Center for Auto Safety v. Raymond A. Peck, Jr., Administrator of the National Highway Traffic Safety Administration, Motor Vehicle Manufacturers Association of the United States, Inc., Automobile Importers of America, Inc., Intervenors. State Farm Mutual Automobile Insurance Company v. Department of Transportation, Elizabeth Dole, as Secretary, National Highway Traffic Safety Administration, and Raymond A. Peck, Jr., as Administrator, Motor Vehicle Manufacturers Association of the United States, Inc., Automobile Importers of America, Inc., Intervenors. Allstate Insurance Companies v. Department of Transportation, Elizabeth Dole, as Secretary, National Highway Traffic Safety Administration, and Raymond A. Peck, Jr., as Administrator

751 F.2d 1336 243 U.S.App.D.C. 117 CENTER FOR AUTO SAFETY, Petitioner, v. Raymond A. PECK, Jr., Administrator of the National Highway Traffic Safety Administration, Respondent, Motor Vehicle Manufacturers Association of the United States, Inc., Automobile Importers of America, Inc., Intervenors. STATE FARM MUTUAL AUTOMOBILE INSURANCE COMPANY, Petitioner, v. DEPARTMENT OF TRANSPORTATION, Elizabeth Dole, As Secretary, National Highway Traffic Safety Administration, and Raymond A. Peck, Jr., As Administrator, Respondents, Motor Vehicle Manufacturers Association of the United States, Inc., Automobile Importers of America, Inc., Intervenors. ALLSTATE INSURANCE COMPANIES, Petitioner, v. DEPARTMENT OF TRANSPORTATION, Elizabeth Dole, As Secretary, National Highway Traffic Safety Administration, and Raymond A. Peck, Jr., As Administrator, Respondents. Nos. 82-1782, 82-1783 and 83-1164. United States Court of Appeals, District of Columbia Circuit. Argued Dec. 5, 1983. Decided Jan. 8, 1985. [243 U.S.App.D.C. 118] Petitions for Review of an Order of the National Highway Traffic Safety Administration. James F. Fitzpatrick, Washington, D.C., with whom John M. Quinn, Washington, D.C., was on the brief, for State Farm Mut. Auto. Ins. Co. and Allstate Ins. Companies, petitioners in Nos. 82-1783 and 83-1164. Peter R. Maier and Donald P. McHugh, Washington, D.C., also entered appearances for petitioners in Nos. 82-1783 and 83-1164. Katherine I. Hall, Washington, D.C., was on the brief for Center for Auto Safety, petitioner in No. 82-1782. Michael F. Hertz, Atty., Dept. of Justice, Washington, D.C., with whom J. Paul McGrath, Asst. Atty. Gen., Dept. of Justice, Frank Berndt, Chief Counsel, Stephen P. Wood, Asst. Chief Counsel, David W. [243 U.S.App.D.C. 119] Allen, Asst. Chief Counsel, Eileen T. Leahy and J. Edward Glancy, Attys., National Highway Traffic Safety Administration, Washington, D.C., were on the brief, for respondents in Nos. 82-1782, 82-1783 and 83-1164. William H. Allen, Richard D. Copaken, Oscar M. Garibaldi, Washington, D.C., and William H. Crabtree, Detroit, Mich., were on the brief for Motor Vehicle Manufacturers Association of the United States, Inc., et al., intervenors in Nos. 82-1782 and 82-1783. Harris Weinstein, Washington, D.C., and Edward P. Good, Detroit, Mich., also entered appearances for Motor Vehicle Manufacturers Association of the United States, Inc., et al. Milton D. Andrews and Lance E. Tunick, Washington, D.C., were on the brief for Automobile Importers of America, Inc., intervenor in Nos. 82-1782 and 82-1783. Nathan Lewin and Anne Shere Wallwork, Washington, D.C., were on the brief for American Insurance Association, amicus curiae, urging that National Highway Traffic Safety Administration's Final Rule amending the Bumper Standard should be vacated in Nos. 82-1782 and 82-1783. Mary Todd Foldes, Washington, D.C., was on the brief for United States Congressmen Moss and Eckhardt, amici curiae, urging reversal in Nos. 82-1782, 82-1783 and 83-1164. Charles A. Taylor, III, Washington, D.C., was on the brief for National Association of Independent Insurers, amicus curiae, urging reversal in Nos. 82-1782 and 82-1783. Robert E. Litan, Washington, D.C., was on the brief for National Association of Insurance Commissioners, amicus curiae, urging reversal in Nos. 82-1782, 82-1783 and 83-1164. Before WRIGHT, TAMM and SCALIA, Circuit Judges. Opinion for the Court filed by Circuit Judge SCALIA. Dissenting opinion filed by Circuit Judge J. SKELLY WRIGHT. SCALIA, Circuit Judge: — 1 — 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 — 2 — 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 impacts1 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 [243 U.S.App.D.C. 120] impacts.2 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. — 3 — 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 this 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 systems 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: — 4 — [M]any accidents of the parking lot variety 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. — 5 — 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. — 6 — 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 Secretary 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 reduction 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: — 7 — (A) the cost of implementing the standard and the benefits attainable as the result of implementation of the standard; — 8 — (B) the effect of implementation of the standard on the cost of insurance and prospective legal fees and costs; — 9 — (C) savings in terms of consumer time and inconvenience; and — 10 — (D) considerations of health and safety, including emission standards. — 11 — 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). — 12 — 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 [243 U.S.App.D.C. 121] 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. — 13 — 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). — 14 — 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: — 15 — 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. — 16 — Letter from Joan Claybrook to Richard Copaken 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 [243 U.S.App.D.C. 122] 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, — 17 — 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 feasible the standard that is most cost beneficial to the public and the consumer, that is, where marginal benefits equal marginal costs. — 18 — 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. — 19 — 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. — 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 $11-$29 advantage over the 2.5 mph standard. Id. — 21 — 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 [243 U.S.App.D.C. 123] 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. — 22 — In July 1981, NHTSA released its Preliminary Regulatory Impact Analysis ("PRIA"), which analyzed five alternative bumper standards. PRIA at IV-2 to -3, J.A. 399-400. It then published a notice of proposed rulemaking soliciting comments on nine alternative bumper standards3 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. — 23 — 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 — 24 — 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)). — 25 — Petitioners assert that "because the agency has reversed a longstanding policy," this court must exercise a "more [243 U.S.App.D.C. 124] 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. — 26 — 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 — 27 — 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,4 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 standard, id. at § 1912(b)(2). — 28 — 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). — 29 — 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 [243 U.S.App.D.C. 125] new rule continues to require uniform bumper height. A. Protection of Vehicle Safety Equipment — 30 — The agency concluded that "reduction of the 5/5 bumper standard would not have any significant effect on safety." 47 Fed.Reg. 56,643.5 In arriving at this conclusion 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 Tri-Level Study of the Causes of Traffic Accidents: Executive Summary (May 1979) ("Executive Summary"), J.A. 2741-2809. [243 U.S.App.D.C. 126] 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. — 31 — 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 noted, 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. — 32 — 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). — 33 — Petitioners charge that NHTSA's safety conclusions conflict with the only "real world" data comparing the effects on safety [243 U.S.App.D.C. 127] 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 1980.6 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. — 34 — Brief for Insurance Company Petitioners at 20 (footnotes omitted). — 35 — 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. — 36 — 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 correct for the effect of bumper height standardization or [243 U.S.App.D.C. 128] 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.7 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.8 — 37 — 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-de minimus [sic ] number of failures." United States v. General Motors [243 U.S.App.D.C. 129] Corp., 518 F.2d 420, 438 n. 84 (D.C.Cir.1975). Surely substituting the Latin 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-case fashion is made clear by the totality of the Supreme Court's analysis which the dissent chooses to quote in misleading part: — 38 — [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." — 39 — 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. — 40 — 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 IV-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.9 As acknowledged by State [243 U.S.App.D.C. 130] Farm's Vice President of Research, the relationship between safety and damage to safety-related vehicle components — 41 — 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. — 42 — Department of Transportation, NHTSA, Public Hearing on Bumper Standard Number 581 at 77 (Oct. 22, 1981) (Statement of Wayne W. Sorenson), J.A. 1282. — 43 — 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 — 44 — "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. — 45 — The agency's analysis finds ample evidentiary 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 [243 U.S.App.D.C. 131] (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. — 46 — 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 824.10 III. COST SAVINGS ACT ANALYSIS — 47 — 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 assertedly 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: — 48 — [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. — 49 — 118 Cong.Rec. 18,229 (1972) (Statement of Rep. Danielson). — 50 — 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. — 51 — 118 Cong.Rec. 18,229-30 (1972) (Statement of Rep. Eckhardt).