-110 of the Clean Air Act, and “hazardous air pollutant” emission controls under § 112. As noted above, the courts have consistently read § 108, § 109, and § 110 to preclude the Administrator from considering technological and economic feasibility in setting ambient air quality standards. It makes no sense to allow the Administrator greater latitude in the establishment of standards for pollutants Congress explicitly recognized as highly dangerous than he is allowed in the regulation of less dangerous pollutants. In fact, § 112’s language directing the Administrator to establish regulations which “in his judgment provide an ample margin of safety to protect the public health” appears on its face to be more strict than § 109’s language “adequate margin of safety.” 42 U.S.C. §§ 7412(b)(1), 7409(b)(1); Environmental Defense Fund v. EPA, 598 F.2d 62, 81 (D.C.Cir.1978) (noting that “ample” is more than “adequate” in these provisions of the Clean Air Act).
The majority makes reference to this fact, but fails to explain it. Maj. op. at 724-26. Instead, the majority points out that § 109 explicitly refers to criteria to be considered in setting ambient air quality standards, and specifically fails to mention feasibility, whereas § 112 contains no such list. Maj. op. at 725. While this is certainly true, it does not distinguish congressional intent in § 109 from the intent behind § 112. Section 112 mentions feasibility as a permissible consideration no more than does § 109. Granted, unlike § 109, § 112 does not contain a list of criteria to be used by the Administrator. But the absence of a detailed list does not mean §112 allows the Administrator to consider any factor he pleases. On the contrary, § 112 lists only one proper consideration— the public health. If anything, this focus on a single factor seems to indicate an even stronger bias against feasibility considerations than exists under § 109. Moreover, the language in the two sections is so clearly parallel that if § 109 forbids consideration of feasibility, § 112, absent strong evidence to the contrary, presumptively precludes consideration of feasibility as well. See Japan Whaling Ass’n v. American Cetacean Society, — U.S. —, 106 S.Ct. 2860, 2870, 92 L.Ed.2d 166 (1986); Sedima, S.P.R.L. v. Imrex Co., — U.S. —, 105 S.Ct. 3275, 3281, 87 L.Ed.2d 346 (1985).
The majority’s interpretation muddles what has been a remarkably coherent body of law addressing the proper place of technological and economic feasibility considerations under the Clean Air Act. The role of § 112 within the Clean Air Act as a whole, as well as the internal structure of § 112 itself, make abundantly clear that EPA’s evaluation of technological and economic feasibility in its decision to withdraw the 1977 standard was impermissible under the Act.
This is not my view alone. Virtually every commentator to examine the question agrees that Congress prohibited the Administrator from considering technological and economic feasibility in the establishment of § 112 hazardous pollutant emission standards. See D. Currie, Air Pollution: Federal Law and Analysis 3-90 (1984) (“While Congress ought to have allowed costs to be taken into account [in § 112], the EPA has taken excessive liberties [in reading cost considerations into the statute]; the difference between health standards and technology standards is too obvious to be explained away as accidental.”); F. Anderson, D. Mandelker & A. Tarlock, Environmental Protection: Law & Policy 511, 513 (1984) (“[Section] 112 appears to direct EPA to set a highly protective standard that eliminates possible serious risks *732without regard to cost. * * * [Technical and economic feasibility apparently cannot be taken into account * * *.); R. Findley & D. Farber, Environmental Law 456 (1985) (“[Under § 112] [t]here is no provision for EPA consideration of the cost or difficulty of compliance in determining the emission standards. * * * Section 112 seems to require EPA to set standards which will completely eliminate emissions * * *.”); W. Rodgers, Environmental Law 277 (1977) (“[T]he administrative judgment on an emission standard for a hazardous pollutant should be dictated solely by public health considerations without regard to the costs, or feasibility of control.”); Graham, The Failure of Agency-Forcing: The Regulation of Airborne Carcinogens Under Section 112 of the Clean Air Act, 1985 Duke L.J. 100, 134 (“[T]he EPA has attempted to convert section 112 from a health-based statute to a technology-based statute. * * * The Agency’s departure from health-based emission standards is highly questionable * * *.”); Schroeder, Foreword: A Decade of Change in Regulating the Chemical Industry, 46 Law and Contemp. Prob. 1, 30-36 (Summer 1983) (“EPA rewrote the statute,” EPA’s “approach is totally unjustified” and “textually implausible,” EPA’s interpretation is “dead wrong”). Even a General Accounting Office report “finds little support for EPA’s position,” and comments that EPA seems to be “at odds with section 112.” Comptr. Gen. of the U.S., Delays in EPA’s Regulation of Hazardous Air Pollutants 44, 51 (1983). And in Hercules, Inc. v. EPA, 598 F.2d 91 (D.C.Cir.1978), this court itself stated in dicta that “Congress enacted section 112 * * * without provision for considerations of feasibility.” Id. at 112.
