Full opinion text
Opinion for the Court filed by Senior Circuit Judge BAZELON. Concurring opinion filed by Circuit Judge WILKEY. Dissenting opinion filed by Circuit Judge WALD. BAZELON, Senior Circuit Judge: Petitioner General Motors Corporation (GM) brings three consolidated petitions for review of final actions of the Environmental Protection Agency (EPA) under the Clean Air Act, as amended. In these petitions we are asked to decide whether the recall provision of section 207(c)(1) of the Act permits the EPA to require automobile manufacturers to recall and repair at their own expense all members of a class of vehicles — a substantial number of which have been found to be in nonconformity with applicable emissions standards during their useful lives — regardless of the age or mileage of any individual vehicle when presented for repair. We take jurisdiction under section 307(b)(1) of the Act. For reasons detailed below, we reverse the actions of the Administrator. Background Through the Clean Air Act, Congress sought “to protect and enhance the quality of the Nation’s air resources so as to promote the public health and welfare and the productive capacity of its population ....” To this end, Title II of the Act establishes a comprehensive program for the control of motor vehicle emissions. The Act authorizes the Administrator to prescribe standards for motor vehicle emissions within the broad guidelines set out by the statute. Such standards are applicable to the vehicles throughout their “useful lives.” The useful life of light duty vehicles, such as automobiles, is defined by statute as “a period of use of five years or fifty thousand miles (or the equivalent), whichever first occurs ....” The Act also provides an elaborate enforcement system designed to ensure that vehicles comply with emissions standards, not only when they leave the assembly line but also while in actual use. Among the enforcement mechanisms at the disposal of the EPA is the authority to order manufacturers to recall and repair at their own expense an entire class or model of vehicles or engines should the Administrator determine that a substantial number of vehicles in that class, although properly used and maintained, have failed to meet applicable emissions standards during their useful lives. The precise scope and application of this recall authority are at issue in this lawsuit. In May 1975, the EPA began an investigation of the emissions performance of 1975 Cadillacs of the 60V43 engine family. On March 21, 1977, following a program of testing fifteen sample vehicles by both EPA and GM, the Administrator officially notified GM that he had determined that a substantial number of 1975 Cadillacs equipped with the 230-carburetor, although properly maintained and used, failed to meet federal emissions standards during their useful lives. The Administrator ordered GM to submit a plan for the recall and repair of the 230-carburetor class. He also indicated that the EPA staff would continue its investigation of emissions problems in 193-carburetor Cadillacs and urged GM to recall these vehicles voluntarily. GM did not contest the finding of nonconformity and “volunteered” to recall the 193-carburetor vehicles. Nevertheless, negotiations concerning GM’s proposed remedial plan dragged on from May 1977 until December 1979. Finally, on December 26, 1979, EPA withdrew one of its principal objections and agreed to approve a modified plan that had been submitted by GM on February 15, 1978. At this point, however, GM informed the EPA that “[ajbout forty percent of the subject vehicles are beyond five years old and many more will have accumulated more than 50,000 miles.” GM urged the Administrator to cancel the recall and averred that, if the EPA were to go forward with the recall, “only those vehicles within the lesser of five years or 50,000 miles of operation at the time of presentation to the dealer for repairs will receive the campaign adjustments at General Motors (sic) expense.” EPA responded on May 30, 1980 by promulgating what it labelled an “interpretive rule” setting out the EPA’s position that the Clean Air Act requires “manufacturers to submit a plan to remedy all vehicles within the class or category of vehicles subject to an ordered recall which experienced the nonconformity during their useful lives regardless of their age or mileage at the time of repair.” Shortly thereafter, the EPA wrote to GM, refusing to withdraw the recall order. The EPA letter formally approved GM’s remedial plan of February 15,1978 insofar as it applied to vehicles still within their useful lives at the time of repair. But, relying on the May 30 rule, the agency disapproved the plan insofar as it related to vehicles beyond their useful lives. The letter also reflected EPA’s finding that, at least with respect to vehicles beyond their useful lives, GM had “failed to submit a [remedial] plan as required by section 207(c)(1) of the Act....” GM petitioned this court, seeking review of both the May 30 rule and EPA’s partial disapproval of GM’s remedial plan. Analysis A. Standard of Review We begin by noting that we are called upon in this case to review an interpretative, not a legislative, rule. While this observation may seem apparent, the nature of the May 30 rule has been hotly contested by the parties. Although EPA has from the start characterized the rule as “interpretive,” GM argues that, because the rule substantially expands the scope of a manufacturer’s liability under the recall provisions of the Act, the rule should be viewed instead as an improperly promulgated legislative rule. Where, as here, an agency has the authority to issue both legislative and interpretative rules, the line between the two is often blurred. Nevertheless, the determining factor is the agency’s authority and intent in promulgating the rule. Although an agency’s own labelling of a rule is not dispositive of the question of intent, it is indicative and is entitled to judicial deference. There is nothing in this record to suggest that EPA ever intended this rule to carry any weight beyond that ordinarily attendirig an agency’s interpretation of a statute. The Supreme Court has pointed out that “[ojrdinarily, administrative interpretations are given important but not controlling significance.” The precise weight to be accorded an interpretative rule “in a particular case will depend upon the thoroughness evident in its consideration, the validity of its reasoning, its consistency with earlier and later pronouncements, and all those factors which give it power to persuade, if lacking power to control.” The EPA rule does not receive “high marks when judged by [these] standards ....” The rule was not a contemporaneous interpretation of the Clean Air Act, and there is no evidence that it reflects a longstanding interpretation of the Act by the agency. Nor does the rule involve the kind of fact-intensive questions concerning which great deference need be given the agency’s technical expertise; rather, as the agency itself concedes, “[s]ince the rule simply expresses an interpretation of the law based on the language, legislative history and policy of the Clean Air Act, no factual data need be analyzed or commented on.” Consequently, although some deference is to be accorded to the May 30 rule, our inquiry will focus on whether EPA’s interpretation is reasonable and supportable in light of the statutory language and legislative