[243 U.S.App.D.C. 132] [The Secretary] will not select a 2-mile-per-hour standard when there is a 5-mile-per-hour standard already in the law. — 52 — 118 Cong.Rec. 18,230 (1972) (Statement of Rep. Harvey). — 53 — [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. — 54 — 118 Cong.Rec. 18,222 (1972) (Statement of Rep. Moss). — 55 — 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." — 56 — 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. — 57 — 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 — 58 — NHTSA identified three different sources of cost differences among alternative bumper systems. First, more protective bumper systems, which require additional [243 U.S.App.D.C. 133] 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).11 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. — 59 — 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. — 60 — Petitioners do not challenge NHTSA's overall methodology.12 Moreover, although [243 U.S.App.D.C. 134] there is admitted uncertainty with respect to nearly every variable, they take serious issue only with NHTSA's primary and secondary weight estimates.13 1. Primary Weight — 61 — 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 2.5/2.5 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. — 62 — 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. — 63 — 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, [243 U.S.App.D.C. 135] 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. — 64 — 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 — 65 — 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). — 66 — 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. — 67 — 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 [243 U.S.App.D.C. 136] 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. — 68 — 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 — 69 — By reducing the incidence and the severity of damage-causing accidents, protective bumper systems save repair costs and insurance costs,14 and by reducing the incidence they save intangible costs of personal delay.15 Petitioners raise a number of challenges to the agency's analysis of [243 U.S.App.D.C. 137] the relative benefits of alternative bumper systems. — 70 — 1. Calculation of Effectiveness in Limiting Damage — 71 — 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. — 72 — 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. 21,83 1-32; 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-7 to -8. — 73 — 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. — 74 — NOTE--Some parts of this form are wider than one screen. To view — 75 — material that exceeds the width of this screen, use the right arrow — 76 — key. To return to the original screen, use the left arrow key. — 77 — TABLE III Estimated Reduction in Lifetime Repair Costs Per Car Assuming Unrepaired Damage Valued at 50% and 75% of Full Cost to Repair 50% Value for Unrepaired Damage --------------------------- Repaired Unrepaired Damage Reduction in Damage Lifetime Repair Costs Per Total Accident Value Percent Value Lifeti- Effectiveness Compared of of of me in to Percent of Damage Lifetime Damage Damage Reducing Unregulated Speed Lifetime Per Number of Per Per Damage 4 Bumpers 3 Number (mph) of Accide- Accide- Accidents Accide- Accide- Accident 3 2.5 mph 5 mph 2.5 nts 1 nt 2 1 nt nt 2 mph ------------------------------------------------------------------------------------------------------ 0-3 14.8 $ 150 54.1 $ 75 $63 76% 85% $48 $54 3-5 12.1 225 3.8 113 32 29 68 9 21 5-7 6.6 480 0 240 32 0 40 0 13 7-10 4.4 850 0 425 37 0 6 0 2 10-15 3.6 1,750 0 875 63 0 0 0 0 15 0.5 3,700 0 1,850 20 0 0 0 0 -------- ------ Total $57 $90 2.5/5 mph System=63% (57/90) 75% Value For Unrepaired Damage --------------------------- 0-3 14.8 150 54.1 113 84 76 85 63 71 3-5 12.1 225 3.8 169 34 29 68 10 23 5-7 6.6 480 0 360 32 0 40 0 13 7-10 4.4 850 0 638 37 0 6 0 2 10-15 3.6 1,750 0 1,313 3 0 0 0 0 15 0.5 3,700 0 2,775 20 0 0 0 0 -------- ------ Total $73 $109 2.5/5 mph System=67% (73/109) 1. From page 24 of 1979 Assessment, original source is Ford data, finetuned by State Farm data.1/5 2. From page 28 of 1979 Assessment, for unregulated bumpers.1/5 3. Per accident--values need to be multiplied by the total number oflifetime damage producing accidents to attain lifetime dollar values of damage.1/5 4. From page 30 of 1979 Assessment.1/5 — 78 — [243 U.S.App.D.C. 138] 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 [243 U.S.App.D.C. 139] 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-1 to -7, J.A. 1099-1105.16 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). — 79 — 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 prevented by a bumper system compared to the unregulated bumper, Final Assessment at 29, J.A. 1010, and are reported in Columns 7 and 8. By way of illustration, at a speed of 0.0-3.0 mph, a 2.5 mph bumper prevents 76 percent of the damage and a 5.0 mph bumper prevents 85 percent of the damage that would have resulted with unregulated bumpers. Finally, the agency used the effectiveness values to determine the per-accident lifetime value of all repaired and unrepaired damage prevented by the two bumper systems. FRIA at VI-5 to -10. That computation is provided in Columns 9 and 10. The result, shown on Table III under Columns 9 and 10, is that the 2.5 mph system is 63 percent as effective as 5.0 mph bumpers (assuming that the economic value of damage an owner elects not to repair is 50 percent of the estimated cost of repair) or 67 percent as effective as 5.0 mph bumpers (assuming that unrepaired damage is valued at 75 percent of the estimated repair cost) in reducing the dollar value of damage expected from unregulated bumpers. FRIA at VI-4. — 80 — NHTSA then applied this conclusion to updated data on per-accident costs of unregulated and 5.0 mph bumpers,17 and the number of lifetime accidents per vehicle,18 to derive the present value of lifetime damage prevention for different bumper systems. FRIA at VI-5 to -11. The result is a disadvantage of the 2.5 mph system relative to the 5.0 mph system of $34-$69, FRIA at VI-10. (The range is attributable to the use of ranges for the discount factor for unrepaired damage (50-75 percent of the cost of repair), the number of lifetime accidents (2.45-3.20), and the value of lifetime damage incurred by 5.0 mph bumpers ($390-$578). FRIA at VI-8 to -10.) — 81 — Petitioners focus their challenge not on the soundness of the above-described methodology, but on the derivation of certain of the values shown in Table III. Even if the alleged deficiencies were clearly established, we would have some difficulty in believing that they are of such significance as to render the product before us capricious. For all of them inhered in the 1979 Final Assessment and were not thought worthy of mention by many of these same petitioners who were asked for and who furnished extensive comments at that time. Indeed, their utterly invalidating effect did not impress itself upon any of the commenters in the present rulemaking, until after [243 U.S.App.D.C. 140] the final rule was issued in 1982. We have nonetheless considered each of them and find them to be without merit. — 82 — Amicus National Association of Independent Insurers ("NAII") directs its criticism at NHTSA's computation of the frequency of accidents by speed intervals (Columns 2 and 4 of Table III). As noted above, NHTSA derived these figures from Ford Motor Company data, which in turn relied on barrier crash data for 1969-70 model year cars reported by the Insurance Institute for Highway Safety ("IIHS"), see Affidavit of E.J. Rohn, Addendum to Brief for Intervenor Motor Vehicle Manufacturers Association ("MVMA") at 2a ("Rohn Affidavit"). The Ford study computed accident frequencies for speed cells of 0.0-3.0 mph, 3.0-5.0 mph, 5.0-7.0 mph, 7.0-10 mph, and over-10 mph. See Letter from E.J. Rohn to D.W. McCallum at Chart I (Sept. 19, 1974), J.A. 2413C. The agency adopted these speed cells, FRIA at VI-2, though it broke down the over-10 mph range into 10-15 mph and over-15 mph cells using State Farm data. See Final Assessment at A-4, J.A. 1076. IIHS, however, had not conducted any barrier crashes at speeds less than 5.0 mph, leading NAII to charge here that Ford had no basis for breaking down the crucial 0.0-5.0 mph cell (containing 84% of all accidents in the Ford study) into 0.0-3.0 mph and 3.0-5.0 mph subcells, and that NHTSA therefore acted arbitrarily in relying on that information in constructing its effectiveness methodology. — 83 — This challenge is unfounded. The engineers who conducted the Ford study had collected data on the costs of repaired and unrepaired collision damage to about 18,000 cars, which established a distribution of accidents by damage. In order to translate this distribution into a distribution of accidents by speed, they relied upon IIHS test data that established average damage values of $208 at 5.0 mph, $597 at 10 mph and $853 at 15 mph. Using conventional mathematical extrapolation techniques, the Ford engineers then fitted a "cost-speed" curve to those three points, from which they derived their speed-frequency distribution using the narrower speed cells. Rohn Affidavit at 1a-2a. — 84 — Insurance Company Petitioners reiterate and expand the charge--made earlier by IIHS in its petition for reconsideration--that some of the assumptions underlying the agency's derivation of effectiveness values (Columns 7 and 8 of Table III) are questionable. Those values had been derived from so-called "effectiveness curves" constructed by the agency in 1979 on the basis of several critical assumptions: (1) that no bumper system provides complete protection at any speed; (2) that at their design speed, aluminum and steel bumpers are 60 percent effective; (3) that effectiveness falls to zero at replacement speed; and (4) that the effectiveness of bumpers is described by a curve of a particular shape. Department of Transportation, NHTSA, Calculations and Supporting Material for the Preliminary Analysis of the Bumper Standard at B-1 (Feb. 26, 1979), J.A. 1218 ("NHTSA Calculations"). By applying these assumptions to the available data for each bumper system,19 three points were established on a speed/effectiveness graph (corresponding to replacement speed, design speed, and 0.0 mph), through which the appropriate curve was drawn. The effectiveness values of the system for given speeds were then read right off the curve; and the effectiveness values for given speed intervals were "derived by estimating the area under each curve and calculating [243 U.S.App.D.C. 141] a mean value for each incremental speed range." Id. — 85 — The principal challenge to this analysis made by IIHS in its petition for reconsideration was that crash test data refuted the assumption that the 5.0 mph bumper loses all its effectiveness at its replacement speed (10 mph), showing that even Standard 215 5.0 mph bumpers "resulted in substantial reductions in damage in barrier impacts of 10 and 15 mph." IIHS Petition for Reconsideration at 8, J.A. 20. NHTSA responded fully to this point, noting that the IIHS tests were conducted under laboratory conditions and therefore failed to reflect certain factors in real-world collisions, that IIHS's conclusions were based on a small number of vehicles, and that there was "a great deal of year-to-year scatter in the data for individual vehicle types." 47 Fed.Reg. 56,649. More importantly, NHTSA concluded that if an adjustment were made for 5.0 mph bumpers, a comparable adjustment would have to be made for 2.5 mph bumpers. Id. — 86 — In this appeal for the first time, Insurance Company Petitioners seek to attack not merely the use of theoretical analysis (rather than crash test data) but the integrity of the analysis, describing the agency's initial assumptions of 60 percent effectiveness at design speed and zero effectiveness at replacement speed as "guesswork," and asserting that the shape of the curves drawn through the three points (as described above), being unsupported by any equation in the rulemaking record, was "arbitrary in the purest sense," Brief for Insurance Company Petitioners at 35. Such an ambush at this late stage cannot be allowed. The agency's analysis accompanying its final rule asserts that the assumptions "were based on engineering judgment of the agency's experts," 47 Fed.Reg. 21,831. We have no reason to disbelieve that statement, and engineering judgment is assuredly the sort of expertise that NHTSA preeminently possesses. The only real issue, then, is whether the details of the agency's engineering analysis on these narrow points were so critical to the ability to participate in the rulemaking as to cause their absence from the record to invalidate it, see Portland Cement Ass'n v. Ruckelshaus, 486 F.2d 375, 392-94 (D.C.Cir.1973), cert. denied, 417 U.S. 921, 94 S.Ct. 2628, 41 L.Ed.2d 226 (1974). It is impossible to think so, since if they were there would have been strenuous objection much before this. Both the assumption of 60 percent effectiveness at design speed and the shape of the effectiveness curves have been a part of the agency's methodology since the January 1979 Preliminary Assessment and were carried forward (with minor modifications in light of IIHS comments) to the June 1979 Final Assessment, which the insurance industry found not only acceptable, but congenial. — 87 — The Insurance Company Petitioners' response to this reality is disingenuous. They assert that NHTSA is incorrect to state "that it first heard of problems with its effectiveness curves in the IIHS petition for reconsideration," since "IIHS responded to the ... 1979 [Preliminary] Assessment within a few months of its issuance and submitted alternative curves." Reply Brief for Insurance Company Petitioners at 18-19 n. 27. The statement is true, but neglects to note that the IIHS alternative curves displayed the same alleged defects which petitioners claim invalidate NHTSA's curves--i.e., were drawn through three points, one of which represented the assumption of 60 percent effectiveness at design speed. Indeed, IIHS adopted without change the agency's effectiveness curve for 5.0 mph bumpers. Comments of IIHS, Docket No. 73-19-N.25-097, App. C at 4 (Apr. 30, 1979), J.A. 2240. Insurance Company Petitioners boldly assert that "no 'engineer' would attempt to draw a curve solely on the basis of only three points," id. at 18. But NHTSA did not purport to have designed the curve solely on the basis of three points. It drew the curve through the three points, but the shape of the curve was evidently determined by the location of those points and engineering theory. NHTSA Calculations at B-1, J.A. 1218. It is interesting that some "engineer" did the same thing in the Ford study of distribution [243 U.S.App.D.C. 142] of accidents by speed, discussed at pages 1359, supra. — 88 — The dissent's response is equally unsatisfactory, consisting essentially of the assertion that late presentation of the issue must be disregarded because validity of the curves is essential to validity of the agency's conclusion. Dissent at 1388-89. It is simply not the case, however, that all of the essential postulates for an agency rule must be contained in the record. Every judgment of any consequence is constructed upon an infinitude of other judgments, of greater or lesser certitude, in a progression of logical dependency terminating in a first principle the equivalent of 1 + 1 = 2. They cannot all possibly be included in the statement of basis and purpose for a rulemaking. We do not have authority to require that all elements underlying a rule be set forth in a fashion "understandable to a layman," see Vermont Yankee Nuclear Power Corp. v. Natural Resources Defense Council, Inc., 435 U.S. 519, 557, 98 S.Ct. 1197, 1218, 55 L.Ed.2d 460 (1978). And it may as well be disclosed that in scientific fields we judges ourselves are laymen, ill equipped to determine where the line falls between requisite explanation of problematic analysis and useless replication of what for the cognoscenti amounts to a textbook on basic physics. Thus, our "feel" for the adequacy of an analysis in such cases is necessarily governed by the reaction that it elicits from knowledgeable commenters. From the point at which it reaches common ground we can reasonably assume that no further explication is required. We will hear on appeal assertions that needful elaborations fairly requested were not provided; but we must be implacably skeptical of belated recognition at the appellate stage that elements of scientific analysis unchallenged during a contested proceeding are incomprehensible without further explanation. To credit such post-appeal pleas of inadequate information is to threaten the integrity of all rulemaking in fields beyond our own limited scientific ken. The present challenge to the effectiveness curves presents the threat in a particularly flagrant form. NHTSA specifically asked the petitioners (and other rulemaking participants) in 1979: "Do the existing analyses represent the most appropriate methods of approaching a study of bumper standards at different impact speeds and levels of damage resistance? If not, what method should be used?" Comments of IIHS, Docket No. 73-19-N.25-097 at 5 (Apr. 30, 1979), J.A. 2200; Comments of State Farm, Docket No. 73-19-N.25-098 at 1 (Apr. 30, 1979), J.A. 2059. Though State Farm and IIHS, among others, responded with substantial comments--see Comments of IIHS, supra, Apps. A-B, J.A. 2212-35; Comments of State Farm, supra, 1-6, J.A. 2059-64--there was not even a suggestion that the effectiveness curves represented a fundamentally invalid methodology. The dissent provides a list of reasons why petitioners might not have raised these objections during the rulemaking proceeding, Dissent at 1387-88 all of which boil down to petitioners' probable