Obviously, none of these comments has any binding effect on this court. We are free to ignore them if we wish, and the majority has done so. Nevertheless, the commentators’ unanimity on the meaning of § 112 is a powerful indication of the clarity of Congress’ intent on this issue. Despite the majority’s attempt to show otherwise, Congress made § 112 an unambiguously health-based standard. We need not consider the legislative history of the provision in any great detail, nor need we consider whether EPA's interpretation is “reasonable” under the second tier of the Chevron deference test. Our only duty is to vindicate the clear intention of Congress.
But even if some ambiguity actually surrounded § 112, the EPA’s construction of that section still should be invalidated. If Congress had no clear intent on a particular question, and did not explicitly delegate the interpretation to the agency — as the majority claims is the case here — this court has specifically held that a petitioner need only show that the agency’s interpretation of the statute is unreasonable or impermissible in order to prevail. State of Montana v. Clark, 749 F.2d 740, 745 (D.C.Cir. 1984), cert. denied, — U.S. —, 106 S.Ct. 246, 88 L.Ed.2d 255 (1985); American Cetacean Society v. Baldrige, 768 F.2d 426, 433 (D.C.Cir.1985), overruled on other grounds sub. nom. Japan Whaling Ass’n v. American Cetacean Society, — U.S. —, 106 S.Ct. 2860, 92 L.Ed.2d 166 (1986); Chevron, 467 U.S. at 843-44, 104 S.Ct. at 2782. Given the structure of the Clean Air Act and the existence of closely parallel precedent on other provisions within the Act, EPA’s interpretation of § 112 simply cannot stand as “permissible.”
II. The Legislative History of Section 112
The majority, confronted with the clarity of § 112’s language in light of the Clean Air Act’s overall structure, searches for ambiguity in the voluminous legislative history of the 1970 Clean Air Act Amendments. At the outset, it is important to note that the majority does not find that the legislative history clearly indicates the face of the statute is misleading and that Congress intended economic and technological feasibility to be considered by the Administrator in setting § 112 emission standards. If the legislative history clearly contradicted the language of the statute, perhaps the intent of Congress could legitimately be called “ambiguous,” and Chevron’s “reasonableness” review of EPA’s *733interpretation would become appropriate. See 467 U.S. at 844,104 S.Ct. at 2782. The majority, however, stops far short of this evaluation of the legislative history. Indeed, it finds that “(a ]£ most * * * the legislative history of the 1970 Clean Air Amendments [is] ambiguous with respect to what the Administrator may consider * * Maj. op. at 719 (emphasis added).3
The majority uses “ambiguous” legislative history, therefore, to indict a provision that is otherwise clear on its face. This is a contorted approach to statutory interpretation. The words of a statute are presumptively conclusive of legislative intent. State of Montana, 749 F.2d at 747. Ambiguous legislative history alone cannot overcome that presumption. The majority effectively suggests that judicial deference to administrative agencies is due whenever any ambiguity can be found in the legislative history of a provision, regardless of ambiguity or lack thereof in the actual language and structure of the statute. This approach comes perilously close to establishing an absolute rule of judicial deference to agency interpretations. Virtually all legislative histories of any size are plagued with some degree of contradiction and ambiguity. The majority would impose upon Congress a duty to be clear that goes far beyond what that body, or any human drafter, could possibly hope to achieve on a consistent basis in an age of complex statutory schemes. In our caution not to rob the Executive Branch of its proper role in the constitutional system, we must be extremely careful not to deprive Congress of effective legislative control over agency action.