history. B. Statutory Language In reviewing an agency’s interpretation of a statute, a court should first examine the language of that statute to determine whether the interpretation falls within the statute’s plain meaning. In the instant case, however, the statutory language is ambiguous at best. Section 207(c)(1) of the Act provides in relevant part: If the Administrator determines that a substantial number of any class or category of vehicles or engines, although properly maintained and used, do not conform to ... [emissions standards], when in actual use throughout their useful life ..., he shall immediately notify the manufacturer thereof of such nonconformity, and he shall require the manufacturer to submit a plan for remedying the nonconformity of the vehicles or engines with respect to which such notification is given. The plan shall provide that the nonconformity of any such vehicles or engines which are properly used and maintained will be remedied at the expense of the manufacturer. EPA construes the requirements of the entire section as “class-based.” Under the agency’s interpretation, once the Administrator determines that a substantial number of vehicles in the class fail to conform to standards during their useful lives, the manufacturer is notified of the nonconformity of the class. The manufacturer must then submit a remedial plan for the entire class of vehicles, providing for the repair at the manufacturer’s expense of all vehicles or engines that are members of the class. The only exception is for vehicles that have not been properly used and maintained. Indeed, the very existence of this exception for improperly used and maintained vehicles is said to highlight the absence of an explicit exclusion of vehicles beyond their useful lives. GM focuses instead on the concept of nonconformity. Although the manufacturer is notified of the nonconformity of the class of vehicles on the basis of the failure of a substantial number of its members to conform to standards during their useful lives, the remedial plan must address the nonconformity of the individual vehicles in the class, and it is the nonconformity of such vehicles that is to be remedied at manufacturer expense. However, “[nonconformity” is a word having meaning only for vehicles for which the Act and EPA regulations set emissions standards. The Act does not set nor does it authorize EPA 'to prescribe emissions standards for any vehicle which has exceeded its 5 years/50,000 miles “useful life”. A vehicle which has passed that limit cannot be in “nonconformity” with any emissions standard and, therefore, the law cannot require that any “remedy” be provided for a nonexistent nonconformity. GM’s viewpoint is not without merit. Classes of vehicles can be recalled, and remedial plans can be prepared, on a class-wide basis. Yet the actual remedy of the nonconformity of a class of vehicles can only be accomplished by the repair of individual vehicles. The repair of a vehicle presupposes the existence of a defect or nonconformity in that vehicle, but a vehicle cannot fail to conform to a standard which does not apply to it. Thus, vehicles beyond their useful lives have no nonconformity to be remedied. Indeed, their “repair” does not even contribute to the remedy of the nonconformity of the class, since vehicles beyond their useful lives are no longer part of the nonconforming class. EPA attempts to circumvent this reasoning by conceding that the manufacturer is not required to repair vehicles which first exceed standards after five years or 50,000 miles. Moreover, “for those vehicles [beyond their useful lives] which are subject to repair at manufacturer expense, the manufacturer is only responsible to carry out the remedy contained in the approved plan and is not responsible to remedy any other non-conformities.” However, these concessions appear to generate more problems than they resolve. First, EPA’s interpretation relies upon the assumption that, because a “substantial "number” of cars are found to exceed vehicle emissions standards during their useful lives, the entire class of cars of that engine-family and model-year is suspect and ought to be subject to recall. The function of this class-based analysis is to shift the focus of attention away from the time-consuming and expensive task of establishing the nonconformity of each individual vehicle so that those resources may be devoted to remedying the nonconformity of the class as a whole. Yet EPA’s exception with respect to vehicles first exceeding standards after the expiration of their useful lives undermines its own class-based inquiry. Instead, the EPA invites manufacturers to indulge in fact-specific controversies concerning whether a given car, beyond its useful life when presented for repair, experienced nonconformity during its useful life or first exhibited excessive emissions after its useful life had expired. EPA’s concession not only erodes its own class-based theory, but also offers little relief to manufacturers in practice. If the car is not presented for repair until it has exceeded its useful life, how will the timing of the onset of excessive emissions be established? On whom will the burden of proof rest? This difficulty of proof is compounded by the manner in which the recall system operates. The presumption of class nonconformity is based on testing of a relatively small sample of cars. On the basis of such sampling, an estimate of the percentage of cars in nonconformity is made; this percentage will often be substantially less than one hundred percent. In this case, the EPA originally estimated that forty-three percent of the class would exceed standards; this estimate was later increased to sixty-eight percent. Thus, on the basis of the statistical evidence in this case, it would appear that between thirty-two and fifty-seven percent of the cars in this class — whether or not they have exceeded their useful lives at the time that notice of class nonconformity was given or at the time of repair — are likely to have been in compliance with the applicable emissions standards throughout their useful lives. In the case of cars still in their useful lives, the manufacturer may establish, through testing once the cars have been recalled, that a particular car is in compliance and, therefore, no repair would be necessary for that car. Under the EPA’s interpretation, however, such a showing may be impossible with respect to cars that have exceeded their useful lives when presented for repair. Congress clearly has contemplated some increase in a car’s emissions once the vehicle has surpassed its useful life. However, when a car brought in for repair has exceeded its useful life, the manufacturer will be unable to demonstrate whether a present emissions excess is attributable to this sanctioned erosion in performance or to a nonconformity existing during the vehicle’s useful life. In essence, the May 30 rule establishes an absolute and irrebuttable presumption that all older cars were among the percentage failing to meet standards during their useful lives. The establishment of such a presumption might be within the agency’s authority, but it goes well beyond simple statutory interpretation. The agency’s answer to the problem posed by requiring cars to be brought into conformity with a standard which no longer applies to them is equally disingenuous. The agency explains that manufacturers would not be required to bring cars beyond their