Full opinion text
Opinion for the Court filed by Circuit Judge SCALIA. Dissenting opinion filed by Circuit Judge J. SKELLY WRIGHT. SCALIA, Circuit Judge: Since 1971 the National Highway Traffic Safety Administration (“NHTSA”) has prescribed a minimum performance standard for automobile bumpers, originally under the National Traffic and Motor Vehicle Safety Act of 1966 (“the Safety Act”), Pub.L. No. 89-563, 80 Stat. 718 (1966) (codified as amended at 15 U.S.C. §§ 1381-1431 (1982)), and later under concurrent authority of the Safety Act and the Motor Vehicle Information and Cost Savings Act of 1972 (“the Cost Savings Act”), Pub.L. No. 92-513, 86 Stat. 947 (1972) (codified as amended at 15 U.S.C. §§ 1901-1991 (1982)). In the consolidated petitions before us we are asked to review the most recent change in that standard, which reduces the primary test impact speed from 5.0 mph to 2.5 mph, with the effect that lighter, less protective, and less costly bumpers will now satisfy the standard and presumably be made available to consumers. I. Background On April 9, 1971, NHTSA adopted the first federal bumper standard pursuant to its authority under the Safety Act. Federal Motor Vehicle Safety Standard No. 215, 36 Fed.Reg. 7218 (1971) (codified at 49 C.F.R. § 571.215 (1971)) (“Standard 215”), required passenger cars to withstand specified collision impacts without sustaining damage to the vehicles’ safety systems, i.e., “lighting, fuel, exhaust, cooling, or latching systems.” Id. For cars manufactured after September 1, 1972 (model year 1973), the rule specified a series of perpendicular barrier impacts at 5.0 mph for front and 2.5 mph for rear bumper systems. Post-1973 model year cars were required to meet the same damage criteria after additional front, rear, and corner pendulum impacts. After two amendments to the standard — see 36 Fed.Reg. 11,852 (1971); 36 Fed.Reg. 20,369 (1971) — the pendulum impact test speeds were set at 5.0 mph for the front and rear and 3.0 mph for corners, at a height of 16-20 inches, and the perpendicular barrier impact speed for rear bumpers was raised to 5.0 mph. Standard 215 was predicated on two safety-related concerns. First, in higher speed collisions bumpers are prone to over- or underride and interlock with other bumpers and guardrails. Vehicles immobilized in tes fashion pose a safety hazard to oncoming traffic and to vehicle occupants who leave their vehicles to remedy the condition. See Federal Motor Vehicle Safety Standard No. 215, 35 Fed.Reg. 17,999, 17,999 (1970) (proposed Nov. 13, 1970). The pendulum test, applied uniformly at a height of 16-20 inches, effectively required standardization of bumper heights, and this alone significantly reduced the likelihood of bumper interlock and subsequent damage or injury. Second, the agency assumed an indirect relationship between bumper performance in low speed accidents and motor vehicle safety. Although the low-speed test damage criteria were not expected to assure protection for the vehicle’s occupants in collisions at speeds likely to result in death or injury, they were meant to prevent damage to the vehicle’s safety systerns in minor accidents. It was thought that such damage, if left unrepaired, might cause, contribute to, or increase the severity of a subsequent, more serious accident. As the agency concluded in its notice of proposed rulemaking: [M]any accidents of the parking lot variéty impair the safety but not the mobility of the involved vehicles. A vehicle driven without proper lights, or with leaks in its fuel, exhaust, or cooling systems is a hazard. Similarly, a hood or trunk with a damaged latch may spring up to block the driver’s vision, and a defective door latch may either open unexpectedly or fail to open in an emergency. Id. The agency did not test or otherwise provide empirical support for the proposition that safety system damage in low speed accidents was a significant factor in subsequent accidents, Shortly after NHTSA promulgated Standard 215, Congress enacted the Cost Savings Act, the stated purpose of which was “to reduce the economic loss resulting from damage to passenger motor vehicles involved in motor vehicle accidents.” 15 U.S.C. § 1911(a). The Act directs the Sec-rotary of Transportation and, by delegation, NHTSA, see 49 C.F.R. § 1.50(f) (1983), to promulgate a bumper standard that will seek to obtain the maximum feasible re-Auction of costs to the public and to the consumer— 15 U.S.C. § 1912(b)(1). In conducting the statutorily required cost-benefit analysis, NHTSA is to take into account: ^) the cost of implementing the stan-^ard an<^_ the benefits attainable as the result of implementation of the standard; (®) the effect of implementation of the standard on the cost of insurance and prospective legal fees and costs; (C) savings in terms of consumer time and inconvenience; and (D) considerations of health and safety, including emission standards, Id. At the same time, the Cost Savings Act prohibits NHTSA from promulgating a bumper standard under that Act which conflicts with any safety standard under the Safety Act. Id. at § 1912(b)(2). See S.Rep. No. 413, 92d Cong., 1st Sess. 20 (1971). NHTSA issued its first post-Cost Savings Act bumper standard, referred to as the Part 581 standard, on February 27, 1976. 41 Fed.Reg. 9346 (1976) (codified at 49 C.F.R. Part 581 (1976)). Before the rule was finally promulgated, however, NHTSA had issued no fewer than three different proposals. The first proposal incorporated the test speed impacts and procedures of Standard 215 and established — beyond the Standard 215 criteria prohibiting damage to safety systems — a “no-damage” standard requiring that, after testing, the vehicle should experience no dents in or separations of exterior surface material, no breakage or release of fasteners or joints, and no separations of paint, polymeric coatings, or other bonded coverings. 38 Fed.Reg. 20,-899 (1973) (proposed July 30, 1973). In December 1974, this was replaced by a proposal that would have reduced the Standard 215 test impact speeds to 1.5 mph for corner impacts and 2.5 mph for front and rear impacts until 1979, and to 2.5 mph for corner impacts and 4.0 mph for front and rear impacts thereafter; and would have reduced the severity of the damage-resistance criteria of the 1973 proposal. 40 Fed. Reg. 10 (1975) (proposed Dec. 27, 1974). In March 1975, NHTSA withdrew the second proposal and issued a notice that differed only slightly from the original Part 581 proposal. 40 Fed.Reg. 11,598 (1975) (proposed Mar. 7, 1975). Finally, in February 1976, the agency promulgated the first final Part 581 standard. 41 Fed.Reg. 9346. This adopted the Standard 215 test speeds and procedures (reducing the number of front and rear pendulum impacts to two) and established new no-damage criteria to become effective in two “phases.” Model year (“MY”) 1979 vehicles had to satisfy Phase I criteria, which consisted of the old Standard 215 safety criteria (no damage to lights, fuel and cooling systems, etc.) and the new requirement of no damage to any exterior surface other than the bumper face bar itself (including related components and fasteners). Phase II criteria, applicable to MY 1980 cars, required in addition that the bumper face bar suffer only minimal damage or displacement. The agency’s ongoing examination of bumper standards did not end with the final rule in 1976. In 1977, NHTSA issued two proposed amendments to the standard. The first would have delayed the effective date of the Phase II no-damage criteria one year. 42 Fed.Reg. 10,862 (1977) (proposed Feb. 22, 1977). The second proposed three alternatives to the Phase II criteria: (1) a one-year delay of Phase II, (2) a one-year delay with a consumer information program on bumper performance in the interim, and (3) a substitution of the information program for Phase II. 42 Fed.Reg. 30,655 (1977) (proposed June 13, 1977). On November 1, 1977, both proposals were withdrawn. 