In any event, the legislative history is not ambiguous. On the contrary, the history of § 112 supports its otherwise clear meaning. Section 112 had its origins in § 115 of Senate bill S. 4358 and § 112 of House bill H.R. 17255. The Senate bill contained stiff provisions that prohibited non-threshold hazardous pollutants, that is, hazardous pollutants for which no safe level of exposure exists. Only health factors were to be considered. See S. 4358, 91st Cong., 2d Sess. § 6(b), 116 Cong.Rec. 32375 (1970). The House bill, in contrast, explicitly called for consideration of feasibility in setting emission standards for all new sources of hazardous pollutants. The House version did not attempt to handle existing sources. See H.R. 17255, 91st Cong., 2d Sess. § 5(a), 116 Cong.Rec. 19225-26 (1970). The conference committee arrived at a compromise between the two versions, and it is that compromise that is at issue here.
First of all, the final version of § 112 clearly resembles the Senate bill. Like the Senate bill, § 112 regulates all emissions, not simply those from new sources. Like the Senate bill, the final version does not include a provision that would give the Administrator broad discretion to grant exceptions from emissions standards. The House bill contained such a provision. See H.R. 17255, 91st Cong., 2d Sess. § 5 (1970). The majority denies that the similarity between the Senate bill and the final Act says anything as to whether Congress intended to allow the Administrator to consider economic and technological feasibility in setting the standards. Maj. op. at 718. This overstates the case. If the final bill more closely resembles the Senate bill, then that fact should be taken at face value. The Senate version seems to have substantially prevailed in the conference committee deliberations. As the Senate version forbade consideration of feasibility, we can draw the cautious inference, absent evidence to the contrary, that the final bill does so as well. It would be odd to discover a “rule” of statutory construction that indicated that final bills resembling the version of one house are to be evaluat*734ed according to the legislative history of the other house’s version.
This analysis alone, however, does not illuminate every dim corner of the compromise reached between House and Senate on § 112. Conceivably, the House conferees allowed the Senate to retain its regulation of all sources, both new and existing, in exchange for an implicit agreement to allow the Administrator to consider feasibility in setting emission standards. The insertion of the language “ample margin of safety” into the final provision weakly supports this hypothesis, though it is important to reemphasize that this language on its face speaks only to health considerations, and conspicuously fails to make any mention of feasibility. See Conference Report, H.R.Rep. No. 1783, 91st Cong., 2d Sess. 195-96 (1970).
But the conference committee added other language to § 112 as well. The committee, after wrangling over the scope of the regulation and the precise language of the command to the Administrator, made a point of inserting § 112(c), the presidential waiver provision. See Conference Report, H.R.Rep. No. 1783, 91st Cong., 2d Sess. 197 (1970). This, it seems to me, is the heart of the compromise between House and Senate. The waiver provision, as previously noted, see page 730 supra, allows for temporary waiver of emission standards if available technology cannot implement the standards at a source of importance to national security. This provision is unnecessary unless the “ample margin of safety” language otherwise bars consideration of available technology. The conference committee added both pieces of language at approximately the same time, a fact that makes it difficult to believe the committee was unaware of the interaction of the two phrases.
Given the resemblance of the final bill to the Senate bill, and the fairly clear evidence of the nature of the compromise between the two houses of Congress provided by the changes made in conference, the legislative history tips toward petitioner’s position. As noted above, however, it is far from clear that even an ambiguous legislative history would require us to find for respondents. On the contrary, petitioner’s position is amply supported by the face of the statute, which makes resort to the legislative history a subsidiary endeavor. In any event, however, the ambiguity the majority seeks in § 112 can no more be found in its legislative history than in its actual language.
III. The Majority’s “Uncertainty” Rationale
To explain its inference of broad discretion for the Administrator to consider economic and technological feasibility under § 112, the majority relies heavily on what it regards as the inevitable discretion of the EPA Administrator when faced with “scientific uncertainty.” In essence, the majority would create an “uncertainty” exception to the “health-based” character of § 112. The questions raised by the majority’s discovery of this exception to the general rule that feasibility is not to be considered in areas where Congress has not authorized such an approach, American Textile Manufacturers Institute v. Donovan, 452 U.S. 490, 510, 101 S.Ct. 2478, 2491, 69 L.Ed.2d 185 (1981) (when Congress intends cost-benefit analysis to be used, it says so); Union Electric, 427 U.S. at 257 n. 5, 96 S.Ct. at 2525 n. 5 (same), are sufficiently important to warrant separate discussion.