reluctance to upset favorable determinations. Even if that were the only consideration, it would be of questionable wisdom to reward their tactical decision to leave the agency in the dark--and thus encourage benighted agency action in the future. But in any event, the dissent's speculations do not explain why the proponents of the 2.5 mph standard did not raise these arguments. In fact, some commenters on the June 1979 Final Assessment (which supported the 5.0 mph standard) did object that the "effectiveness ratios for 5.0 mph systems are overstated to a considerable extent," Department of Transportation, NHTSA, Commentary on Critiques of the June 1979 Bumper Standard Assessment 35 (Dec. 1979), J.A. 951 (emphasis added), but that objection was based upon the lack of correspondence with repair cost data rather than theoretical inadequacy of the effectiveness curves. — 89 — We conclude that the curves were considered part of the common ground of expert analysis that required no further explanation. Petitioners' objections on this score put us in mind of the Supreme Court's injunction in Vermont Yankee that[243 U.S.App.D.C. 143] administrative proceedings should not be a game or a forum to engage in unjustified obstructionism by making cryptic and obscure reference to matters that "ought to be" considered and then, after failing to do more to bring the matter to the agency's attention, seeking to have that agency determination vacated on the ground that the agency failed to consider matters "forcefully presented." — 90 — 435 U.S. at 553-54, 98 S.Ct. at 1217. These instructions are even more appropriate when cryptic and obscure reference to theoretical inadequacy--or in fact, even less than that, mere assertion of lack of explanation of theoretical adequacy--is first made on appeal. — 91 — Insurance Company Petitioners also complain that, after correcting its initial assumption about the speed at which bumpers lose all effectiveness, in response to the IIHS comments (see note 19, supra ), NHTSA "did not redraw the [effectiveness] curves," yet increased from 73 to 76 percent the presumed effectiveness of 2.5 mph bumpers in the 0.0-3.0 mph speed interval. Brief for Insurance Company Petitioners at 35-36. In fact, however, the change in the 0.0-3.0 mph speed cell was not the only change in effectiveness values made by the agency between February 1979 and the publication of the Final Assessment in June of that year. The figures for 2.5 mph and 7.5 mph metal bumpers were adjusted in higher speed cells, and substantial changes were made in the effectiveness values for soft-face bumpers. Compare NHTSA Calculations at B-2, J.A. 1220 with Final Assessment at 30, 32, J.A. 1011, 1013. The logical inference is that the agency did in fact rethink and redraw its curves following IIHS's suggestion, noted above, that the zero-effectiveness point be set at twice design speed. 2. Value of Delay and Inconvenience — 92 — Petitioners challenge the range of $26-$50 per damage-producing accident assigned by NHTSA as the value of delay and inconvenience. In both the PRIA and the 1979 Final Assessment, the agency had used a flat figure of $26 calculated directly from estimates of the average hourly wage in the private sector, the number of occupants in the average car, the average amount of time lost on the scene of the accident and in obtaining estimates for repairs, and the cost of being without the car while it is being repaired. FRIA at III-39 to -41. (In 1979, State Farm commented that this analysis "reflects a reasonable effort ... to correctly account for the variety of time and cost savings due to the impact of [the] 5 mph bumper standard." Comments of State Farm, Docket No. 73-19-N.25-098 at 5 (Apr. 30, 1979), J.A. 2063 (quoted in FRIA at III-39).) It was precisely in response to comments suggesting that some of these estimates were low that the agency decided to use a range of $26-$50 instead of a flat $26 figure. FRIA at III-41. — 93 — Petitioners claim that NHTSA arbitrarily rejected the "only reliable evidence in the record--the survey of 1,000 consumers conducted by the Opinion Research Corporation for the Insurance Institute for Highway Safety--[which] showed that consumers place a value on such losses at $100 to $200 or more." Brief for Insurance Company Petitioners at 37 (footnote omitted). The agency rejected this survey because it concluded that many of the questions were biased or could only be answered by people with specialized knowledge--see 47 Fed.Reg. 21,835; FRIA at III-36 to -38--and because the Cost Savings Act does not contemplate accepting raw statements of preference of this type in calculating the costs and benefits of bumper systems, 47 Fed.Reg. 56,642. NHTSA rejected a second survey submitted by IIHS for similar reasons. Id. We readily affirm the agency in this regard. — 94 — C. Comparison of Costs and Benefits for Various Systems — 95 — If NHTSA had simply established best estimates for every component of cost and benefit, its identification of an optimal bumper standard would have required no more than addition of the estimates and [243 U.S.App.D.C. 144] comparison of the totals for each system. But in order to be realistic about uncertainty and conflicts in the data, the agency established ranges of possible values, rather than point estimates of the most likely values, for several of the key components of bumper system costs and benefits. Absolute cost and benefit figures were then converted into relative advantages (in terms of bumper system costs and benefits) using the 5.0/5.0 mph system as the baseline. The resulting product was a range of relative benefits and costs for each alternative based on the maximum possible variance obtainable from the variance in the constituent cost and benefit items.20 The FRIA sets out the agency's figures in Tables X-5 through X-8, FRIA at X-12 to -15, and summarizes those findings in Table X-9, FRIA at X-16, which is reproduced below as Table IV. Each horizontal row of the Table corresponds to the designated combination of extreme assumptions used: e.g., the first row employs the highest value in the "benefit" range (advantage in the cost of bumpers over the 5.0/5.0 system), and the lowest value in the "cost" range (disadvantage in expected damage cost reduction relative to the 5.0/5.0 system). — 96 — [243 U.S.App.D.C. 145] NHTSA's first task was to determine whether the existing standard--the 5.0/5.0 system--met the requirements of the Cost Savings Act. On that issue the outcome was clear: it provided significantly fewer benefits to the public and consumers than several of the alternatives. 47 Fed.Reg. 21,82 5. Only under the fourth set of extreme assumptions--one which the agency concluded would be especially unlikely to occur in reality, id. at 21,826--did the existing [243 U.S.App.D.C. 146] standard prove superior to the alternatives. That conclusion was not independently challenged by petitioners. — 97 — The next step was the agency's selection of the 2.5/2.5 standard over the remaining alternatives. NHTSA rejected, for reasons not at issue, all remaining alternatives except two--the 2.5/2.5 and 5.0/2.5 systems. Id. at 21,825-26. It then compared those two directly "under all examined sets of extreme assumptions, as well as under those sets of assumptions deemed by the agency most representative or most likely to occur." Id. at 21,826. The agency employed no fewer than four methods of comparison. First, it averaged the benefit and cost range end points and compared the resulting net benefits. This resulted in a $1 net advantage for the 5.0/2.5 alternative ($24 vs. $23 for the 2.5/2.5 alternative). FRIA at X-18 to -19. Second, it computed the simple arithmetic average of the net benefits associated with the four combinations of extreme assumptions. This again produced a $1 net advantage for the 5.0/2.5 bumper ($23 vs. $22 for the 2.5/2.5 alternative). Table IV, supra. Third, it recomputed the arithmetic average after excluding the values from the fourth combination of assumptions, which it had concluded was "highly unlikely." 47 Fed.Reg. 21,826. That indicated substantial superiority of the 2.5/2.5 bumper system--a $9 net advantage ($42 vs. $33 for the 5.0/2.5 alternative). Table IV, supra. Fourth, it computed averages for several of the variables for which ranges had been used--e.g., a sales-weighted industry average for primary cost savings. FRIA at XI-19. That gave the 5.0/2.5 system a $4 advantage over the 2.5/2.5 system. Id. In light of these analyses,21 NHTSA concluded that because the 2.5/2.5 bumper was markedly superior to the 5.0/2.5 alternative under the third method of comparison--which, as noted, excluded calculations based on "highly unlikely" assumptions--its cost-benefit analysis favored the less restrictive standard. 47 Fed.Reg. 21,826. — 98 — Petitioners point out that in three comparisons the 5.0/2.5 bumper was shown to be superior, albeit by only $1, $1, and $4 respectively, and they accuse the agency of blatant manipulation in rejecting, for purposes of its other comparison, the most unlikely set of extreme assumptions. It seems to us that such rejection was entirely reasonable. Averaging values from all four sets of extreme assumptions would be statistically valid only if each of the four combinations had an equal probability of occurrence. The agency explicitly found, however, that it was "virtually impossible" for the fourth combination--which included the lowest estimate of consumer price reductions, the highest estimate of the frequency of bumper-related accidents, and the highest values for delay and inconvenience--to occur in the real world. 