42 Fed.Reg. 57,979 (1977). That same year, NHTSA undertook a multi-year evaluation of the benefits and costs associated with the 1976 standard and alternatives. See 47 Fed.Reg. 21,821 (1982). Congress was also concerned about the inadequacy of the economic basis for the standard. By 1977, when NHTSA began its multi-year evaluation, the only cost-benefit analysis undertaken by the agency had supported a 2.5 mph rather than a 5.0 mph standard. See 40 Fed.Reg. 10. The agency’s adoption of the higher standard in 1976 was due, not to any new data or recalculations of costs and benefits, but to the agency’s belief that the 5.0 mph standard “best carries out the intent of Congress with respect to bumper protection.” 40 Fed.Reg. 11,599. As explained by Joan Claybrook, then Administrator of NHTSA: While setting a standard which results in the greatest net benefit to the consumer is certainly consistent with Title I of the Cost Savings Act, much of the work that has been conducted thus far applies primarily to a 5-mph Bumper Standard. This was certainly appropriate given the legislative history surrounding the passage of the Motor Vehicle and Cost Savings Act in 1972 [sic]. For example, the debate on the House floor left little question that the initial standards under Title I were to be expressed in terms of a 5-mph requirement. The reaction of Congress in early 1975 when the Department proposed to reduce the 5-mph damage resistance level of the standard is further evidence of Congressional sentiment. Letter from Joan Claybrook to Richard Co-paken at 1 (Oct. 30, 1978), reprinted in Comments of Houdaille Industries on the Bumper Standard, DOT HS-803 867, Submissions to the NHTSA between Dec. 27, 1978, and Feb. 15, 1979, at 30. While that may have been the sentiment of Congress in 1972 and 1975, see pages 1350-51, infra, by 1978 the situation had evidently changed. In its report on NHTSA’s fiscal year 1979 appropriations, Department of Transportation and Related Agencies Appropriation Act, Pub.L. No. 95-335, 92 Stat. 435 (1978), the Senate Appropriations Committee expressed its expectation that $300,-000 of the lump-sum appropriation would be used to study the bumper standard, to assure that the Secretary’s obligations ... in setting bumper standards ... are fulfilled through consideration and evaluation of a range of possible standards ... to determine to the maximum extent the standard that is most cost beneficial to the public and the consumer, that is, where marginal benefits equal marginal costs. S.Rep. No. 938, 95th Cong., 2d Sess. 25 (1978). NHTSA was called on to “modify its bumper standard to reflect the results of this study.” Id. NHTSA began a separate study at once, even though its multi-year evaluation begun in 1977 was still in progress. The first cost-benefit study to be completed, however, was neither of these agency efforts but an analysis submitted to NHTSA late in 1978 by Houdaille Industries, a West Virginia bumper manufacturer. This compared the performance under laboratory conditions of Chrysler’s bumper system designed for the Mexico market with Chrysler’s bumper system designed to meet NHTSA’s 5.0 mph test. Comments of Houdaille Industries, Inc. (Dec. 27, 1978) (“Houdaille Study”), reprinted in Comments of Houdaille Industries on the Bumper Standard, supra, at 9-232. The Mexican bumpers were thought to provide approximately the same level of protection as would be required by a 2.5 mph standard. Houdaille Study at 18. The study concluded that the 2.5 mph system would save the consumer $136 per vehicle relative to the 5.0 mph system. Id. at 20. Early in 1979, NHTSA published the preliminary version of its Senate-requested study. The analysis showed that 2.5 mph bumpers offered greater net benefits than 5.0 mph bumpers. Department of Transportation, NHTSA, Analysis of the Bumper Standard 5 (Jan. 26, 1979), Joint Appendix (“J.A.”) 1250; Department of Transportation, NHTSA, Calculations and Supporting Material for the Preliminary Analysis of the Bumper Standard 17 (Feb. 26, 1979), J.A. 1201. See also Department of Transportation and Related Agencies Appropriations for Fiscal Year 1980: Hearings on H.R. 4440 Before the Subcomm. on Transportation Appropriations of the Senate Comm, on Appropriations, 96th Cong., 1st Sess. 348-49, 360-61 (1979). In June of 1979, however, the agency revised its Preliminary Assessment and published what was represented as a final assessment of the bumper standard. Department of Transportation, NHTSA, Final Assessment of the Bumper Standard (June 1979), J.A. 979-1182. The agency later condensed the report’s many results, each based on separate assumptions, into the conclusion that a 5.0 mph standard was $39 more cost-beneficial than the 2.5 mph standard. Department of Transportation, NHTSA, Commentary on Critiques of the June 1 Bumper Standard Assessment 6 (Dec. 1979), J.A. 922. This conclusion was modified somewhat in response to criticism, the final calculus giving the existing standard an $ll-$29 advantage over the 2.5 mph standard. Id. The 1979 “final” assessment, however, was not to be the last word. In April of 1981, several months after the Reagan administration took office, the Office of the Press Secretary of the White House announced that, as part of the Administration’s program to assist the economically troubled United States auto industry, NHTSA was transmitting for publication in the Federal Register a Notice of Intent (attached to the press release) to modify various regulatory requirements, including the bumper standard. White House Press Office, Actions to Help the U.S. Auto Industry at 4 & A-43 (Apr. 6, 1981), J.A. 2652, 2697. The NHTSA Notice of Intent appeared in the Federal Register three days later, 46 Fed.Reg. 21,203 (1981), soon followed by release of the results of the multi-year study begun by the agency in 1977. This comprehensive report on the Part 581 standard concluded that while 5.0 mph front bumpers might be cost-effective — i.e., provide consumers with strictly positive net benefits relative to no regulation — 5.0 mph rear bumpers clearly were not. Department of Transportation, NHTSA, Evaluation of the Bumper Standard 2-19 (Apr. 1981), J.A. 719. In July 1981, NHTSA released its Preliminary Regulatory Impact Analysis (“PRIA”), which analyzed five alternative bumper standards. PRIA at IY-2 to -3, J.A. 399-400. It then published a notice of proposed rulemaking soliciting comments on nine alternative bumper standards as well as the original Part 581 standard. 46 Fed.Reg. 48,262 (1981) (proposed Sept. 25, 1981). On May 14, 1982, after receiving more than 200 docket submissions from automobile insurers, automobile manufacturers, consumer groups, and interested individuals, NHTSA issued its final decision, explained by an analysis occupying 18 pages in the Federal Register, which incorporated by reference a 263-page Final Regulatory Impact Analysis (“FRIA,” J.A. 106-368). Bumper Standard, 47 Fed.Reg. 21,820 (1982) (codified at 49 C.F.R. §§ 581.