useful lives when presented for repair up to the same standards applicable to “younger” cars. Instead, the manufacturer would only be required to perform the same repair on older cars as is mandated for younger vehicles by the remedial plan. This response is not entirely forthcoming. In this case, the repair required under the remedial plan involves adjustment of the idle screw and the adjustable part throttle plug to a specified RPM drop. Thus, under the precise facts of this case, there is a difference between requiring the “same repair” to be made on all cars and requiring all vehicles, regardless of age or mileage, to be brought into conformity with emissions standards. However, in other recall situations the required repair is to bring the car to a particular emissions ratio. In a recent Chrysler recall, for example, the remedy required a mechanic to attach an exhaust emissions analyzer to the catalyst tap and then to adjust the mixture screws back and forth until the idle carbon monoxide concentration met specifications. In cases in which the required repair is not a “fixed” adjustment but rather a “fine-tuning” process, the effectiveness of which is measured by conformity to emissions standards, a blanket rule disregarding the age of the car when presented for repair in effect requires the manufacturer to make an older car meet standards that Congress intentionally required only younger cars to meet. By imposing such a requirement, the EPA rule again exceeds the limits of permissible statutory interpretation. Even if every recall involved remedies in which the “repair” could be performed on older vehicles without their having to be brought into compliance with standards no longer applicable to them, the EPA rule would nevertheless be an unreasonable interpretation of the statutory mandate. If the “repair” does not bring the vehicle on which it is performed back into compliance with emissions standards, it is not a “repair” at all. In essence, the EPA rule requires a manipulation to be performed on older vehicles at manufacturer expense. But, by the EPA’s own admission, that manipulation would not remedy a nonconformity as the statutory language clearly contemplates. C. Legislative History Whatever doubt may remain concerning the legitimacy of the EPA’s interpretation is laid to rest by a critical passage of the Act’s legislative history. The passage refers to a Senate version of the bill that at that time established a 50,000-mile, but no age, limitation on vehicle useful life. In discussing the recall provisions of the bill, the Senate Committee said: The 50,000-mile period can be assumed to be 4 to 5 years and the manufacturer should be expected to notify any owner of a vehicle that is five years old or less as to failure to continue to perform to the standard. A decision not to require the manufacturer to repair the vehicle could be made after notice and after finding that the vehicle had exceeded the 50,000-mile warranty period. This language indicates the understanding of Congress that manufacturers would not be required even to notify owners of cars that had in all probability exceeded their useful lives; i.e., cars ostensibly belonging to the nonconforming class but more than five years of age at the time the manufacturer was informed by the EPA of the nonconformity would not be recalled in the first instance. Furthermore, if a car less than five years of age were recalled but found to have exceeded the 50,000-mile limitation when presented for repair, the manufacturer could be excused from liability for repair. EPA attempts to downplay this evidence of congressional intent with two responses, First, EPA contends that the legislative history of the Senate bill is irrelevant because the bill had a “fast track” recall system absent in the Act as finally adopted. Under this scheme, there would have been no allowance for protracted negotiation of the specifics of the recall plan and, hence, no opportunity for the sorts of administrative or manufacturer-induced delays that would permit large numbers of vehicles to exceed their useful lives after their nonconformity had come to the attention of the EPA. Instead, all cars would have had to have been recalled within sixty days of the date that notice of nonconformity was given by EPA to the manufacturer. As a result, EPA argues, “it is clear that the Senate may not have focused upon the problems caused by delays in the implementation of a recall, because substantial delays were impossible under the Senate scheme.” EPA’s argument fails to persuade for several reasons. First, the so-called “fast track” recall system precludes only one kind of delay — that caused by negotiations over the content of the recall plans. But EPA itself concedes that most of the delay encountered in previous recalls has not derived from protracted negotiations. Months may be consumed in testing and investigating a potentially nonconforming class prior to the issuance of a recall order. Considerable delay may occur while administrative and judicial appeals are exhausted. Yet although the “fast track” recall system would have limited the length of negotiations over remedial plans, the system would have done nothing to avoid the more substantial delays incurred in the course of investigation or appellate proceedings. Moreover, even if the “fast track” system were able to eliminate all delays encountered under the Act, this fact does not render the legislative history of the bill entirely inapposite. Perhaps the significance of the Committee Report’s language concerning the fate of vehicles found to be beyond their useful lives when presented for repair might be lessened. Yet, the absence of a possibility for manufacturer-induced delay does nothing to mitigate the weight of the report as evidence that Congress did not intend manufacturers to be responsible for the recall of vehicles likely to have exceeded their useful lives at the time the EPA notifies the manufacturer of nonconformity. Even if there were no delay between EPA notice of nonconformity and manufacturer notification of vehicle owners, cars already beyond their useful lives would not be subject to recall by the terms of the report’s language. Yet the May 30 rule would require manufacturers to recall such vehicles. Indeed, the most that can be said for the import of the “fast track” recall system in the Senate bill is that it indicates the possibility that Congress failed to appreciate and to provide relief for the delays that might arise in the recall process when the “fast track” system was removed. However, if the agency wishes to fill in interstices created by a possible congressional failure to anticipate the consequences of amendments to the statutory scheme, the proper recourse is a legislative, not an interpretative, rule. An interpretative rule may neither expand nor contract the statutory form and substance; it must simply construe, without supplementing, the terms of the Act EPA also argues that any inference to be drawn from the language of the Committee Report concerning the manufacturer’s duty to notify and to repair vehicles exceeding their useful lives must be counterbalanced by other segments of the legislative history evidencing Congress’ concern that the industry produce cars capable of meeting emissions standards throughout their actual lives. For example, Senator Muskie, one of the bill’s sponsors, remarked: Throughout discussions with the industry