47 Fed.Reg. 56,65 0. It reasoned that although each of these extreme assumptions may separately have some degree of probability, the probability that they will be accurate in combination is virtually zero. FRIA at XI-15 to -17. This finding was based on two considerations. First, although all four sets of assumptions involved extreme estimates, having only a remote possibility of occurring in combination, the agency considered the estimates in the excluded high-cost, low-benefit combination to be especially unlikely, id., a conclusion which finds ample support in the record. For example, with regard to the low-benefit $18 estimate of consumer savings attributable to the reduced purchase price of 2.5/2.5 bumpers, most companies submitted estimates at least $10 higher than the $18 figure, FRIA at XI-8 to -9, Ford and Chrysler both projected savings of $35, id. at VII-20, and NHTSA's own estimate of bumper component cost savings was $49--$14 higher than the upper-range figure based on manufacturer submissions and $31 (or 172 percent) higher than the $18 figure, id. at [243 U.S.App.D.C. 147] VII-21, -25. In addition, the agency's high-cost estimate of 3.20 average lifetime bumper accidents per vehicle, based on a submission by Allstate, was "not supported by data," id. at III-15, had "no independent factual basis ... in the record," id. at X-3, and indeed "was used solely to test the sensitivity of the results to changes in lifetime accidents," id. at XI-7, i.e., it was never intended to be a substantive part of a real-world cost-benefit analysis. Thus, in making its calculations of the expected real-world consequences of varying bumper standards, NHTSA was fully justified in heavily discounting--or even discarding--the set of assumptions containing both this unsubstantiated high estimate of accidents and the low $18 estimate of consumer cost savings.22 — 99 — Second, the agency suggested that the low-benefit, high-cost estimates involve contradictory behavioral assumptions. 47 Fed.Reg. 21,825; 47 Fed.Reg. 56,650; FRIA at XI-15. The only specific inconsistency pointed to by the agency, however, is the alleged incompatibility of the high-range figures for the number of lifetime bumper accidents (3.20 per vehicle) and the valuation of unrepaired damage (75 percent of the cost of repair). NHTSA reasoned that — 100 — where more than three accidents per year are assumed, the valuation of unrepaired damage at 3/4 the cost of repair would be inconsistent with the probability that at least one of such accidents would in fact lead to repair, and thus to the absence of unrepaired damage, and the absence of at least some of the inconvenience value due to the absence of need for repair estimates. — 101 — Id. at X-4 to -4a. Like petitioners, we are utterly baffled by this reasoning. In addition to wrongly assuming that the 3.20 figure represents accidents per year (whereas in fact it represents accidents per vehicle-life), this passage seems to assume that the unrepaired damage percentages (50 to 75 percent) represent the percentage of all damage that goes unrepaired--whereas they in fact represent, for the quantum of unrepaired damage that is independently calculated (see Column 4 of Table III, supra ), the percentage of the hypothetical repair cost that represents the value of the damage to the consumer. This passage bears every evidence of having been inserted as a make-weight by someone who had not the slightest idea what he was talking about. Even so, the agency's decision on this point is adequately supported by the alternative rationale based on the confluence of independently improbable assumptions. Considering the record as a whole, we cannot say that this single error on an alternative point--blatant though it may be--renders the entire rulemaking arbitrary or capricious. — 102 — Thus, the agency was justified in concluding that the cost-benefit analysis favored, even if only slightly, the 2.5/2.5 bumper. NHTSA further noted, however, that even if it used the methodology urged by petitioners, it would find the 5.0/2.5 and 2.5/2.5 standards "indistinguishable" and would in any event select the latter because it best promotes the policies of the Cost Savings Act in ways not considered by the formal cost-benefit analysis. 47 Fed.Reg. 21,826-27. Contrary to the petitioners' contentions, we think it was permissible for the agency to consider such factors. Once it was determined on cost-benefit grounds that the 2.5/2.5 and 5.0/2.5 standards were both clearly superior to the existing standard, NHTSA had an obligation under the Cost Savings Act to choose one of these alternatives. But if--as would be the case using the cost-benefit methodology recommended by petitioners--the cost-benefit analysis showed that neither alternative was appreciably superior to the other, then it was entirely reasonable for the agency to decide on the basis [243 U.S.App.D.C. 148] of factors that promote the policy of the Cost Savings Act, even if not included within the formal cost-benefit model. — 103 — Petitioners also challenge the propriety of the particular extra-model factors the agency considered. The primary factor relied on by NHTSA was that the less restrictive standard would produce benefits--e.g., a reduction in automobile purchase prices--more certainly and immediately than would the 5.0/2.5 standard. 47 Fed.Reg. 21,826. Petitioners fault this reasoning on the ground that the 2.5/2.5 standard already received full credit for the immediacy of its benefits under the cost-benefit model, because reductions in accident costs, the principal benefit of the more restrictive standard, were discounted to present value. This discounting, however, took account of the cost of money, not the inherently more speculative nature of the future costs. FRIA at III-65. Thus, having already reduced the costs of anticipated accidents to present value, NHTSA could reasonably favor allowing consumers to benefit from immediate concrete price reductions rather than forcing them to spend additional money now in order to forestall accident costs that probably will, but might never, materialize. — 104 — NHTSA's second extra-model factor was that the wider design freedom permitted by a less restrictive standard would promote innovation, which "could result in more effective bumpers at lower cost to the public than would otherwise be available," 47 Fed.Reg. 21,827. Petitioners find this to be "counter-intuitive" and "counter to experience." Reply Brief for Insurance Company Petitioners at 27. It does not seem to us counter-intuitive, since a reduced standard would enable manufacturers to experiment with designs and materials likely, but not certain, to achieve more than 2.5 mph protection without fear of violating the law if they fell short of 5.0 mph protection. Such new designs, if successful, might then make possible the design of bumpers providing 5.0 mph or greater protection at a lower cost than was previously possible. As for prior experience, it is for the agency, not the court, to determine whether conditions in the automobile industry today (as opposed to 1966 or 1972) are incompatible with the agency's assumption that production freedom will promote innovation. Cf. State Farm, 103 S.Ct. at 2872. Furthermore, petitioners do not challenge NHTSA's more general conclusion that reduced standards would promote innovation to achieve values other than increased bumper protectiveness --for example, increased ability to accommodate consumers' automotive design preferences and the collection of data on different bumper performances. See 47 Fed.Reg. 21,827. Petitioners also question whether the record comments support the conclusion that a 2.5/2.5 standard would promote innovation better than the 5.0/2.5 alternative. The conclusion, however, was not based on record comments, but on the (indisputable) proposition that lower standards increase design freedom, which, as discussed above, NHTSA reasonably thought would promote innovation. Id. — 105 — A third extra-model factor considered by NHTSA was that a 2.5/2.5 standard will promote commonality, and hence achieve lower production costs than a standard with different requirements for front and rear bumpers. Petitioners attack this as resting on the "ridiculous" assumption that front and rear bumpers are identical. Reply Brief for Insurance Company Petitioners at 28. This misstates the agency's reasoning, which was based on commonality, not of front and rear bumpers, but of front and rear bumper components. See 47 Fed.Reg. 21,827. It seems to us not at all ridiculous, and not arbitrary or capricious, to assume that front and rear bumpers meeting the same impact standards have more components in common than bumpers required to meet differing impact standards. Petitioners have not even asserted that this plausible assumption is false. — 106 — Finally, NHTSA noted that the less restrictive standard would increase international harmonization. 47 Fed.Reg. 21,828-29. Petitioners assert that this is irrelevant to the requirements of the Cost Savings [243 U.S.App.D.C. 149] Act. Even that contention seems to us questionable, since harmonization produces public benefits by promoting international commerce. In any event, NHTSA did not assert that harmonization was a benefit under the Cost Savings Act. Rather, it was relying upon the Trade Agreements Act of 1979, Pub.L. No. 96-39, Title IV, § 402, 93 Stat. 144, 242-43 (1979) (codified at 19 U.S.C. § 2532 (1982)), which instructs each federal agency that is developing standards to "take into consideration international standards and ..., if appropriate, base the standards on international standards." 