-5, 581.6 (1983)). The new rule reduced from 5.0 mph to 2.5 mph the fixed barrier and pendulum impact tests for both front and rear bumpers, reduced from 3.0 mph to 1.5 mph the corner pendulum impact tests, and eliminated the Phase II damage criteria, i.e., the bumper itself was no longer required to be undamaged after the impact tests. The latter two changes are not here in dispute. After the agency denied petitions for reconsideration filed by the Insurance Institute for Highway Safety and several individuals, 47 Fed.Reg. 56,640 (1982), petitions for review were filed in this court by the Center for Auto Safety (“CFAS”) and two automobile insurance companies — State Farm Mutual and Allstate (“Insurance Company Petitioners”) — pursuant to 15 U.S.C. § 1394 (1982). II. Standard of Review Our review of NHTSA’s action is governed by the provisions of 5 U.S.C. § 706 (1982). Thus, the agency’s decision must be upheld unless it is “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law.” Id. at § 706(2)(A). As we have recently been reminded, “[t]he scope of review under the ‘arbitrary and capricious’ standard is narrow and a court is not to substitute its judgment for that of the agency.” Motor Vehicle Manufacturers Ass’n, Inc. v. State Farm Mutual Automobile Insurance Co., 463 U.S. 29, 103 S.Ct. 2856, 2866, 77 L.Ed.2d 443 (1983) (“State Farm"). This is especially true when the agency is called upon to weigh the costs and benefits of alternative policies, since “[s]uch cost-benefit analyses epitomize the types of decisions that are most appropriately entrusted to the expertise of an agency____” Office of Communication of the United Church of Christ v. FCC, 707 F.2d 1413, 1440 (D.C.Cir.1983). Our role is to determine “ ‘whether the decision was based on a consideration of the relevant factors and whether there has been a clear error of judgment.’ ” State Farm, 103 S.Ct. at 2867 (quoting Citizens to Preserve Overton Park, Inc. v. Volpe, 401 U.S. 402, 416, 91 S.Ct. 814, 823, 28 L.Ed.2d 136 (1971)). Petitioners assert that “because the agency has reversed a longstanding policy,” this court must exercise a “more heightened and exacting scrutiny,” Brief for Petitioner CFAS at 14, and apply “particularly careful judicial review,” Memorandum for Petitioner State Farm on the Impact of the Supreme Court’s Passive Restraint Decision on This Case at 7. Even accepting the premise that this 13-year-old rule, frequently targeted for revision on the basis of vacillating estimates of efficacy, qualifies as a longstanding policy, the conclusion is in error. The Supreme Court has made clear that “the same test” applies to the rescission or modification of a rule as to its initial promulgation — the “arbitrary or capricious” standard of 5 U.S.C. § 706(2)(A) (1982) — and that there is “no difference in the scope of judicial review depending upon the nature of the agency’s action.” State Farm, 103 S.Ct. at 2866. The same “presumption ... against changes in current policy that are not justified by the rulemaking record,” id., exists whether those changes consist of enacting a new rule or of revoking or modifying an old one. To overcome the presumption the agency “must examine the relevant data and articulate a satisfactory explanation for its action.” Id. We proceed to analyze NHTSA’s substitution of standardized-height 2.5 mph bumpers for standardized-height 5.0 mph bumpers under this standard. III. Safety Petitioners charge that NHTSA was arbitrary and capricious in concluding that the amended bumper standard “would not compromise any known safety consideration,” 45 Fed.Reg. 21,835. It is undisputed that the Part 581 bumper standard is, in part, a safety standard, purportedly complying with a mandate to consider safety contained in both the Safety Act and Cost Savings Act. The former directs the Secretary of Transportation (and, by delegation, NHTSA, see 49 C.F.R. § 1.50(a) (1983)) to “establish motor vehicle safety standards” designed to “reduce traffic accidents and deaths and injuries to persons resulting from traffic accidents,” 15 U.S.C. § 1381 (1982); the latter requires NHTSA to consider “health and safety” in conducting its analysis of bumper standard costs and benefits, 15 U.S.C. § 1912(b)(1)(D) (1982), and prohibits the promulgation of any standard that would conflict with a Safety Act stan- • dard, id. at § 1912(b)(2). The Safety Act’s mandate is not, however, categorical. Not all risks of accident or injury are to be eliminated, but only those that are “unreasonable,” 15 U.S.C. § 1391(1); and safety standards cannot be imposed unless they are “practicable,” id. at § 1392(a). This qualifying language was added to ensure that NHTSA would “consider reasonableness of cost, feasibility and adequate lead time.” S.Rep. No. 1301, 89th Cong., 2d Sess. 6 (1966), 1966 U.S.Code Cong. & Ad.News 2709. Moreover, the Safety Act expressly authorizes NHTSA to “amend or revoke any Federal motor vehicle safety standard established under this section.” 15 U.S.C. § 1392(e). Bumper systems have been thought to be related to safety in four ways: (1) they may protect vehicle equipment, such as lighting and braking systems, that directly affect motor vehicle safety; (2) they may facilitate or hinder absorption of the destructive energy created by an automobile collision, i.e., in the patois of this case they may contribute to or impede “crash-energy management”; (3) they may contribute to pedestrian safety; and (4) if not of standard height, they may produce bumper interlock, thereby creating unsafe traffic conditions. The last two considerations are not at issue here — pedestrian protection because petitioners do not challenge the agency’s conclusion that “there is no evidence that the change [to 2.5 mph bumpers] will aggravate risks to pedestrians,” FRIA at IV-15, and bumper interlock because the new rule continues to require uniform bumper height. A. Protection of Vehicle Safety Equipment The agency concluded that “reduction of the 5/5 bumper standard would not have any significant effect on safety.” 47 Fed.Reg. 56,643. In arriving at this con-elusion with regard to the protection afforded by bumper systems to automobile safety equipment, NHTSA relied upon the Tri-Level Study of the Causes of Traffic Accidents (“Tri-Level Study”), published in 1979 by the Indiana University Institute for Research in Public Safety. See TriLevel Study of the Causes of Traffic Accidents: Executive Summary (May 1979) (“Executive Summary”), J.A. 2741-2809. That study, which performed on-site examinations of 2,258 accidents, estimated that only 0.44 to 0.48 percent of accidents, estimated that only 0.44 to 0.48 percent of accidents were certainly caused, and 0.93 to 1.67 percent were either certainly or probably caused, by malfunctioning safety equipment potentially within the zone of protection of a bumper system, vehicle lights and signals being the safety system most frequently implicated. FRIA at IV-12. Petitioners claim that even these low figures constitute an enormous toll in lives and injuries when it is considered that “50,-000 people a year have died in traffic accidents in recent years, and millions of others are injured.” Brief for Petitioner CFAS at 18. The agency’s reasoning on this issue shows petitioners to be wide of the mark. The 0.44 to 0.48 (or 0.93 to 1.67) percent figure is merely the point of departure, and not the conclusion, regarding the