As a threshold matter, the majority’s characterization of EPA’s rationale for withdrawing the proposed vinyl chloride standard is misleading. Throughout its opinion the majority attempts to narrow EPA’s position to more tenable grounds by limiting EPA’s claims to the special “uncertainty” question posed by carcinogen regulation. EPA itself has not been so neal. In its notice of proposed regulation, in its briefs, and in its oral argument to this court, EPA never raises the “scientific uncertainty” problem in any significant manner. And with good reason, because it is not EPA's position that only scientific uncertainty justifies consideration of fea*735sibility under § 112. EPA argues that it may engage in cost-benefit analysis, including consideration of technological and economic feasibility, any time it promulgates regulations under § 112. See 50 Fed.Reg. at 1183, JA 47 (cost-benefit analysis used to arrive at vinyl chloride standard, because elimination of risk would require prohibition of emissions); Brief for Respondents at 12-14.
EPA’s expansive position on its discretion to consider feasibility flows in part, no doubt, from the fact that the uncertainty as to vinyl chloride’s low-concentration health effects seems relatively small. See 50 Fed. Reg. at 1182, JA 46. In fact, EPA has decided to a substantial degree of certainty that the number of lives saved by regulation below the level of the 1976 standards would not be justified by the costs of such regulation. Id. (setting out specific, though estimated, figures on risk). As a consequence, the majority’s attempt to distinguish between “apparent non-threshold toxics” and “non-threshold toxics” is rather odd. Maj. op. at 722 n. 7. EPA does not focus on any such distinction. On the contrary, EPA explicitly recognizes that “any atmospheric concentration of VC poses some public health risk.” 50 Fed.Reg. at 1183, JA 47; compare maj. op. at 722 (EPA not refusing to regulate known health risk). In other words, EPA recognizes that vinyl chloride is effectively a non-threshold pollutant.4 EPA’s decision to withdraw its proposed regulations did not spring from concern over uncertainty, but rather rested on the determination that the costs of regulation exceeded its benefits. When Congress enacted § 112 of the Clean Air Act as a health-based provision, it forbade the use of precisely this rationale.
The majority argues that the word “judgment” in § 112(b)(1) implicitly grants the Administrator discretion to consider feasibility in setting emissions standards when the level of harms from pollutant exposure is uncertain. Maj. op. at 716-17. If health were the only factor to be considered, the majority submits, the statute would effectively require that the Administrator ban outright any emissions of pollutants with no safe level of emissions, that is, any “non-threshold” pollutants. If this were the case the Administrator could make only one decision, his discretion would disappear, and the word “judgment” would have no meaning. This would also be true, says the majority, if any uncertainty whatever existed as to the health risks posed by a pollutant. Three responses to this line of reasoning come to mind. I will explain each in turn. First, “non-threshold” pollutants are not the only pollutants § 112 regulates. A substantial part of the Administrator’s “judgment” lies in classifying pollutants as threshold or non-threshold, and in establishing specific threshold levels. Second, the Administrator has *736some discretion to refuse to regulate “insignificant” health risks. Finally, this court has no authority to allow the agency to ignore clear congressional commands, regardless of our opinion of their wisdom.
Congress did not enact § 112(b)(1) to regulate “non-threshold” pollutants alone. The Senate report explicitly adverted to “threshold” pollutants, that is, pollutants safe at some low level of emissions, in its explanation of the scope of the Administrator’s discretion. If the Administrator found “that a greater than zero emission [of a hazardous pollutant] could be permitted without presenting a hazard to health,” he would be free not to prohibit emissions of that pollutant. S.Rep. No. 1196, 91st Cong., 2d Sess. 20 (1970). Senator Muskie, reflecting back on the legislative process, has said that the Act is based on the assumption that thresholds of safety exist for some hazardous pollutants. See Clean Air Act Amendments of 1977, Hearing Before the Subcomm. on Environmental Pollution of the Senate Comm, on Environment and Public Works (pt. 3), 95th Cong., 1st Sess. 8 (1977).