19 U.S.C. § 2532(2)(A). NHTSA concluded that this statute was not controlling--i.e., did not require it to select the standard that best promoted harmonization, regardless of other factors--but that the 2.5/2.5 standard is "far more compatible" with relevant international regulations than the 5.0/2.5 alternative. 47 Fed.Reg. 21,829. We cannot say that this was an improper consideration. IV. IMPERMISSIBLE AGENCY MOTIVES — 107 — Petitioner CFAS, noting the circumstances surrounding the promulgation of this rule--the economically hard-hit domestic automobile industry and President Reagan's promises of assistance--advances the argument that NHTSA's decision was based on the "impermissible ground of aiding the U.S. automobile industry." Brief for Petitioner CFAS at 42. Petitioner refers to the White House Press Office publication, "Actions To Help The U.S. Auto Industry" (Apr. 6, 1981) ("Press Release"), which included a proposal to modify the bumper standard among thirty-four specific proposed regulatory actions. J.A. 2648-2715. "Three days after the booklet's release by the White House," petitioner observes, "NHTSA published a notice listing seventeen rulemaking actions--identical to those listed by the White House, and including a modification of the bumper standard." Brief for Petitioner CFAS at 44. — 108 — There seems to us nothing either extraordinary or unlawful in the fact that a federal agency opens an inquiry into a matter which the President believes should be inquired into. Indeed, we had thought the system is supposed to work that way. Nor is there anything remarkable in the fact that NHTSA published its proposals--"identical to those listed by the White House"--a mere three days after the Press Release (hardly time enough to give them serious consideration). For although they were published in the Federal Register three days later, they were transmitted to the Federal Register the same day as the Press Release, and indeed the forthcoming Federal Register text was attached as an Appendix to the Press Release. J.A. 2685-93. It is obvious, in other words, that the Press Release was the product of the agency's decisions rather than the cause of them. Nor, finally, is it disreputable that, underlying the President's belief that this rule should be reexamined, there lurked a desire to assist the domestic automobile industry--as numerous pieces of legislation, including multi-million dollar subsidized loans, amply show, see Chrysler Corporation Loan Guarantee Act of 1979, Pub.L. No. 96-185, 93 Stat. 1324 (1980) (codified as amended at 15 U.S.C. §§ 1861-75, 2003, 2512 (1982)). It would be a different matter if the President directed the agency, in the course of its inquiry, to disregard the statutory criteria controlling its actions. That is not even alleged, and could not plausibly be, since the text of the White House Press Release stated that the proposed actions "are described in considerable detail in the attachment," Press Release at 4, J.A. 2652, which attachment in turn (the forthcoming Federal Register Notice) began its statement of purpose by announcing that "NHTSA has undertaken this review to determine what, if any, modifications to its standards and regulations may be appropriate to reduce regulatory burdens upon the regulated industries without jeopardizing the safety or consumer-related goals and policies established by Congress in its related legislation." Id. at A-33, J.A. 2687. — 109 — The most petitioner can reasonably make of the Press Release is that it demonstrates [243 U.S.App.D.C. 150] an expectation, on the part of NHTSA officials, that the President would be pleased if the agency ended its rulemaking by reducing the existing standard, and disappointed if it left the current rule in effect. But lawyers and others in public service frequently must take or recommend action that is not what their superiors, in a world of unfettered discretion, would prefer. It is entirely absurd to suggest that a delegated decision is vitiated by the mere knowledge that the superior would have preferred it to come out the way it did. — 110 — The present regulation was adopted only after the painstaking and intricate analysis described in small part above. It replaced a standard that had been adopted in 1971 on the basis of an assumed connection between low speed accidents and safety having no empirical support, see page 1339, supra, and that had been retained in 1975 (despite agency analysis showing that 2.5 mph rather than 5.0 mph would be the better speed criterion) in part because of "[t]he reaction of Congress" to the proposed reduction, see Letter from Joan Claybrook, supra, at 1 (quoted at page 1340, supra ). It is unreal to suggest that we should reinstate the old rule because the new one bears evidence that the purity of reason has been distorted by politically derived value judgments. V. MOTION FOR A LIMITED REMAND — 111 — Petitioner State Farm has filed what it terms a Motion in the Alternative for a Limited Remand Pursuant to 15 U.S.C. § 1913(b).23 Petitioner requests that, if we find (as we have done) that the 2.5 mph standard should not be overturned, we remand the case to NHTSA to consider evidence which has surfaced since the conclusion of the rulemaking proceeding, demonstrating that NHTSA vastly underestimated the increase in insurance repair and overhead costs resulting from a change to a 2.5 mph bumper standard and did not correctly consider the effects on safety of side marker lamps.24 Assuming that this evidence is material and that there were reasonable grounds for failure to adduce it in the agency proceeding, we nonetheless decline to grant a remand. — 112 — Section 1913(b) provides that, if the two conditions noted above are satisfied, this court "may order such additional evidence ... to be taken before the Secretary ...." (Emphasis added.) This is not an appropriate case in which to exercise that discretion, since a remand would simply produce judicial interference in an ongoing agency review. Federal bumper standards have been subject to review, revision, and proposed revision from the moment they were first promulgated, and the process will not end with this decision. NHTSA has indicated in its denial of petitions for reconsideration that if petitioners submit their data, "the agency will consider them as part of its ongoing review of the Bumper Standard." 47 Fed.Reg. 56,650. Indeed, one of the effects of the new rule will be to generate real-world data on the effects of a 2.5 mph standard for use in future proceedings. 47 Fed.Reg. 21,827. Since nothing prevents petitioners from submitting their new evidence to the agency as part of the continuing review process, [243 U.S.App.D.C. 151] a remand at this juncture would be ill-timed and ill-considered. — 113 — Other factors potentially relevant to disposition of motions such as this one similarly militate against remand: the length of the rulemaking process that would be reopened and the cost to the public of delay in implementation of the rule. See American Optometric Ass'n v. FTC, 626 F.2d 896, 907 (D.C.Cir.1980). — 114 — * * * — 115 — * * * — 116 — The view of this rulemaking that emerges from the foregoing analysis is of course a distorted one. Petitioners have challenged and we have addressed only those estimates which arguably understate the net benefits of the 5.0 mph bumper, but there are perhaps as many instances in which the potential bias goes the other way. For example, while petitioners have charged that NHTSA's analysis of the relative effectiveness of different bumper systems overstates the protection of the 2.5 mph bumper, one study indicates that it may well understate the protection. See Houdaille Study. Similarly, while petitioners assert that NHTSA's assessment of the value of delay and inconvenience was too low, NHTSA recognized that its estimate, even before adjustment upward to take into account some of petitioners' comments, exaggerated the benefits of the 5.0 mph system--because all damage-producing accidents were counted in deriving the comparative bumper benefits, even though many such accidents surely go unrepaired and cause little, if any, delay and inconvenience. FRIA at III-41. Finally, and most important of all, NHTSA's cost-benefit analysis simply assumed that all automobile manufacturers would adopt a 2.5 mph bumper system, whereas it seems probable that some manufacturers will continue to produce and will advertise more protective bumpers.25 The cost-benefit results of this circumstance are most unlikely to be random, since the more protective bumpers will appeal precisely to those consumers whose vehicles are more expensive to repair and whose time is economically more valuable, thus eliminating cases which skewed the agency's average figures against the 2.5 mph standard. Since the Cost Savings Act requires NHTSA to implement the bumper standard, not the bumper system, which minimizes costs for consumers and the public, 15 U.S.C. § 1912(b)(1), it seems that the proper referent for the agency's analysis should have been not a nation with 2.5 mph bumper cars, but a nation with the mix of bumpers that a 2.5 mph standard would produce. The agency's failure even to attempt such a calculation was perhaps its most obvious inadequacy--and it was an inadequacy that went entirely to the petitioners' advantage. — 117 — These comments