indirect relationship between bumpers and safety. The Tri-Level Study did not indicate what fraction of the malfunctioning safety equipment was caused by normal wear and tear (e.g., burned-out lightbulbs) rather than an earlier accident, but did indicate that older vehicles were disproportionately involved in accidents resulting from mechanical problems. FRIA at IV-13. Thus, with regard to light and signal failure, the major source of accidents attributable to systems that could conceivably be protected by bumpers, the agency quite reasonably concluded that “natural wearing out of bulbs [is] ... the most typical cause[ ] of inoperation.” Id. Even the remaining fraction, moreover, does not represent the safety differential between the two bumper systems. What matters for that comparison is the much smaller percentage of accidents caused by a damaged piece, of equipment that would have been protected by the 5.0 mph system but not by the 2.5 mph system. Many of the equipment-damaging accidents involved in the Tri-Level Study may have occurred at higher speeds, where neither the 5.0 mph system nor the 2.5 mph system would have protected the safety components. And even in accidents occurring between the 2.5 mph and 5.0 mph range, it is not the case that the 2.5 mph bumper never protects the safety component and the 5.0 mph bumper always does so. Finally, not all accidents involving failed safety systems result in deaths or injuries. Indeed, the Tri-Level Study noted that 70 to 80 percent of the accidents in its on-site sample resulted only in property damage, a figure which “corresponds closely to national figures for reported accidents.” Executive Summary at A-2, J.A. 2802. Since, as the agency the unreasonableness of risk from a safety standpoint depends, not simply on the number of accidents, but also on “the contribution of such accident[s] to injuries and deaths,” FRIA at IV-14, the significance of any difference between the accident-reduction potentials of the 2.5 mph and 5.0 mph bumper systems must be further discounted to a substantial degree. We cannot contradict the agency’s conclusion that the difference between the two systems resting upon the fraction of the 1 percent of accidents caused by all theoretically protectable safety defects, further discounted to an obviously massive but not precisely established extent by the above-described factors, was not “measurable” and did not constitute a significant safety consideration. 47 Fed.Reg. 21,827; 47 Fed.Reg. 56,643. The Safety Act does not require NHTSA to establish safety standards with an eye toward any conceivable safety hazard, no matter how insignificant; rather, the Act is directed at “unreasonable” risks. 15 U.S.C. § 1391(1). NHTSA could properly conclude that the level of risk here does not rise to such a level. Cf. Industrial Union Dep't, AFL-CIO v. American Petroleum Institute, 448 U.S. 607, 655, 100 S.Ct. 2844, 2870, 65 L.Ed.2d 1010 (1980) (Occupational Safety and Health Administration has “an obligation to find that a significant risk is present before it can characterize a place of employment as ‘unsafe’ ”) (emphasis added) (plurality opinion). Petitioners charge that NHTSA’s safety conclusions conflict with the only “real world” data comparing the effects on safety of the 5.0 mph and 2.5 mph bumpers. These consist of State Farm statistics, submitted to the agency as a comment during the rulemaking proceedings, analyzing its own claims experience from 1973 to 1980. According to petitioners, the statistics demonstrate that cars equipped with 2.5 mph bumpers experience far more damage to safety-related parts than cars equipped with 5.0 mph bumpers. They reach that conclusion by comparing State Farm’s claims experience for MY 1973 cars (equipped with rear bumpers meeting the Standard 215 2.5 mph requirement) with its claims experience for MY 1974 cars (equipped with the Standard 215 5.0 mph bumper systems). The comparison shows that the incidence of damage to safety items protected by rear bumper systems (gas tank, tail lamp, and trunk lid) was significantly and, for the most part, uniformly lower for the MY 1974 cars. For example, the relative incidence of damage to tail lamps, a safety item thought by NHTSA to be particularly important, is given in Table I. TABLE I Number of Safety Items Damaged Per 100 Estimates Written 2.5 m.p.h. 5 m.p.h. Percentage rear bumper rear bumper Improvement of standard standard 5 m.p.h. over Component MY 1973 MY 1974 2[.5] m.p.h. Tail Lamp Subcompact 10.34 5.83 44% Compact 8.19 7.31 11% Intermediate 8.64 6.25 28% Full Size 7.49 8.48 -13% Brief for Insurance Company Petitioners at 20 (footnotes omitted). NHTSA rejected the State Farm data because the cars examined in the study were manufactured to comply with the Standard 215 requirements, which for MY 1974 did, but for MY 1973 did not, include the pendulum test responsible for standardizing bumper heights. The agency therefore concluded that it would be misleading to treat the MY 1973 bumper system as equivalent to the 2.5 mph, uniform height system that NHTSA was considering' in 1981. FRIA at IV-9. Petitioners assert that this overestimates the importance of uniform bumper heights. They point out that in 1974, the year in which State Farm’s data evidences a dramatic improvement in MY 1974 rear protection, most cars on the road which the MY 1974 vehicles encountered were not subject to the height-standardizing pendulum test. Thus, they claim that while the State Farm figures might slightly overstate the contribution of the higher test impact speed, they still demonstrate the contribution to be significant. In this argument, we do well to be guided by the Supreme Court’s most recent instruction that “it is within the agency’s discretion to pass upon the generalizability of ... field studies.” State Farm, 103 S.Ct. at 2872. NHTSA did not conclude that there was no difference between the bumper systems in the protection they afford to safety systems. It merely concluded that “State Farm’s review ... falls short of demonstrating that [the difference] is noticeable.” FRIA at IV-9. State Farm made no effort to coriect for the effect of bumper height standardization or to demonstrate that, with or without such correction, the differences in claims experience were statistically significant. With regard to its comparison of MY 1973 and MY 1974 statistics in particular, there was no basis for assurance that the apparent differences were attributable to anything other than chance. We will not reverse the agency’s considered rejection of these data. They in any case relate, we should note, only to a single factor underlying the agency’s conclusion of no significant safety effect (viz., the difference between the 5.0 mph standard and the 2.5 mph standard in the protection of safety equipment) and in no way affect other factors, notably the agency’s estimation that safety equipment damaged in prior accidents is not a substantial cause of automobile accidents in the first place. We do not know what to make of the dissent’s complaint that the agency “formulated no criteria for defining what an insignificant safety risk might be,” Dissent at 1381. The assertion that this failure represents “the essence of unreasoned and unsupported decisionmaking,” id,., is refuted by the passage from our own opinion quoted in the same portion of the dissent, wherein we say that “[w]e use the term ‘significant’ to indicate that there must be a