Despite the majority’s assertions to the contrary, therefore, the word “judgment” simply does not require a technological and economic feasibility overlay to make sense in this provision. The term “judgment” in § 112(b)(1) gives the Administrator discretion to determine the existence and level of “thresholds” for various pollutants, consistent with the public health. See Lead Industries, 647 F.2d at 1152 (EPA position stresses that Administrator’s “judgment” regards scientific and technical questions). It does not grant discretion to make wide-open evaluations of technological and economic feasibility. The existence of uncertainty in the determination of threshold levels not only fails to undercut this interpretation of congressional intent, but actually supports it. Congress was aware that safe emissions levels might always be scientifically uncertain to some extent, and therefore gave the Administrator discretion to make the final determination of the threshold levels of various pollutants by evaluating existing scientific evidence. Id. at 1152-53.
The majority worries that interpretations of § 112 that do not allow for considerations of feasibility might require the EPA to prohibit all emissions of non-threshold pollutants. This could lead, says the majority, to the elimination of entire industries for the sake of negligible and uncertain improvements in the public health. The majority’s picture of the calamities that could befall the nation if this court obeys the will of Congress is greatly exaggerated. Though the statute might be read to ban emissions of non-threshold pollutants in every case, I am unconvinced that this would be proper. Congress clearly banned consideration of feasibility under § 112, but it did not clearly require an absolute ban on non-threshold pollutant emissions. At times EPA may in fact be bound for health reasons to ban outright all emissions of a given non-threshold pollutant. But it is well established that agencies have some limited discretion to refuse to regulate “insignificant” harms. In Alabama Power Co. v. Costle, 636 F.2d 323 (D.C.Cir.1979), for example, Judge Leventhal noted that “[c]ourts should be reluctant to apply the literal terms of a statute to mandate pointless expenditures of effort,” id. at 360, and that “there is likely a basis for an implication of de minimis authority to provide exemption when the burdens of regulation yield a gain of trivial or no value.” Id. at 360-61; see also Ethyl Corp. v. EPA, 541 F.2d 1, 13-14 (D.C.Cir. 1975) (era banc), cert. denied, 426 U.S. 941, 96 S.Ct. 2663, 49 L.Ed.2d 394 (1976). If, in the judgment of the Administrator, health harms from some low level of emissions are trivial or nonexistent, § 112 does not require a ban.
But that is not the situation presented by this case. Discretion to allow “de minim-is ” emissions of toxic pollutants if harm to the public health would be truly insignificant is a far cry from the EPA’s claimed discretion to examine technological and economic feasibility under all circumstances, or from the majority’s attempt to give EPA license to consider feasibility factors in the *737face of any scientific uncertainty whatever. The majority highlights just how sweeping an approach it takes by stating that even if “the area of uncertainty [of harms associated with a toxic], as with vinyl chloride, covers all non-zero levels of emission[, that] does not alter our conclusion [that cost-benefit analysis can be applied when any uncertainty exists].” Maj. op. at 722. Trivial and insignificant harms are not the only harms encompassed by this language.
Finally, even if the statute did require an outright ban on non-threshold pollutant emissions, this court should not set about “correcting” perceived deficiencies in that policy. If Congress has made the decision to impose such costs on society, that is the prerogative of Congress. The Supreme Court made this absolutely clear not long ago, in a situation in some respects not very different from that before us today. See TVA v. Hill, 437 U.S. 153, 194, 98 S.Ct. 2279, 2301, 57 L.Ed.2d 117 (1978). Though we owe the agency here substantial deference, see Chevron, 467 U.S. at 844, 104 S.Ct. at 2782, that deference pales by comparison with that we owe Congress. Congress has spoken clearly in § 112, and we must obey its command.
The extent to which the majority’s grant of wide discretion to the EPA to consider feasibility in the face of “uncertainty” is at loggerheads with congressional intent becomes even more clear if we consider the practical effect of this exception. As has become all too apparent in the last fifteen years, the degree of risk posed by carcinogens is virtually always a matter of some uncertainty. See, e.g., Industrial Union Dep’t, AFL-CIO v. American Petroleum Institute, 448 U.S. 607, 613,100 S.Ct. 2844, 2849, 65 L.Ed.2d 1010 (1980). The majority itself recognizes this fact, admitting that science may never eliminate uncertainty in the evaluation of carcinogenic risk. Maj. op. at 716. The existence of a hazard may be absolutely clear, as in the present case, but the magnitude of that hazard is likely to be imprecise.