suggest why it is appropriate that our standard of review of the substance of agency decisions be as limited as the APA provides. We do not reverse simply because there are uncertainties, analytic imperfections, or even mistakes in the pieces of the picture petitioners have chosen to bring to our attention, see American Textile Mfrs. Institute, Inc. v. Donovan, 452 U.S. 490, 528-29 & n. 52, 101 S.Ct. 2478, 2500 & n. 52, 69 L.Ed.2d 185 (1981); but only when there is such an absence of overall rational support as to warrant the description "arbitrary or capricious." That description is plainly not appropriate here. While the points on which the present petitioners can plausibly disagree with NHTSA's conclusions are numerous, that is to be expected in an enterprise of such difficulty and scope. The enterprise was pursued, however, through a methodology that was sensible; during the course of the proceeding NHTSA responded to, and often accommodated, every major criticism and suggestion; and its conclusions are within the range of those that a reasonable person could derive from the evidence presented. No more is necessary to survive our review. — 118 — Petition denied. — 119 — [243 U.S.App.D.C. 152] J. SKELLY WRIGHT, Circuit Judge, dissenting: — 120 — The court today upholds the National Traffic Highway and Safety Administration's (NHTSA) rollback of its bumper performance standard from the old requirement that front and rear bumpers withstand impacts of 5.0 mph (5.0/5.0 mph standard) to the new requirement that front and rear bumpers need only withstand impacts of 2.5 mph (2.5/2.5 mph standard). The majority finds that the agency did not act arbitrarily and capriciously in determining that the lessened standard (1) would be in accordance with the safety requirements of the National Traffic and Motor Vehicle Safety Act of 1966, 15 U.S.C. § 1391 et seq. (1982) (Safety Act), and (2) would represent, overall, the most cost-effective standard and thus the proper standard under the Motor Vehicle Information and Cost Savings Act of 1972, 15 U.S.C. § 1911 et seq. (1982) (Cost Savings Act). — 121 — I cannot agree with the majority's conclusion or with most of its steps in reaching that conclusion. In my judgment, NHTSA's actions in this rulemaking were clearly arbitrary and capricious and not in accordance with its statutory mandates. NHTSA first dealt in a wholly inadequate manner with the safety concerns which prompted promulgation of the bumper standard originally. Then, having dismissed these concerns, NHTSA proceeded to perform a cost-benefit analysis that appears, given the contortions that the agency went through to reach its final conclusion, to have been solely a formalistic exercise aimed at justifying a preordained result. Throughout these deliberations, the agency ignored clear legislative direction as to what policies were to be given precedence. The majority's approval of these actions in the name of deference to agency expertise is, I believe, an unacceptable retreat from our judicial responsibility to carefully review agency decisionmaking and an unsupportable condonation of agency failure to act in accordance with explicit instructions from the legislative branch. Consequently, I dissent. I. BACKGROUND — 122 — As the majority notes, the bumper standard at issue here was promulgated under two separate statutes: The Safety Act and the Cost Savings Act. Thus the 5.0/5.0 mph standard that was revoked in this rulemaking represented both a motor vehicle safety standard under the Safety Act and a bumper standard under the Cost Savings Act. Consequently, alteration of this standard must, as the majority recognizes, be reviewed to ensure that such alteration was in accordance with the requirements of both relevant Acts. A. Statutory Framework — 123 — The Safety Act provides for promulgation of motor vehicle safety standards to protect the public "against unreasonable risk of accidents" and "against unreasonable risk of death or injury to persons, in the event accidents do occur." 15 U.S.C. § 1391(1), (2). The 5.0/5.0 mph bumper standard was originally set forth as a motor vehicle safety standard under this Act in 1971. This safety standard remained in its original form until the agency decision under review here.1 Thus the alteration of [243 U.S.App.D.C. 153] the bumper standard from 5.0/5.0 mph to 2.5/2.5 mph was, first of all, a change in a safety standard. Consequently, the court must first review that alteration to ascertain that the agency's decision was in accordance with the strictures of the Safety Act. — 124 — Our review under the Safety Act, however, is only the first step. For the 5.0/5.0 mph standard was also a bumper standard promulgated under the Cost Savings Act, and we must also review the alteration of the standard for its compliance with that Act. Under the Cost Savings Act, NHTSA is to formulate a bumper standard so as to obtain "the maximum feasible reduction of costs to the public and to the consumer." 15 U.S.C. § 1912(a), (b)(1). This Act, in addition to specifying this overall criterion, also specifies particular factors for the agency to consider in determining the most cost-reducing bumper standard: — 125 — (A) the cost of implementing the standard and the benefits attainable as the result of implementation of the standards; — 126 — (B) the effect of implementation of the standard on the cost of insurance and prospective legal fees and costs; — 127 — (C) savings in terms of consumer time and inconvenience; and — 128 — (D) considerations of health and safety, including emission standards. — 129 — 15 U.S.C. § 1912(b)(1). The Cost Savings Act also states explicitly that the Safety Act considerations are to be given priority: "Bumper standards under this subchapter shall not conflict with motor vehicle safety standards promulgated under * * * the * * * Safety Act." Id. § 1912(b)(2). This statutory language is supported by consistent indications throughout the legislative history of the Cost Savings Act that safety concerns were to be overriding. See, e.g., S.Rep. No. 92-413, 92d Cong., 1st Sess. 20 (1971) ("For example, if a cost-beneficial bumper standard increased the likelihood of, or severity of, injury to occupants, then the Secretary would not promulgate that particular standard. The considerations of safety would take precedence."); H.R.Rep. No. 92-1033, 92d Cong., 2d Sess. 23 (1972) ("Safety must remain the paramount and overriding purpose of the Department."). Thus any alteration of the bumper standard to achieve cost reductions as mandated by the Cost Savings Act can take place only after a determination that such alteration meets all of the criteria for an alteration of the bumper standard as a motor vehicle safety standard under the Safety Act. Any review under the Cost Savings Act, therefore, must give precedence to the Safety Act review. B. Standard of Review — 130 — As the Supreme Court noted recently in Motor Vehicle Mfrs. Ass'n v. State Farm Mutual Auto. Ins. Co., 463 U.S. 29, 103 S.Ct. 2856, 2865, 77 L.Ed.2d 443 (1983), revocation or amendment of a motor vehicle safety standard under the Safety Act is to be reviewed according to the judicial review provisions of the Administrative Procedure Act. Thus the reviewing court is to set aside the agency's action if it is found to be "arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law." 5 U.S.C. § 706(2)(A) (1982).2 And, since the Cost Savings Act [243 U.S.App.D.C. 154] also provides for the applicability of the APA review provisions, see 15 U.S.C. § 1913(c), a revocation or amendment of a bumper standard under that Act is to be reviewed under the same standard. — 131 — As discussed by the Supreme Court in State Farm, the scope of review under the arbitrary and capricious standard is narrow. This does not mean, however, that an agency has unlimited discretion or that a reviewing court may abdicate its review responsibility by simply deferring to the agency without thoroughly investigating and evaluating the agency's actions. As the Supreme Court noted, the agency is required to "examine the relevant data and articulate a satisfactory explanation for its action including a 'rational connection between the facts found and the choice made.' " State Farm, supra, 103 S.Ct. at 2866-2867, quoting Burlington Truck Lines v. United States, 371 U.S. 156, 168, 83 S.Ct. 239, 245, 9 L.Ed.2d 207 (1962). For example, an agency has failed to meet this general requirement, and has acted in an arbitrary and capricious manner, — 132 — if the agency has relied on factors which Congress has not intended it to consider, entirely failed to consider an important aspect of the problem, offered an explanation for its decision that runs counter to the evidence before the agency, or is so implausible that it could not be ascribed to a difference in view or the product of agency expertise. — 133 — State Farm, supra, 103 S.Ct. at 2867. — 134 — Judicial review of agency action requires that the reviewing court "consider whether the decision was based on a consideration of the relevant factors and whether there has been a clear error of judgment." Id., quoting Bowman Transportation, Inc. v. Arkansas-Best Freight System, Inc., 419 U.S. 281, 285, 95 S.Ct. 438, 441, 42 L.Ed.2d 447 (1974); Citizens to Preserve Overton Park, Inc. v. Volpe, 401 U.S. 402, 416, 91 S.Ct. 814, 823, 28 L.Ed.2d 136 (1971). As noted by this court in Int'l Ladies' Garment Wkrs. Union v. Donovan, 722 F.2d 795, 815 (D.C.Cir.1983), cert. denied, --- U.S. ----, 105 S.Ct. 93, 83 L.Ed.2d 39 (1984), such review "is