non-cfe minimus [sic] number of failures.” United States v. General Motors Corp., 518 F.2d 420, 438 n. 84 (D.C.Cir. 1975). Surely substituting the Latín phrase “de minimis” for the English adjective “insignificant” does not constitute “formulating a criterion”; nor are we aware of any case that imposes such a requirement — whereby an agency would, presumably, announce a rule that more than 1,000 injuries is significant and less than 1,000 is not. To the contrary, the agencies’ ability to identify and pronounce significant risk in case-by-ease fashion is made clear by the totality of the Supreme Court’s analysis which the dissent chooses to quote in misleading part: [T]he requirement that a “significant” risk be identified is not a mathematical straitjacket. It is the Agency’s responsibility to determine, in the first instance, what it considers to be a “significant” risk. Some risks are plainly acceptable and others are plainly unacceptable. If, for example, the odds are one in a billion that a person will die from cancer by taking a drink of chlorinated water, the risk clearly could not be considered significant. On the other hand, if the odds are one in a thousand that regular inhalation of gasoline vapors that are 2% benzene will be fatal, a reasonable person might well consider the risk significant and take appropriate steps to decrease or eliminate it. Although the Agency has no duty to calculate the exact probability of harm, it does have an obligation to find that a significant risk is present before it can characterize a place of employment as “unsafe.” Industrial Union Dep’t, AFL-CIO v. American Petroleum Institute, 448 U.S. 607, 655, 100 S.Ct. 2844, 2870-2871, 65 L.Ed.2d 1010 (1980) (footnote omitted). As we said in General Motors, immediately following the “de minimis” remark which the dissent elevates to a criterion: “The question whether a ‘significant’ number of failures have \sic ] taken place must be answered in terms of the facts and circumstances of each particular case.” 518 F.2d at 438 n. 84. In any case, even if NHTSA had established the sort of numerical “criterion” of significance the dissent evidently demands, it would have been of no assistance in resolution of the present issue, since it was impossible to quantify the present risk for purposes of comparison with any such criterion. As the FRIA made entirely clear, no reliable data were available from which hard numbers or even percentages could be derived, either with regard to the differential in protection of safety systems afforded by 5.0 mph and 2.5 mph bumpers, FRIA at IV-7 to IV-9, or with, regard to the number of accidents attributable to safety system malfunctions caused by prior accidents, id. at IY-13. Perhaps data regarding the former could somehow have been developed from extensive crash-test experiments with 2.5 mph-bumper cars — which would have required, for American cars that constitute the bulk of the United States market, specially produced bumpers, and would also have required some means of accounting for the fact that not all safety systems, and in particular lights, are identically situated and protected on all car models. But it is utterly impossible to imagine how the second category of nonexistent data could have been developed, i.e., the number of accidents caused not merely by faulty safety systems but by safety systems damaged in an earlier accident — and in an earlier accident at particular speeds where the variable protectiveness of the two bumpers would make a difference. As acknowledged by State Farm’s Vice President of Research, the relationship between safety and damage to safety-related vehicle components is probably the most difficult possible thing to study and I agree with one of the former speakers that you would have to put some judgment on here____ I’m not optimistic we’ll be able to provide a definitive answer, but fortunately I think that burden rests with you more than us. Department of Transportation, NHTSA, Public Hearing on Bumper Standard Number 581 at 77 (Oct. 22, 1981) (Statement of Wayne W. Sorenson), J.A. 1282. The agency did “put some judgment on,” and State Farm’s Vice President was wrong only in his assumption regarding the burden of proof. The burden was on the agency, to be sure, to justify the change from the status quo, see State Farm, 103 S.Ct. at 2866. But that justification need not consist of affirmative demonstration that the status quo is wrong; it may also consist of demonstration, on the basis of careful study, that there is no cause to believe that the status quo is right, so that the existing rule has no rational basis to support it. The agency did precisely that here, stating that “the 5.0-mph safety criteria of Part 581 have been determined to be unsupported,” 47 Fed.Reg. 21,830. That finding was sufficient to support the agency’s action, and was itself amply supported by the record. The position in effect espoused by the dissent — that obviously small but precisely unknowable risks of injury must be presumed to be significant once they have been treated as such by the agency — is a formula for rendering an erroneously adopted standard, itself unsupported, by any scientific analysis, essentially irremediable. B. Crash Energy Management “Crash energy management” involves the absorption of collision energy by a vehicle’s structural components, lessening the risk of injury to occupants. On the basis of the record evidence, NHTSA concluded that (1) the total contribution of 5.0 mph bumper systems to crash energy management in collisions at speeds likely to cause death or serious injury is small, (2) the difference between 5.0 mph and 2.5 mph systems is small, and (3) it may be that 5.0 mph systems are less protective as a theoretical matter. 47 Fed.Reg. 56,643. The agency’s analysis finds ample evi-dentiary support in the record. Automobile manufacturers submitted estimates of the bumper system’s percentage contribution to total crash energy management in 30 mph barrier-impact tests. Those estimates were uniformly low (less than 5 percent for most commenters), and indicated little or no difference between 5.0 mph and 2.5 mph systems. FRIA at IV-4 to -5. Some manufacturers noted that the use of relatively rigid bumper energy absorbers, required to comply with the 5.0 mph standard, could adversely affect crash energy management, Comments of Volkswagen of America, Inc., Docket No. 73-19-N.27-058 at II (Nov. 25,1981), J.A. 1722, and that “in certain cases it may be possible to have a better optimized [sic ] and more controllable occupant protection system for high speed protection if one didn’t have to take into account the requirements of low speed protection,” Comments of Volvo of America Corp., Docket No. 73-19-N.27-078 at 3 (Nov. 30, 1981), J.A. 1655 (quoted in FRIA at IV-5 to -6). See also Comments of General Motors Corp., Docket No. 73-19-N.27-053 at A-5 (Nov. 25, 1981), J.A. 1773; Comments of Ford Motor Co., Docket No. 73-19-N.27-047 at 24 (Nov. 25, 1981), J.A. 1835. Petitioners do not advance any data showing that there is better crash energy management with the 5.0 mph bumper system, but merely recite, without explanation, their own conclusion that “strong bumpers help to absorb force in automobile crashes; they thereby contribute to crash energy management by reducing the violence of impact forces reaching the occupants of the vehicle.” Brief for Insurance Company Petitioners at 10 (footnote omitted). This does