over the past 6 or 7 years, that is what they were stating, 50,000 miles. They do not consider that technology would be effective or worthwhile, in terms of cost to the consumer, unless it meets the 50,-000-mile test. So we are asking for that, because unless automobiles will perform for a practical proportion of their life, meeting standards initially may not be worthwhile. Fifty thousand miles is not all their life, 100,000 miles being nearer to a measure of the life of a motor vehicle, but we have taken 50,000 miles, .. . and we have used that 50,000-mile test on performance. The Committee Report also observed: The [Senate] Committee hopes that, if the motorist complied with [the manufacturer’s maintenance] instructions, emission controls would not deteriorate after 50,000 miles to the extent that ambient air quality would be impaired. The Committee further expects the manufacturer to endeavor to either improve the quality control of emission systems or explore better ways to assure compliance beyond 50,000 miles of use. EPA’s reliance on these passages of the legislative history is misplaced. Neither GM nor this court have any doubt that Congress believed that the actual life of an automobile is closer to ten years and one hundred thousand miles than it is to five years and fifty thousand miles. Nor do we doubt that Congress hoped that, by rigidly enforcing emissions standards during the five-year/fifty thousand-mile statutory useful life, emissions performance would be improved throughout the balance of the car’s actual life. Nevertheless, Congress intentionally limited the duration of a manufacturer’s liability for a vehicle’s conformity to standards to a five-year/fifty thousand-mile period in response to industry representations that the technology could not be guaranteed for a longer period. It is not for the agency to impose under the guise of “interpretation” a more stringent requirement that Congress considered and rejected. Conclusion We are not insensitive to the fact that our holding may impede the enforcement of emissions standards even for cars within their useful lives at the time the EPA first notifies the manufacturer of the nonconformity of a vehicle or engine class. Nonconformity may be discovered only late in a vehicle’s useful life, and only an unusually uninventive lawyer will be unable in the context of today’s backlogged court dockets to prolong administrative and judicial appeals for at least several years. The recall system may be rendered useless if the combination of belated identification of noncon-formities and protracted negotiations and appeals combine to place all or most vehicles involved in a recall beyond their useful lives once owner notification commences. However, this is a difficulty which can only be redressed by congressional intervention, or at a minimum, legislative rulemak-ing with attendant notice and comment procedures. “It is not for an administrative agency ... to preempt Congressional action or to ‘fill in’ where it believes some federal action is needed.” Although the May 30 rule may have been a well-intentioned response to an obvious and precipitous loophole in the statutory framework, the rule, and as a result the GM recall order, exceeded the bounds of reasonable statutory interpretation. Therefore, we vacate both the May 30 rule and that part of the June 23, 1980 order that found that GM had failed to submit a remedial plan with respect to 1975 Cadillacs beyond their useful lives when presented for repair. So ordered. . 42 U.S.C. §§ 7401 et seq. (Supp. V 1981) (hereinafter “the Act”]. . 42 U.S.C. § 7541(c)(1) (Supp. V 1981). . 42 U.S.C. § 7607(b)(1) (Supp. V 1981). . Clean Air Act § 101(b)(1), 42 U.S.C. § 7401(b)(1) (Supp. V 1981). . 42 U.S.C. §§ 7521-7574 (Supp. V 1981). . Clean Air Amendments of 1970 § 6(a), 84 Stat. 1676, 1690 (1970) (current version at 42 U.S.C. § 7521 (Supp. V 1981)). . Id. § 6(a)(1), 84 Stat. 1676, 1690 (1970) (current version at 42 U.S.C. § 7521(a)(1) (Supp. V 1981)). . Id. § 6(a), 84 Stat. 1676, 1692 (1970) (current version at 42 U.S.C. § 7521(d)(1) (Supp. V 1981)). . Id. § 8(a), 84 Stat. 1676, 1694-98 (1970) (current version at 42 U.S.C. §§ 7525, 7541 (Supp. V 1981)). . Clean Air Act § 207(c)(1), 42 U.S.C. 7541(c)(1) (Supp. V 1981). Other aspects of the enforcement mechanism require presale and assembly line testing and certification, see id. § 206, 42 U.S.C. § 7525 (Supp. V 1981), and warranties for the repair of isolated, individual vehicle failures, see id. § 207(a), (b), 42 U.S.C. § 7541(a), (b) (Supp. V 1981). . The investigation was initiated in response to data from state emissions inspections and GM assembly line audits indicating significant hydrocarbon and carbon monoxide problems. See Joint Appendix (J.A.) 1. . The 60V43 engine family consisted of approximately 220,000 vehicles equipped with carburetor part number 7045230 or 7045193 [hereinafter referred to as the 230-carburetor or 193-carburetor, respectively]. See EPA Brief at 5 n. 6. . See J.A. 123. This determination was based on EPA ordered testing of 15 sample vehicles, permitting the EPA to project with 95 percent statistical confidence that at least 68 percent of the vehicles in the class were exceeding standards. See id. at 122-23, 125. Earlier projections estimated a 43 percent nonconformity rate. See J.A. 35. . See J.A. 123. . See J.A. 123-24. . See J.A. 126. GM’s agreement to recall vehicles equipped with 193-carburetors was not purely altruistic. Although GM continued to contend that the 193-carburetor vehicles complied with emissions standards, there was no practical method of determining which of the carburetors was installed in a given 1975 Cadillac until it had been brought into the repair shop. See id. . A manufacturer notified of a determination of nonconformity is required to submit a remedial plan for the Administrator’s approval. See Clean Air Act § 207(c)(1), 42 U.S.C. § 7541(c)(1) (Supp. V 1981); 40 C.F.R. §§ 85.-1803-85.1804 (1982). The manufacturer must demonstrate both that the proposed remedy is technically sound and that it can and will be properly implemented. See 40 C.F.R. §§ 85.-1803-85.1804 (1982). . See J.A. 137, 142, 145, 170, 182, 203, 224, 226, 230, 236, 247, 252. Much has been made by both parties of the time lag between initial notification of nonconformity and the EPA’s ultimate approval of the remedial plan and of the appropriate allocation of blame for the delay. The EPA maintains that its original refusal to approve GM’s proposed remedial plan was premised on two principal concerns. First, GM had failed to demonstrate that the repair proposed for the 193-car-buretor would actually remedy the nonconformity in those vehicles. Second, the adjustment’s actual or perceived adverse effect on the drivability of the Cadillacs might induce mechanics, fearing customer dissatisfaction, to refuse to perform the repair properly. See J.A. 137-38, 142. GM tested the 193-carburetor vehicles and discovered that its proposed repair did not in fact remedy the nonconformity, yet did not submit a revised remedial plan until January 30, 1978. See J.A. 203. A few minor modifications to this plan were made at the request of the EPA and were incorporated in a new revised remedial plan submitted on February 15, 1978. See J.A. 224. This is the plan that was ultimately approved by the EPA. See J.A. 252-55, 261-64. GM attempts to excuse its delay on the ground that the 193-carburetor recall was voluntary