Many of the pollutants regulated under § 112, or being considered for regulation under that section, are carcinogens. Comptr. Gen. of the U.S., Delays in EPA’s Regulation of Hazardous Air Pollutants 8-11 (1983); Summary of Data on Specific Pollutants and Cancer Risk Estimates Excerpted from EPA Draft Study, 15 Env’t Rep. (BNA) 616, 616-18 (Aug. 10, 1981). As a consequence, if the majority’s interpretation of § 112 prevails, many pollutants regulated under this provision will be regulated under a “technology-based” feasibility standard. This would be perverse. Granted, Congress may not have realized the full extent to which carcinogenic risk is subject to intrinsic uncertainty. Nevertheless, Congress was clearly aware that some toxics pose a danger to health at any concentration, requiring the strictest regulation without regard to cost, even if the extent of that danger might be unclear at the third decimal point. See 116 Cong.Rec. 16091 (1970) (comments of Sen. Muskie). Congress did not include any language whatever that indicates it embraced a broad “uncertainty” exception of the sort proposed by the majority. After all, the judgment of the Administrator under § 112 is judgment as to an “ample margin of safety necessary to protect the public health,” not judgment as to “a reasonable degree of safety in light of the availability of technology and the costs to be imposed on industry.”
Moreover, the majority’s willingness to allow for consideration of feasibility if harms are uncertain runs afoul of its own logic. The majority feels that if the exact scope of harm from a hazardous pollutant is not certain, even though the existence of harm may be clear, then “such considerations as technological and economic feasibility seem natural.” Maj. op. at 722. EPA should be allowed to weigh these factors against the harms caused by the toxic, argues the majority. If the harms that stem from emission of a hazardous pollutant are uncertain, however, how can they be “weighed” against technological and economic feasibility? The majority proposes an exception to Congress’ ban on cost-benefit analysis under § 112 that covers precisely the circumstance where cost-*738benefit analysis is of no real assistance to the Administrator. Let us be clear about what the majority would do: Discretion to weigh cost and technology against uncertain levels of harm — even though these harms may be certain at some level — is broad discretion not to regulate low-level hazardous pollutants at all. Without solid statistics for harm, the costs of regulation will always dominate the equation. Perhaps this approach can be justified if Congress has explicitly used the language of “reasonableness,” see, e.g., Industrial Union Department, 448 U.S. at 642, 100 S.Ct. at 2864, but it is hardly acceptable if Congress has gone to considerable lengths to avoid such language, as in the case before us.
III. Conclusion
This court should not read the Supreme Court’s Chevron opinion and general principles of administrative law to allow the insertion of cost-benefit analysis or its equivalent into every statute where Congress makes any technical delegation to an agency. It is the legislature’s right, and at times its duty, to vindicate public values even if they are not cost-justified. See, e.g., TVA v. Hill, 437 U.S. at 194-95, 98 S.Ct. at 2301-02. Kelman, An Ethical Critique of Cost-Benefit Analysis, 1981 Regulation 33 (Jan./Feb.). In § 112 of the Clean Air Act Congress sought to make “the hard choices” necessary to improve our air quality and to eliminate hazardous pollutants from our air. Union Electric, 421 U.S. at 257, 96 S.Ct. at 2525. Congress refused to compromise on matters of public health, and denied the EPA Administrator the discretion to consider technological and economic feasibility in setting hazardous pollutant emissions standards. Simply because the majority feels lack of compromise is unreasonable does not mean the court should enforce its own conceptions of proper policy. TVA v. Hill, 437 U.S. at 185, 98 S.Ct. at 2297 (“It is not for us to speculate, much less act, on whether Congress would have altered its stance had the specific events of this case been anticipated.”).
This analysis is in no way at odds with the Chevron decision. The Supreme Court has directed the lower federal courts to consider the reasonableness of an agency’s interpretation of a statute carefully and deferentially if the intent of Congress on the issue is not clear. 467 U.S. at 844, 104 S.Ct. at 2782. As I have argued above, § 112 is as devoid of ambiguity on the permissibility of cost-benefit analysis, given the overall structure of the Clean Air Act, as any statutory provision probably could be. But even accepting, for the sake of argument, the majority’s claim that § 112 is somehow unclear, the agency here should still be directed to reconsider its decision. The “reasonableness” of an agency interpretation must be determined in light of the statutory scheme under examination. Id. at 845, 104 S.Ct. at 2783. Chevron’s, “reasonableness” review does not give the courts — and agencies — power to apply “laws” of efficiency and cost-benefit optimality to legislative schemes if the structure and history of those schemes indicate that Congress intended otherwise.