not suffice to cast doubt upon the agency’s judgment, based upon informed estimates from knowledgeable sources. Petitioner CFAS reiterates the argument, advanced by the Insurance Institute for Highway Safety in its petition for reconsideration, that NHTSA’s crashworthiness conclusions were based on 30 mph tests, but “many severe injuries occur in speeds as low as 10-15 mph.” Brief for Petitioner CFAS at 19. NHTSA responded to this criticism, pointing out that it had considered the crash energy management issue at all speeds, “notwithstanding the fact that certain of the illustrations cited by the agency involved 30 mph crashes,” 47 Fed.Reg. 56,643, and explaining its reasoning in considerable detail. Id. NHTSA has more than adequately articulated its consideration of the relevant factors, and we will not “substitute [our] judgment [on the issue] for that of the agency.” Citizens to Preserve Overton Park, Inc. v. Volpe, 401 U.S. at 416, 91 S.Ct. at 824. III. Cost Savings Act Analysis Once NHTSA has discharged its obligations under the Safety Act, its remaining statutory duty is to apply § 102(b)(1) of the Cost Savings Act, 15 U.S.C. § 1912(b)(1) (1982). The Insurance Company Petitioners, joined as amici curiae by former Congressmen Eckhardt and Moss, the principal sponsors of the Cost Savings Act, advance the threshold argument that the legislative history of the Cost Savings Act limits NHTSA’s discretion in fashioning a no-damage bumper standard. They marshall passages from the committee reports and statements by individual legislators assert-edly establishing the congressional “intent” that the no-damage standard be set at a level of no less than 5.0 mph. The following statements are representative: [T]he chairman and the other committee members have so ably demonstrated it is their belief, and I am sure it is well founded, that the Department of Transportation is going to promulgate standards eventually which will be set in excess of 5 miles per hour. 118 Cong.Rec. 18,229 (1972) (Statement of Rep. Danielson). It seems inconceivable to me that after already enacting a 5-mile standard with respect to safety that the Federal Government would establish a 2.5 mile standard with respect to repairability. 118 Cong.Rec. 18,229-30 (1972) (Statement of Rep. Eckhardt). [The Secretary] will not select a 2-mile-per-hour standard when there is a 5-mile-per-hour standard already in the law. 118 Cong.Rec. 18,230 (1972) (Statement of Rep. Harvey). [I]n the entire span of discussion in the committee there was a consensus that we were talking of a 5 mile an hour standard. We did not feel it desirable to freeze that into the statute, but rather [wanted] to make clear our intent and acquaint the Secretary with that intent. 118 Cong.Rec. 18,222 (1972) (Statement of Rep. Moss). This argument is luminescently invalid. Congress does not act, and cannot legally bind, through its intent and expectation as such, whether individually or collectively expressed, but only through the laws that it enacts. Thus, the only intent or expectation of Congress pertinent to our task is its intent regarding the meaning of statutory language or its expectation regarding the manner in which that language will be interpreted. It is, as the Supreme Court has often said, a bad enough idea to consult extra-statutory expressions of even that sort of intent or expectation when the law is clear on its face. See, e.g., Gulf Oil Corp. v. Copp Paving Co., 419 U.S. 186, 200-01, 95 S.Ct. 392, 401-402, 42 L.Ed.2d 378 (1974). But it is absurd — indeed, lawless — to give legal effect to such expressions that purport to relate, not to the meaning of the statute, but to the manner in which a legally unconstrained agent of the Executive will behave under it. All of the statements pressed upon us here are in the latter category. Then-Representative Moss told us all we need to know when he said, in the excerpt quoted earlier, that “we did not feel it desirable to freeze [the 5.0 mph standard] into the statute.” It should not be thought, of course, that the expression of such noninterpretive intent and expectation by Members of Congress and their committees is a pointless exercise. As long as those particular Members and committees hold sway, the Executive agent, however unconstrained he may be in law, will disregard their intent and upset their expectation at his budgetary and programmatic peril. (As noted earlier, NHTSA Administrator Claybrook was influenced by this “Congressional sentiment” in favor of the 5.0 mph standard as late as 1978. See page 1340, supra.) But we decline to convert this transient political incentive into a permanent legal compulsion, enabling the past thinking of prior legislators to control current social choices through the dead hand of legislative history unrelated to statutory text. Viewing NHTSA’s task, therefore, as that set forth in the text of the Cost Savings Act, unmodified by unlegislated congressional expectations, it is an exercise quite familiar to economists and public policy analysts. The agency was to identify the costs and benefits of alternative standards, measure them, and select the standard which displays the greatest net benefit. This is more easily said than done, since — given the fact that the alternatives to the Part 581 standard had never been in existence — the process was as much one of prediction as of analysis. The agency proceeded in full recognition of this difficulty, seeking to minimize the number of imponderables and acknowledging the residual imprecision by expressing its estimates in the form of high-low numerical ranges, rather than with the misleading certitude of a single figure. Our discussion below makes no attempt to describe with any completeness the full process of analysis, which is summarized in the agency’s 18-page Federal Register statement of the basis and purpose for the rule, Bumper Standard, 47 Fed.Reg. 21,820 (1982), and is set forth more fully in its 263-page FRIA. The reader should be aware, however, that the items of the analysis under challenge here form a very small (though not necessarily unimportant) part of a much more lengthy and complex whole. A. Bumper System Costs NHTSA identified three different sources of cost differences among alternative bumper systems. First, more protective bumper systems, which require additional components such as energy-absorbing pistons, increase the weight of the car and thus the cost of the fuel that it will consume over its lifetime. This increase occurs not only in the “primary weight” of the bumper system, but also in the “secondary weight” of additional body and frame structural reinforcement needed to accommodate the increased primary weight. Second, more protective bumper systems cost more to manufacture and install — costs which can likewise be divided into “primary” (attributable to the bumper itself) and “secondary” (attributable to related structural reinforcement). Third, and finally, that portion of the increased (primary and secondary) bumper system price which is financed imposes additional finance costs on the consumer. See FRIA at VII-12. NHTSA solicited from the automobile industry confidential production cost data and weight estimates for the alternative bumper systems. Based upon those submissions, and taking into account its own assessment of expected changes in fleet composition over the next decade, id. at VII-2 to -4, it estimated that bumpers satisfying the adopted standard