and did not require EPA approval. See GM Reply Brief at 12-13. However, once GM’s own testing had revealed that half of the sample of eight vehicles failed to meet carbon monoxide emissions standards and that the repair originally proposed failed to remedy this problem, see J.A. 170, it seems fair to assume that the EPA refrained from ordering a recall of the 193-carburetor class only because the details of the “voluntary” recall were currently being negotiated. To alleviate its concerns with respect to mechanics’ motivation to perform the repairs properly, EPA proposed, among other measures, that GM conduct a performance audit. See J.A. 142^13. GM agreed to some of the measures but refused to assume responsibility for an audit, arguing both that an audit was unnecessary and that it would place GM in an adversary relationship with its own dealers. See J.A. 203-06. Although EPA continued to insist on its authority to require GM to perform the audit, EPA ultimately dropped its demand that GM conduct the audit and decided instead to conduct an audit itself. See J.A. 254. GM argues that, because EPA’s insistence on a GM-conducted audit which “it had no right to demand and which it dropped in the final approval” was the real cause of the delay, see GM Reply Brief at 12, “it is arbitrary and capricious for EPA to assert in 1980 that GM now should repair vehicles which have exceeded their useful lives.” GM Brief at 35. Because of our disposition of EPA’s interpretative rule and order on other grounds, we need not reach the question of EPA’s authority to require manufacturers to conduct recall audits nor need we assign to either party full responsibility for the delay that resulted in the vast majority of the vehicles having exceeded their useful lives by the time the recall plan was approved. See J.A. 258, 261 (estimating only 52,000 of the 220,000 vehicles remained within their useful lives at the time of conditional approval of the remedial plan). We do note, however, that the audit issue was not a frivolous question, that it was hotly contested by both parties for some time in what we can presume to have been good faith, and that it is illustrative of the type of controversy that can prolong approval negotiations despite the best intentions and cause the “useful life” problem to arise again in the future. . See J.A. 252-55. See supra note 18. . J.A. 257. . J.A. 259. GM maintains that this statement, contained in a letter to the EPA dated February 5, 1980, merely reiterated the position GM had first taken in January 1979. See GM Brief at 5. EPA insists that the February 1980 letter “was the first time that GM had proposed to impose a 5/50 limitation on a remedial plan.” EPA Brief at 11. Whether there was verbal communication between GM and EPA on this issue before February 1980 we cannot say; there is, however, nothing in the record before this court to document any exchange between GM and EPA concerning the 5-year/50,000-miIe limitation prior to the February 1980 letter. After the February letter, GM imposed but later withdrew 5-year/50,000-mile limitations in several other pending recalls. See EPA Brief at 11 & nn. 17-18. Consequently, the 1975 Cadillac recall is the only situation which to the court’s knowledge presently involves a dispute between the EPA and a manufacturer concerning the extent of a manufacturer’s liability to recall and repair vehicles beyond their useful lives. . Section 307(d)(2)-(6) of the Clean Air Act, 42 U.S.C. § 7607(d)(2)-(6) (Supp. V 1981), establishes certain procedural requirements in connection with agency promulgation of rules. The Act requires, inter alia, establishment of a rule-making docket, publication of notice of proposed rulemaking in the Federal Register, a specified period for receipt of public comment, public access to the docket materials, and response by EPA to each significant comment, criticism, and submission of data. Id. However, section 307(d)(1) of the Act provides, in relevant part: “This subsection shall not apply in the case of any rule or circumstance, referred to in subparagraphs (A) or (B) of subsection 553(b) of title 5 [of the United States Code].” 42 U.S.C. § 7607(d)(1) (Supp. V 1981). The relevant provision of title 5 provides in turn that rulemaking procedures need not be applied “to interpretative rules, general statements of policy, or rules of agency organization, procedure, or practice .... ” 5 U.S.C. § 553(b)(A) (1982). Thus, by casting the regulation as an interpretative rule, the agency was able to circumvent most of the rulemaking requirements of the Clean Air Act. The EPA did create and provide access to a public docket in connection with the promulgation of the May 30 rule (although the utility of even that exercise must be questioned in light of the absence of the public comments that would normally arouse interest in the docket), but none of the other rulemaking procedures were observed. See 45 Fed.Reg. 36,396, 36,397 (May 30, 1980). . 45 Fed.Reg. 36,396, 36,396-97 (May 30, 1980) (presently codified at 40 C.F.R. § 85.-1803, app. A to subpart S (1982)). . See J.A. 261-64 (EPA Letter dated June 23, 1980). . Id. GM agreed to proceed with the recall of these vehicles. See J.A. 308. . See J.A. 261-64. . J.A. 263. The Administrator warned GM that “[t]he failure to submit a plan for these vehicles is considered to be a separate offense under sections 203(a)(4)(B) and 205 of the Act for each vehicle and can potentially subject GM to fines of up to $10,000 per vehicle.” Id.; see Clean Air Act §§ 203(a)(4)(B), 205, 42 U.S.C. §§ 7522(a)(4)(B), 7524 (Supp. V 1981). . On November 26, 1980, EPA published a notice in the Federal Register declaring its partial approval and partial disapproval of GM’s remedial plan to be “final.” See 45 Fed.Reg. 78,798, 78,798 (Nov. 26, 1980). GM subsequently filed a protective petition, consolidated with its previous appeals, seeking review of this Federal Register notice. . See 45 Fed.Reg. 36,396, 36,396-97 (May 30, 1980). . GM Brief at 12-22. GM also argues that the rule must be considered legislative because of the limitations on judicial review imposed by the Act. GM Brief at 21-22. One of the distinguishing features of an interpretative rule is its lack of binding force upon courts; such rules are always subject to challenge in later judicial proceedings. See, e.g., Batterton v. Francis, 432 U.S. 416, 425 n. 9, 97 S.Ct. 2399, 2405 n. 9, 53 L.Ed.2d 448 (1977) (“[A] court is not required to give effect to an interpretative regulation.”); Citizens to Save Spencer County v. EPA, 600 F.2d 844, 876 (D.C.Cir.1979); Joseph v. United States Civil Serv. Comm’n, 554 F.2d 1140, 1153 n. 24, 1154 n. 26 (D.C.Cir.1977); Gibson Wine Co. v. Snyder, 194 F.2d 329, 331-32 (D.C.Cir.1952); see generally 2 K. Davis, Administrative Law Treatise §§ 7:8, 7:13 (2d ed. 1979 & Supp.1982). However, section 307(b)(1) of the Act provides in relevant part: A petition for review of ... any other nationally applicable regulations promulgated, or final action taken, by the Administrator under this chapter may be filed only in the United States Court of Appeals for the District of Columbia .... Any petition for review under this subsection shall be filed within sixty days from the date notice of such promulgation, approval, or action appears in the Federal Register.... 