Congress did not enact the Clean Air Act in order to reach a “reasonable” accommodation between air free of hazardous pollutants and economic considerations. Congress moved with grim determination to clear the skies of these toxics, and imposed upon the country a policy of stringent “technology-forcing” regulation as “a drastic remedy to what was perceived as a serious and otherwise uncheckable problem of air pollution.” Union Electric, 421 U.S. at 256, 96 S.Ct. at 2525. Train v. NRDC, 421 U.S. at 91, 95 S.Ct. at 1487. Today, in diluting the effectiveness of that remedy, this court ignores both the letter of the Act and the uncompromising spirit behind it.
In refusing to read the clear intent of Congress in § 112 of the Clean Air Act, the majority overemphasizes the Chevron deference principle and effectively converts the unambiguously “health-based” § 112 into a “technology-based” provision. This metamorphosis not only contravenes congressional intent but flies in the face of an *739entire line of this court’s precedent, as well as precedent of the Supreme Court. When Congress intended EPA to consider technological and economic feasibility under the Clean Air Act, it clearly stated that fact. I cannot agree with the majority’s refusal to overturn EPA’s unjustified and unreasonable contrary position, and therefore
I respectfully dissent.
. The majority argues that the Lead Industries court's language barring consideration of technological or economic feasibility was unnecessary to the judgment in that case, and is therefore nonbinding dicta. Maj. op. at 725. I must disagree. Although it is formally distinguishable from the holding, the language barring consideration of feasibility under § 109 is an integral part of the opinion's rationale, as demonstrated by its repeated appearance. Lead Industries, 647 F.2d at 1148, 1149, 1150. Furthermore, EPA is clearly bound by this language in the § 109 context. For practical purposes, and certainly for purposes of interpreting parallel provisions of the Clean Air Act, the language barring consideration of feasibility is not dicta.
. Section 112(e)(2) defines "not feasible to prescribe or enforce an emissions standard” to include situations in which (A) control techniques are not available or (B) measurement methodology is impracticable "due to technological or economic limitations." 42 U.S.C. § 7412(e)(2).
. Similarly, the majority discusses whether Congress implicitly ratified EPA’s consideration of technological and economic feasibility in its 1977 amendments to the Clean Air Act, and concludes that the history of the 1977 amendments "is far short of legislative ratification of the EPA’s construction.” Maj. op. 722. Although this evaluation seems reasonable, it does not obscure the clear congressional intent that underlies § 112, and certainly does not make the language and structure of the Act ambiguous.
. The majority, not surprisingly, denies this, and insists that EPA has made a point of distinguishing between "apparent non-threshold” pollutants and "non-threshold” pollutants. Maj. op. at 722. I agree with the majority that such a distinction can be made as an intellectual matter. I do not see, however, that EPA has rested its decision in this case on any such distinction. On the contrary, EPA quite clearly states that it believes vinyl chloride, for all practical purposes, is always hazardous, even though it concedes that at some very low level of exposure adverse health effects might not accrue. Moreover, it seems to me that the majority’s distinction is one without a difference. Science may never be able to prove definitively that any given pollutant is harmless at some low concentration. See text at 736 infra; maj. op. at 716. To do so would be to prove a negative. See, e.g., Ethyl Corp. v. EPA, 541 F.2d 1, 25 n. 52 (D.C.Cir. 1976) (en banc), cert. denied, 426 U.S. 941, 96 S.Ct. 2663, 49 L.Ed.2d 394 (1976) (ultimately all scientific "fact” is uncertain at some level). Thus, if we adhere to the majority’s definitions of "apparent non-threshold” pollutant and "non-threshold” pollutant, we soon discover that the second category is an empty set. As a consequence, the distinction between the two does not offer any particular insight into EPA’s approach to vinyl chloride regulation or, more importantly, into Congress’ intent on non-threshold pollutant regulation. On the contrary, it tends to obscure the otherwise obvious fact that EPA has decided, with only a very small degree of uncertainty, that the costs of this regulation exceed its benefits. As EPA is confident enough to make decisions on the assumption that vinyl chloride is for all practical purposes a non-threshold pollutant, we should not hestitate to do so as well.