would result in production cost savings (over the 5.0 mph system) of $18-$35 in primary production cost, id. at VII-19 to -25, and $6-$20 in secondary production cost, id. at VII-35 to -36. It estimated that the new standard would eliminate 15-33 pounds in primary weight, id. at VII-28, and 11-33 pounds in secondary weight, id. at VII-35 to -36. Using gasoline price forecasts supplied by the Department of Energy, id. at VII-43, its own data on lifetime mileage distribution, id. at VII-44, and an estimated fuel consumption/weight ratio of 1 gallon: 1 pound over the lifetime of the car, id. at VII-40 to -42, NHTSA estimated that those weight reductions would produce fuel savings over the life of the car (reduced to present value using the Office of Management and Budget’s 10 percent discount rate and standard discounting methodology, id. at III — 66) of $28 — $70, id. at VII-46. Finally, the agency estimated a $2-$6 per car reduction in finance charges to consumers. Id. at VII-47 to -53, -60 to -61. The sum of these production cost, fuel cost and finance cost savings produced a cost advantage of $54 — $131 for the 2.5 mph system. Petitioners do not challenge NHTSA’s overall methodology. Moreover, although there is admitted uncertainty with respect to nearly every variable, they take serious issue only with NHTSA’s primary and secondary weight estimates. 1. Primary Weight Insurance Company Petitioners, repeating the argument made by the Insurance Institute for Highway Safety in its petition for reconsideration, claim that, in arriving at a 15-33 pound range of primary weight reduction on the basis of estimates by automobile manufacturers, NHTSA “simply excluded from its analysis the estimates of Chrysler, Volkswagen, and Mitsubishi that weight savings would be significantly less than the agency’s 15-pound minimum.” Brief for Insurance Company Petitioners at 37. It is true enough that NHTSA excluded those submissions, but not without sound reasons. As noted in some detail in the agency’s response to petitions for reconsideration, Chrysler’s 10-pound estimate was for a Phase II bumper system. Compliance with a 2.512.5 Phase I standard could be expected to yield additional weight savings of 1-2 pounds. 47 Fed.Reg. 56,-645. More importantly, Chrysler noted that its estimate was a minimum figure. Id. Similarly, Volkswagen indicated that its 8-pound savings estimate referred only to the savings from one particular design change; total savings were expected to be higher. Id. The exclusion of Mitsubishi’s low estimate of 11-13 pounds was explained by the fact that Mitsubishi sales were relatively insignificant. Since the savings estimates were not weighted to reflect the United States sales volumes of particular cars or manufacturers, the agency deleted low and high extreme estimates submitted by manufacturers with very small sales in the United States. Thus, it excluded Mitsubishi’s low estimate of 11-13 pounds but also Volvo’s high estimate of 39 pounds. Id. Petitioner CFAS charges that, in estimating weight reductions, NHTSA “failed to consider” the effect of the fuel economy test procedures of the Environmental Protection Agency (“EPA”). Brief for Petitioner CFAS at 30. The argument is that consumer estimation of new-car fuel economy is based upon EPA mileage ratings; and that the EPA classifies vehicles for purposes of its fuel efficiency tests according to weight classes. See 40 C.F.R. §§ 86.129-80, 600.111-80 (1983). Because vehicles in the same class, even if of different weight, will be tested as though they weigh the same, manufacturers lack the incentive to reduce a vehicle’s weight unless they can achieve a weight reduction large enough to move the vehicle into the next lower class. Since the weight savings achievable under the new bumper system are small (15-33 pounds) relative to the weight class increments used by EPA (250 pounds for cars over two tons and 125 pounds for cars under two tons), manufacturers, according to CFAS, will save money by using heavier, cheaper materials and forgo the potential weight reductions. It is patently false to assert that the agency “failed to consider” this point. See 47 Fed.Reg. 56,646. Petitioner’s quarrel is with the resolution of it — and even there we think the agency not only acted within the bounds of reason but clearly has the better of the argument. It is fallacious, to begin with, to assume that consumer expectation of fuel economy is the manufacturer’s only incentive to reduce weight. The agency observed that there are quite independent incentives, such as the fact that many weight-reducing changes (e.g., the removal of energy-absorbing pistons) reduce manufacturer costs. Id. Even assuming, however, that consumer expectation of fuel economy is the only incentive, that that expectation is based entirely upon EPA estimates, and that the manner of computing EPA estimates, will remain unchanged (none of which assumptions we consider the agency was obliged to indulge); the petitioner’s argument would still be flawed. The lumpiness of current EPA weight classes does not necessarily reduce the incentive, on average, to take advantage of small weight reductions. In designing cars, there are many small potential weight savings. The 15 to 33-pound savings offered by the new bumper standard — alone or in combination with secondary weight reductions that it makes possible — may be just enough in some cases to make it worth the manufacturer’s while to implement a whole host of changes that together would be enough to move the vehicle into the next weight category. When this occurs, a far larger weight reduction is achieved than would be made possible by the bumper change itself. Thus, one might reasonably conclude that, on average, the bumper weight reduction potential would be given its full effect. In addition, the agency noted that it had calculated its weight savings range from what manufacturers had told it they would in fact do, not from what they thought was theoretically possible. Id. Nothing in the record calls those intentions about future weight reductions into question. Moreover, the agency used its own studies as a check on manufacturer estimates of weight savings, id., concluding that, if anything, the 33-pound upper-range estimate was too low. 47 Fed.Reg. 21,832. We find no basis for overturning the agency’s judgment on this point. 2. Secondary Weight NHTSA found that each pound of reduced bumper weight would be accompanied by further reductions, ranging from 0.7-1.0 pounds, in the weight of basic automobile structures. Petitioners contend that this conclusion finds no support in the record, because no manufacturer promised in its submission to make secondary reductions. This is an unduly narrow view of the kind of evidence that the agency may properly rely on. NHTSA’s prediction of secondary weight reductions was based upon “generally recognized and accepted design theory, the incentives facing manufacturers and manufacturer comments.” 