42 U.S.C. § 7607(b)(1) (Supp. V 1981) (emphasis supplied). Section 307(b)(2) goes on to provide that any “[ajction of the Administrator with respect to which review could have been obtained under paragraph (1) shall not be subject to judicial review in civil or criminal proceedings for enforcement.” 42 U.S.C. § 7607(b)(2) (Supp. V 1981). Apparently, both GM and EPA read this language as rendering any rule, whether such rule was originally intended to be interpretative or legislative in character, binding on the courts if not challenged within 60 days after its promulgation. See GM Brief at 21-22; EPA Brief at 42-46; GM Reply Brief at 2-3, 14-19. Consequently, GM argues, if the review provisions render any nationally applicable regulation binding on the courts after 60 days, then all such regulations must by definition be legislative and not interpretative. See GM Brief at 21-22; GM Reply Brief at 14-19. The Supreme Court has previously taken note of the constitutional problems that review limitations such as section 307(b) may present. See, e.g., Harrison v. PPG Indus., Inc., 446 U.S. 578, 592 n. 9, 100 S.Ct. 1889, 1897 n. 9, 64 L.Ed.2d 525 (1980) (opinion of the Court); id. at 594-95, 100 S.Ct. at 1898-99 (Powell, J., concurring opinion); Adamo Wrecking Co. v. United States, 434 U.S. 275, 289-91, 98 S.Ct. 566, 575-76, 54 L.Ed.2d 538 (1978) (Powell, J., concurring opinion). The issue, however, has not thus far arisen in a posture that would permit its resolution, nor does it so arise in this case. Suffice it to say that a limitation on judicial review does not lend to an interpretative rule any binding force not already provided by the underlying statute. See Citizens to Save Spencer County v. EPA, 600 F.2d 844, 876 (D.C.Cir.1979). A court always has the power to substitute its judgment for that of the agency in the case of an interpretative rule, even though courts customarily accord some measure of deference to an agency’s interpretation of a statute for which it has been assigned the responsibility for enforcement. See Batterton v. Francis, 432 U.S. 416, 425 n. 9, 97 S.Ct. 2399, 2405 n. 9, 53 L.Ed.2d 448 (1977); General Elec. Co. v. Gilbert, 429 U.S. 125, 141-42, 97 S.Ct. 401, 410-11, 50 L.Ed.2d 343 (1976); Skidmore v. Swift & Co., 323 U.S. 134, 140, 65 S.Ct. 161, 164, 89 L.Ed. 124 (1944); 2 K. Davis, supra, at §§ 7:8, 7:13. Thus, in cases arising after the 60-day review period has expired, the court may undertake and enforce its own interpretation of the statute without reviewing the interpretative rule as such. And, in any event, the 60-day limitation period may not begin to run with respect to interpretative rules until a fact-based controversy is ripe for judicial review. See Baltimore Gas and Elec. Co. v. ICC, 672 F.2d 146, 147-50 (D.C.Cir.1982) (interpreting an analogous 60-day review limitation). . See Citizens to Save Spencer County v. EPA, 600 F.2d 844, 873-74 (D.C.Cir.1979). . Chamber of Commerce v. Occupational Safety and Health Admin., 636 F.2d 464, 468 (D.C.Cir.1980) (opinion of the court); id. at 471 (Bazelon, J., concurring opinion). . Id. at 468; Joseph v. United States Civil Serv. Comm’n, 554 F.2d 1140, 1153 n. 24 (D.C. Cir.1977); see generally 2 K. Davis, supra note 30, at §§ 7:8-7:13, 7:15. . See Columbia Broadcasting Sys., Inc. v. United States, 316 U.S. 407, 416, 62 S.Ct. 1194, 1199, 86 L.Ed. 1563 (1942); Chamber of Commerce, 636 F.2d at 468; Citizens to Save Spencer County, 600 F.2d at 879 n. 171. . See Chamber of Commerce, 636 F.2d at 468; see also Udall v. Tallman, 380 U.S. 1, 16-17, 85 S.Ct. 792, 801-802, 13 L.Ed.2d 616 (1965); Energy Reserves Group, Inc. v. Department of Energy, 589 F.2d 1082, 1092 (Em.App.1978). . When the rule was first published it was classified as “interpretive” and described as setjting] forth EPA’s interpretation regarding one aspect of a motor vehicle or motor vehicle engine manufacturer’s recall liability under section 207(c)(1) of the Clean Air Act . EPA interprets this section as requiring manufacturers to submit a plan to remedy all vehicles within the class or category of vehicles subject to an ordered recall ... regardless of their age or mileage at the time of repair. The interpretation set out in this rule will provide guidance to vehicle and engine manufacturers to better enable them to submit acceptable remedial plans. 45 Fed.Reg. 36,396, 36,396-97 (May 30, 1980) (emphasis supplied). Throughout this litigation, EPA has never claimed any authority for its rule other than that conveyed by the statute itself. Thus, there is no need to turn to the effect of the rule in order to ascertain agency intent. If the effect of the rule exceeds that of the statute itself, then the rule will have surpassed the bounds of permissible statutory interpretation. . Batterton v. Francis, 432 U.S. 416, 424, 97 S.Ct. 2399, 2405, 53 L.Ed.2d 448 (1977). . General Elec. Co. v. Gilbert, 429 U.S. 125, 142, 97 S.Ct. 401, 411, 50 L.Ed.2d 343 (1976) (quoting Skidmore v. Swift & Co., 323 U.S. 134, 140, 65 S.Ct. 161, 164, 89 L.Ed. 124 (1944); see id. 429 U.S. at 141-42, 151-53, 97 S.Ct. at 410-11, 415-16; Batterton, 432 U.S. at 425 n. 9, 97 S.Ct. at 2405 n. 9 (“[A] court is not required to give effect to an interpretative regulation. Varying degrees of deference are accorded to administrative interpretations, based on such factors as the timing and consistency of the agency’s position and the nature of its expertise.”). . General Elec. Co., 429 U.S. at 143, 97 S.Ct. at 411. . EPA argues that its interpretation is longstanding. Yet, with the exception of the dispute with GM leading to the promulgation of the May 30 rule, the agency can point to only one controversy, the Chrysler recall in February 1979, during which the useful life issue arose. See EPA Brief at 5 & n. 5. However, by the EPA’s own admission, the agency did not press the issue with Chrysler, because the question of nonconformity was itself being litigated. See id. Thus, even assuming that a lifespan barely exceeding one year suffices to render an interpretation longstanding, it would appear that the public was never placed on notice with respect to EPA’s interpretation. The dissent contends that, nevertheless, the EPA interpretation should be considered longstanding precisely because the question of a manufacturer’s liability for repair of vehicles beyond their useful lives has never previously arisen. See Dissenting Opinion at 1003 n. 5. Yet, this novel approach to defining “longstanding” interpretations is not supported by the cases on which the dissent relies. In Esquire, Inc. v. Ringer, 591 F.2d 796, 801 (D.C.Cir. 1978), we did hold that an administrative interpretation that “has been consistently followed for a significant period of time” is entitled to judicial deference. Yet we found that the administrative interpretation under review in that case was longstanding, because it had been consistently asserted and followed by the agency since 1909 in the face of a long series of challenges. Id. at 802 & nn. 19-20. Similarly, in DeLano v. United States, 393 F.2d 517, 521-22, 183 Ct.Cl. 379 (1968), the railroad’s failure to challenge overtime billing for customs inspections performed on board trains was seen as indicative of the “reasonableness” (not the longstanding nature) of the Immigration Service’s interpretation of its overtime regulations. There, the agency’s policy had been publicly announced and implemented for over 15 years. Id. 393 F.2d at 521. Railroads had been affirmatively required to pay on the basis of the agency’s overtime interpretation. In the instant case, however, the EPA has