47 Fed.Reg. 56,647. (As noted earlier, the premise that there are incentives to reduce weight is reasonable.) It would be a novel legal proposition to hold that such commonplace agency prediction of market behavior is arbitrary and capricious unless supported by manufacturer commitments in rulemaking submissions. Since NHTSA’s reasoning was made part of the record and has a rational basis, it is sufficient to support the agency’s conclusion. See National Tour Brokers Ass’n v. ICC, 671 F.2d 528, 533 (D.C.Cir.1982); Ryder Truck Lines, Inc. v. United States, 716 F.2d 1369, 1385 (11th Cir.1983), cert. denied, — U.S. -, 104 S.Ct. 1707, 80 L.Ed.2d 181 (1984). Petitioners next argue that even if NHTSA properly concluded that reduced bumper weights would lead to some secondary weight reductions, these reductions would be realizable only over a lengthy period of time, as manufacturers gradually redesign their fleets to take advantage of potential weight savings. Thus, NHTSA’s cost-benefit analysis should have allocated and discounted the projected secondary savings over several years (or longer). Instead, NHTSA treated them as available in the first year of implementation of the new standard, FRIA VII-35 to -38, XI-19, thus greatly overstating the present value of the benefits of the 2.5 mph bumper standard. In its denial of petitions for reconsideration, NHTSA acknowledged that the anticipated secondary weight savings would not be fully available right away, but noted that the same could be said of primary weight reductions, since the new bumper standard would not result in the immediate disappearance of all 5.0 mph bumpers from the market. “Production requirements and market pressure will constrain any move immediately to 2.5 mph bumpers only.” 47 Fed.Reg. 56,647. Though inelegantly stated, the agency’s position is understandable if not clear: No objection has been raised to the agency’s unreal assumption (for purposes of its cost-benefit analysis) that it is dealing with a universe of vehicles entirely equipped, here and now, with the 2.5 mph bumpers that the new standard would permit — for while this assumption distorts the magnitude of the costs and benefits that the new standard will produce (in the first few years at least), it does not distort what is important here, the relative net benefit of one standard over the other. The same is true of structural design changes accompanying the new bumper, and there is no need to treat them differently. See Small Refiner Lead Phase-Down Task Force v. EPA, 705 F.2d 506, 535 (D.C.Cir.1983) (“we must defer to the agency’s decision on how to balance the cost and complexity of a more elaborate model against the oversimplification of a simpler model”). In calling for discounting, petitioners confuse the physical state of affairs that the new standard will produce (including both the new lighter bumpers and the new lighter supporting structures), which can properly be assumed to exist at once, with the costs and benefits engendered by that state of affairs, some of which (i.e., accident costs, finance costs and fuel savings) will only accrue over time and must be discounted. The agency did discount the latter — for secondary weight reduction as for primary weight reduction — at a rate of 10 percent. FRIA at III — 70. Finally, petitioners claim inconsistency between NHTSA’s treating secondary weight reductions as presently implemented and its failure to consider the effects of future design improvements that might reduce the weight of 5.0 mph bumpers. The two are not inconsistent. The latter involves technological speculation while the former is a known consequence of the new standard which, like the new bumpers themselves, is immediately achievable as far as technology is concerned. Moreover, NHTSA could reasonably assume that any such technological improvements would also reduce the weight of 2.5 mph bumpers and thus not affect the relative cost-benefit analysis. 47 Fed.Reg. 21,834; 47 Fed.Reg. 56,646. B. Bumper System Benefits By reducing the incidence and the severity of damage-causing accidents, protective bumper systems save repair costs and insurance costs, and by reducing the incidence they save intangible costs of personal delay. Petitioners raise a number of challenges to the agency’s analysis of the relative benefits of alternative bumper systems. 1. Calculation of Effectiveness in Limiting Damage The most satisfactory method of comparing the ability of different bumper systems to prevent or reduce damage is to examine systems that have actually been produced and used. Although there was a wealth of real-world data for comparing the Part 581 5.0 mph system with unregulated bumpers, none existed for the primary alternative — a 2.5 mph standard with uniform bumper heights. NHTSA considered several methodologies for estimating the effectiveness of the 2.5 mph alternative. One was to use real-life data on the performance of MY 1973 cars equipped with 2.5 mph rear bumpers. That was rejected, primarily because the MY 1973 rear bumpers were not subject to the height-standardizing pendulum test, and thus were not comparable to the proposed 2.5 mph bumpers. 47 Fed.Reg. 21,-831; FRIA at III — 3 to -4. Also rejected was a method proposed in the comments of insurance companies — reliance on laboratory condition crash test results comparing the performance of three pairs of vehicles purportedly equipped with 5.0 mph and 2.5 mph bumper systems — because there was no evidence that the bumpers used in the tests had been tested or designed in accordance with the relevant bumper standards. 47 Fed.Reg. 21,831-32; 47 Fed.Reg. 56,-647-48; FRIA at III — 5 to -6. There were also questions about the generalizability of the test data. Id. at III — T to -8. The agency decided instead to use the methodology employed in its January 1979 and June 1979 assessments of the bumper standard, see Department of Transportation, NHTSA, Analysis of the Bumper Standard (Jan. - 26, 1979), J.A. 1245-52 (“Preliminary Assessment”); Department of Transportation, NHTSA, Final Assessment of the Bumper Standard (June 1979), J.A. 979-1182 (“Final Assessment”), which relied on engineering estimates of the theoretical relative effectiveness of different bumper systems. The data used in that exercise were compiled by NHTSA in a convenient table form in the FRIA, VI-3, reproduced here as Table III. First, the agency adopted the estimates from the 1979 Final Assessment of the percentage of damage-producing accidents that occur over the lifetime of a car at each of several speed intervals (Column 1 of the Table), FRIA at VI-2, providing separate figures for accidents resulting in damage subsequently repaired (Column 2) and damage left unrepaired (Column 4). (These figures were based on Ford Motor Company studies of relative bumper effectiveness. Final Assessment at 24, J.A. 1005.) Second, for each speed interval, NHTSA estimated separately the dollar value of repaired and unrepaired damage per accident for a car equipped with unregulated (baseline) bumpers. These data, also derived from the 1979 Final Assessment, were based on an assumed relationship between repair cost and impact speed, id. at 25-26, J.A. 1006-07, and closely approximated the results of calculations based on studies relating damage to the angle of impact, id. at B-l to -7, J.A. 1099-1105. The figures are provided in Column 3 of the Table for repaired and Column 5 for unrepaired damage. Third, by multiplying the values in Columns 2 and 3, and 4 and 5, respectively, and summing the resultant products, NHTSA obtained values of the total lifetime damage per accident for each speed category (Column 6). NHTSA then adopted the 1979 estimates of the relative effectiveness of the 5.0 mph and 2.5 mph bumpers as against unregulated bumpers at each speed interval. 47 Fed.Reg. 56,648-49; FRIA at VI-2. Those relative effectiveness values are defined as the percentage of damage preve