previously announced its position only once and, in that instance, recanted. We are pointed to no other occasion on which the EPA required manufacturers to include vehicles beyond their useful lives in their remedial plans; the issue simply was never raised by either party. Judge Wald’s view of the EPA’s position as “longstanding” apparently presumes that manufacturers (1) prophetically foresaw EPA’s heretofore unannounced interpretation of their liability for repair of vehicles beyond their useful lives; (2) conceded the “reasonableness” of EPA’s position; and (3) therefore, “voluntarily” included older vehicles in their remedial plans, refraining from challenging this manifestly “reasonable” interpretation. A far more plausible explanation for the lack of 5/50 controversies avails itself. The statistical data supplied by the EPA indicates that, with the exception of the Chrysler and Cadillac recalls in which useful life disputes did arise, virtually all completed emissions recalls have involved vehicle classes less than three years of age at the time of owner notification of nonconformity by the manufacturer. See J.A. 278-96. The median class age is approximately 2.5 years; the mean, slightly less than 2.6 years, and the mode, less than one year. EPA data also reveal that, at an age of one year, only 2% of vehicles are likely to have exceeded 50,000 miles; at age 2, 6%; at age 3, 20%; at age 4, 39%; and at age 5, 60%. J.A. 274. Finally, EPA statistics indicate that the owner response rates drop substantially for vehicles more than four-years old (median owner response for vehicles between four and five years of age is between 20 and 30%, and no owner response is indicated for vehicles more than five years of age). See J.A. 298. This information forms a strong basis for the inference that both the EPA and manufacturers in previous recalls have failed to raise the question of liability for repair of cars beyond their useful lives because the involvement of such vehicles in the recalls has been de minimis. Few vehicles beyond their useful lives were putatively subject to recalls; of these, only a handful could have been expected actually to have been presented for repair. It simply would not have been economical for manufacturers to have expended the resources necessary to identify those vehicles beyond their useful lives or to risk owner alienation and possible reaction from the EPA (although the EPA’s position concerning such matters was at the time unknown) by refusing to repair the tiny percentage of older cars which might have found their ways to dealerships in response to a recall. In contrast, in the two major recalls in which 5/50 limitations were imposed, a substantial percentage of the vehicles in the class were beyond their useful lives. The EPA recalled 1975 Chryslers in December 1976. J.A. 281. After a hotly contested dispute over the existence of the emissions nonconformity, Chrysler imposed a 5/50 limitation in December 1978. Id. At that point, the class was 51 months old and over 40% of the vehicles had exceeded their useful lives. Moreover, since Chrysler’s appeal of the EPA’s recall order was still pending in this court, virtually the entire class could have been expected to have exceeded their useful lives before owner notification took place. Similarly, in this case, the model was beyond the five-year point by February 1980 when, after a prolonged and often hostile negotiation process, GM finally imposed a 5/50 limitation on repair of its 1975 Cadillacs. See J.A. 286. The dissent responds that “in six out of the fifteen recalls described in detail in the record, ... the EPA issued a recall notice to the manufacturer 34-55 months after the cars were on the market.” Dissenting Opinion at 1003 n. 5. However, we believe that the dissent’s method of computation produces an inflated picture of the average age of recall classes. For example, the dissent includes two recalls which were incomplete at the time the record in this case was submitted. However, it is impossible to tell, until owner notification has begun, whether a manufacturer will ultimately assert a 5/50 limitation. For example, in this case GM submitted several remedial plans without asserting such a limitation; indeed, the 5/50 limitation was not imposed until after the EPA had approved GM’s remedial plan. Thus, incomplete recalls are useless for purposes of discerning why manufacturers failed to impose 5/50 limitations prior to the Cadillac recall. The dissent also presents figures on three recalls that “involved” large numbers of older vehicles (1975 and 1976 Pontiacs, 1975 Fords, 1974 AMCs); yet, in each case, these vehicles were part of a larger recall class, spanning several model years, in which the majority of the cars involved were younger. We respectfully suggest that employing the mean age of the recall class as a whole would present a more accurate picture of the proportion of older cars in a given recall that the manufacturer might have had to fix without a 5/50 limitation. For example, 54% of the 1975 model Pontiacs may have been beyond their useful lives in 1979, yet 1975 Pontiacs represented only one-quarter of the cars involved in the Pontiac recall which included model years 1975-1978. Similarly, the 1978 Ford recall to which the dissent refers included 1975 and 1976 Fords, and the 1978 AMC recall spanned the model-years 1974 through 1976. The AMC recall points to another flaw in the dissent’s statistical formulation. In that recall, the EPA ordered the recall of all 1976 AMC cars in May 1978. At that point, the model was approximately two-and-a-half years old and fewer than 20% of the vehicles could have been expected to have exceeded their useful lives. We are told that AMC “indicate[d] it [would] include 1974, 1975, and 1976 AMC cars sold in California in its plan.” J.A. 291. The exact motive for this inclusion is not disclosed by the record, but it is clear that AMC included these vehicles voluntarily, perhaps out of a concern for their ability to meet more stringent California state emissions requirements. However, the dissent points to the 1974 AMCs included in this recall as evidence of a class involving large numbers of cars beyond their useful lives for which the manufacturers did not impose a 5/50 limitation. We would suggest, with all due deference, that such use of data is inappropriate (1) because only California vehicles from the 1974 and 1975 AMC model-years were included in the recall, thus making them a relatively small proportion of the recall class which was overwhelmingly composed of 1976 AMCs; and (2) these vehicles were included voluntarily, thus indicating the manufacturer’s willingness to go beyond EPA’s requirements. Finally, the dissent presents its data only in terms of the percentage of older cars it contends were present in the recall class; however, that figure must be multiplied by the projected owner response rate to determine the percentage of cars the manufacturer would have anticipated actually having to fix. Thus, even if 20% of the cars in a given class were beyond their useful lives at the time of owner notification, and a 20% owner response rate could be expected for those cars, only 4% of the cars presented for repair could be expected to be beyond their useful lives. (This figure, of course, would have to be adjusted by the response rates for vehicles of varying ages within their useful lives.) In fact, as indicated above, the response rate for vehicles beyond their useful lives approaches zero. Therefore, even assuming the statistical validity of the dissent’s calculations, manufacturers in past recalls would have had to repair few, if any, vehicles beyond their useful lives, whether or not 5/50 limitations had been imposed. In the few recalls, other than the Chrysler and Cadillac recalls, in which significant numbers of vehicles beyond their useful lives were involved, potent reasons existed for the manufacturers to refrain from invoking 5/50 limitations. One manufacturer “requested additional time to develop a durable replacement component to use in a recall campaign.” J.A. 268. The manufacturer’s failure to adhere to its own production timetable resulted in approximately 70,000 vehicles (most of them apparently from earlier model years voluntarily included in the recall, see J.A. 291) having exceeded their useful lives by the time owner notification began. J.A. 268. Similarly, another manufacturer requested additional time to develop a less expensive remedial repair part, resulting in savings to the manufacturer of $3.2 million. J.A. 269. A third carmaker convinced the EPA to consent to "two audited sample recalls in an unsuccessful attempt to demonstrate that defective carburetors could be repaired instead of replaced; the delays consumed three years. Id. Manufacturers who hope to call upon the EPA’s good graces in the future are hardly in a position to invoke 5/50 limitations when the delays they have asked the EPA to tolerate in order to mitigate the expense of a recall result in large numbers of vehicles having exceeded their useful lives when presented for repair. Indeed, EPA provides information on only two completed recalls in which large numbers of vehicles beyond their useful lives could have been expected to have been presented for repair and special circumstances militating against a manufacturer’s invocation of a 5/50 limitation were not involved. These were the Chrysler and the Cadillac recalls. In both, the manufacturers imposed 5/50 limitations. In the former, the EPA ultimately acceded to the limitation; in the latter, obviously, it has not. Such a history does not a “longstanding” interpretation make. . EPA Brief at 41. . Whirlpool Corp. v. Marshall, 445 U.S. 1, 11, 100 S.Ct. 883, 890, 63 L.Ed.2d 154 (1980); Southeastern Community College v. Davis, 442 U.S. 397, 411, 99 S.Ct. 2361, 2369, 60 L.Ed.2d 980 (1979). . See Consumer Prod. Safety Comm’n v. GTE Sylvania, Inc., 447 U.S. 102, 108, 100 S.Ct. 2051, 2056, 64 L.Ed.2d 766 (1980); Touche Ross & Co. v. Redington, 442 U.S. 560, 568, 99 S.Ct. 2479, 2485, 61 L.Ed.2d 82 (1979); Caminetti v. United States, 242 U.S. 470, 485, 37 S.Ct. 192, 194, 61 L.Ed. 442 (1917); Symons v. Chrysler Corp. Loan Guar. Bd., 670 F.2d 238, 241 (D.C.Cir.1981); Higgins v. Marshall, 584 F.2d 1035, 1037 (D.C.Cir.1978), cert. denied, 441 U.S. 931, 99 S.Ct. 2051, 60 L.Ed.2d 659 (1979). . Clean Air Act § 207(c)(1), 42 U.S.C. § 7541(c)(1) (Supp. V 1981). .See EPA Brief at 22 & n. 26. . GM Brief at 20; see id. at 19-20, 23-26. . The dissent, like the EPA, makes much of the classwide nature of the manufacturer’s remedial obligation. See Dissenting Opinion at 1002-1005, 1006, 1007 nn. 11 & 12, 18-19; see also infra note 50. Yet, assuming arguendo that it is the nonconformity of the class—and not the nonconformity of individual vehicles—that the statute requires manufacturers to remedy, see Dissenting Opinion at 1007 n. 11 (“Congress intended to provide classwide remedies for classwide ... defects that showed up during the useful life of the cars listed as representative samples of the class. A defective class — its members (sic) — can therefore be deemed ‘nonconforming.’ ”), a manufacturer still need not “repair” vehicles beyond their useful lives to effect a remedy of the class. If a class is nonconforming when a representative sample of its members demonstrates nonconformity with emissions standards during the useful lives of the sample vehicles, then it follows that a class is restored to conformity when the EPA can no longer muster such a sample. Since vehicles beyond their useful lives can never form part of the sample upon which an inference of class conformity or nonconformity is founded, “repair” of such vehi-cíes, as noted above, contributes nothing to the remedy of the class. . See 45 Fed.Reg. 36,396, 36,397 n. 2 (May 30, 1980); EPA Brief at 20 n. 25. . 45 Fed.Reg. 36,396, 36,397 n. 2 (May 30, 1980). . The dissent’s apparent approval of EPA’s concession, see Dissenting Opinion at 1007 n. 12, is similarly inconsistent with the dissent’s position that all vehicles which are part of a nonconforming class are subject to repair at manufacturer expense, see id. at 1002-1005, 1007 & n. 12. If, as the dissent believes, the “statute ... require[s] the manufacturers to include a car in their remedial plans if (1) a substantial number of cars exhibit nonconformity during their useful lives, and (2) the car is a member of that class,” id. at 1002; see id. at 1006 n. 10, then the time of the onset of nonconformity in a particular vehicle should be irrelevant to the manufacturer’s liability for repair of that car. Indeed, the dissent’s analysis produces the “bizarre result,” id. at 1002, that even the existence of an individual vehicle’s conformity or nonconformity is beside the point; class conformity is all that matters! Id. at 1007 n. 11 (“Congress intended to provide classwide remedies for classwide design or performance defects that showed up during the useful life of the cars listed as representative samples of the class. A defective class — its members (sic) — can therefore be deemed ‘nonconforming.’ ” (emphasis supplied)); id. at 1008-1009 (“The statute thus mandates class-wide notice without limitation, and requires the manufacturer to repair all cars and engines ‘with respect to which [such] notice is given.’ ” (emphasis and alteration in original)); id. at 1009 (“The statute may fairly be read to require classwide notice, regardless of whether some members of the class have exceeded their useful lives.”); see generally id. at 1002-1005. Yet, pragmatically, the only way to remedy class nonconformity is by repairing individual vehicles. And, to paraphrase the familiar adage, “if it ain’t broke, you can’t fix it.” . In this case, the samples appear to have been 14 or 15 cars of the 230-carburetor family and 8 of the 193-carburetor family. See J.A. 125, 170. . See J.A. 35. . See J.A. 123. . Thus, contrary to the contention of the dissent, better than half of the cars in the class may not “have been riding the roads for years in violation of pollution standards.” Dissenting Opinion at 1001. Moreover, whatever “public benefit” might be derived from improving the emissions performance of vehicles beyond their useful lives, see id. at 1001-1002, may not be achieved by imposing on manufacturers a liability beyond that which Congress has seen fit to prescribe. . See S.Rep. No. 1196, 91st Cong., 2d Sess. 30 (1970), reprinted in 1 Senate Comm, on Pub. Works, 93rd Cong., 2d Sess., A Legislative History of the Clean Air Amendments of 1970, at 430 (1974) [hereinafter cited as 1970 Legislative History]. Indeed, Congress reduced the useful life to which emissions standards would apply from 10 years to 50,000 miles in response to industry representations that manufacturers could not guarantee conformity for a 10-year useful life. See id. . See J.A. 165-67. . Se