not merely perfunctory. We are to engage in a 'searching and careful' inquiry, the keystone of which is to ensure that the [agency] engaged in reasoned decisionmaking." See also Telocator Network of America v. FCC, 691 F.2d 525, 537 (D.C.Cir.1982). — 135 — This formulation points to an acknowledged disparity between the depth of our review and the ultimate scope of that review: Although the ultimate scope may be narrow, the depth must be sufficient for us to be able to comprehend the agency's handling of the evidence cited or relied upon. The purpose of this in-depth review is to educate ourselves so that we can properly perform our reviewing function: determining whether the agency's conclusions are "rationally supported." United States v. Allegheny-Ludlum Steel Corp., 406 U.S. 742, 749, 92 S.Ct. 1941, 1946, 32 L.Ed.2d 453 (1972); Ethyl Corp. v. EPA, 541 F.2d 1, 34-35 n. 74 (D.C.Cir.) (en banc ), cert. denied, 426 U.S. 941, 96 S.Ct. 2663, 49 L.Ed.2d 394 (1976). For, although data interpretation and analysis are functions that often lie within an agency's realm of expertise, see, e.g., Office of Communication of United Church of Christ v. FCC, 707 F.2d 1413, 1440 (D.C.Cir.1983), it is our duty to review those functions to ascertain whether the agency's actions were complete, reasoned, and adequately explained. See id. The mere fact that an agency is operating in a field of its expertise does not excuse us from our customary review responsibilities. And, where the agency's reasoning, although complex, is rational, clear, and complete, we must affirm. Contrarily, where the agency's reasoning is irrational, unclear, or not supported by the data it purports to interpret, we must disapprove the agency's action. See Ethyl Corp., supra, 541 F.2d at 36. — 136 — [243 U.S.App.D.C. 155] With this standard of review in mind, I will first examine whether the agency's rulemaking decision was a reasoned and lawful decision under the Safety Act; that is, I will ascertain what the agency's decisionmaking processes were and determine whether they were arbitrary and capricious. Second, I will examine whether or not the agency's decision was arbitrary and capricious under the Cost Savings Act.3 Only if the rulemaking survives both of these inquiries can it be upheld. I turn now to these inquiries. — 137 — II. REVIEW OF THE AGENCY'S DECISION FOR COMPLIANCE WITH THE SAFETY ACT — 138 — As the majority notes, the agency's findings at issue here under the Safety Act relate to two separate safety issues: (1) the issue of bumper protection, in low-speed collisions, of vehicle safety features which, if damaged and malfunctioning, may later cause additional accidents; and (2) the issue of bumper contributions to energy management in collisions at speeds above the design speed of the bumper (crashworthiness). The majority finds that the agency adequately fulfilled its responsibilities under the Safety Act. I disagree with this conclusion for several reasons. First, I am uncertain, due to a lack of clarity on the part of the agency, as to precisely what the agency did in terms of its Safety Act obligations. I am thus dubious that the majority's assumptions as to the agency's actions are correct. Second, even if the majority's assumptions are correct, I disagree with its conclusion that the agency's actions complied with the Safety Act. In this section, I will state and discuss my concerns. — 139 — A. The Agency's Lack of Clarity and the Majority's Assumptions With Respect to the Agency Actions — 140 — Any investigation of an agency rulemaking decision begins with an examination of the agency's conclusions and the findings that purportedly support those conclusions. This elementary step, however, is hindered in this case by the agency's complete lack of clarity regarding its findings. It is clear that the agency concluded that safety concerns did not prevent its proposed rulemaking. Precisely what findings it relied upon in reaching that conclusion are unfortunately obscure. Specifically, it is unclear whether the agency found that the reduction in the bumper standard would have no reduction in safety or no significant reduction in safety. — 141 — In its final rulemaking statement, NHTSA noted that it had found that the rollback of the bumper standard from 5.0/5.0 mph to 2.5/2.5 mph would "cause no reduction in vehicle safety" with regard to either safety feature protection or crashworthiness. 47 Fed.Reg. 21820 (1982), Joint Appendix (JA) 85 (emphasis added). Specifically, the agency found that "adoption of the 2.5-mph/2.5-mph alternative [would] not have any measurable effect on the risk that future accidents might be caused by safety components which malfunction due to damage incurred in prior low speed collisions and which are left unrepaired." Id. at 21827, JA 92. The agency also found that "reducing the test speeds for the safety criteria [would] not measurably affect the high-speed crash energy management of cars." Id. Later, however, in its findings relating to its denial of the petition for reconsideration of its decision, NHTSA stated simply that it had "concluded that reduction of the 5/5 bumper standard would not have any significant effect on safety." 47 Fed.Reg. 56643 (1982), JA 4 (emphasis added). Yet this conclusion was based entirely on the same evidence as its previous statement, with only one exception: the question of crashworthiness at speeds lower than 30 mph. There, the agency noted that it had concluded that the change in the bumper design speeds would "have a negligible effect" on crashworthiness at speeds of 10-15 mph and higher. Id. — 142 — [243 U.S.App.D.C. 156] Thus the findings that we are to review here were expressed in different terms at different times. The finding of no reduction in safety accompanying the final rule was transformed somehow into a finding of no significant reduction in the statement accompanying the denial of the petition for reconsideration. Consequently, precisely what the agency concluded is unclear from its official statements.4 — 143 — Nor is it clear from the underlying record. In fact, my review of the record indicates manifest ambiguity not only on this point but also on the question whether the agency ever considered safety concerns under the Safety Act at all. In the Preliminary Regulatory Impact Analysis (PRIA) the four-page discussion of safety was extremely cursory and entirely speculative; no conclusions were stated and the Safety Act was not mentioned. In the preliminary notice of rulemaking the only discussion of safety was a solicitation of data regarding any reduction on meeting the motor vehicle safety standard for crashworthiness (not the bumper standard), see 46 Fed.Reg. 48267 (1981), JA 375; the agency did not mention the Safety Act and, in fact, seemed to assume that the bumper standard was no longer a motor vehicle safety standard under that Act. See id. at 48262, JA 369. — 144 — In the Final Regulatory Impact Analysis (FRIA), responding to heated comments that safety concerns were paramount, the agency did mention the Safety Act briefly, noting that it applied only to significant safety risks and stating that it had concluded that "no meaningful adverse safety effects [would] occur." FRIA at IV-3, JA 196. In its more detailed discussion of the safety issues, however, it did not appear to be considering safety effects in light of the significance criterion. Thus it was not clear whether it had determined that the effects did not require application of the Safety Act or whether they were in compliance with its requirements. — 145 — The agency's conclusions were also unclear in the more detailed discussion of the safety issues. The agency stated that the safety effects would be "at the upper range very minor." FRIA at IV-15, JA 208. Specifically, it stated that there would be "no discernible effect on crash energy management." Id. It also stated that — 146 — lower impact speed requirements with no change in uniform height requirements, will bring about any significant increase in damage to bumper standard-related safety components [sic ]. Even in such cases the inoperation of such safety components is itself a very minor factor in accident causation. Such accidents show at most a conjectural relationship with injuries and fatalities[.] — 147 — Id. — 148 — Finally, the final notice of rulemaking, in addition to its stated conclusions of no reduction in safety, contained two short incomprehensible paragraphs stating in a wholly conclusory manner that the agency was not conceding that the Safety Act criteria applied but that, if they did, the rulemaking here was in accordance with them.5 [243 U.S.App.D.C. 157] In asserting this conclusion, however, the agency in no way explained how those criteria were met in this case. Nor did it point to a relevant explanation in the underlying documents. In fact, only its assertion and its "no significant effect" conclusion support its ultimate conclusion that the change complied with the Safety Act. — 149 — I believe this lack of clarity on the part of the agency renders impossible affirmance of its decision here, for several reasons. First, it is inadequately clear that the agency considered the Safety Act requirements at all. This is not a casual or superficial problem that can be glossed over: Safety concerns were mandated by Congress to prevail. Thus failure to properly consider the Safety Act requirements evidences fundamental inconsistency with the legislative mandate. Where such a critical issue is at question, we should hesitate to attribute actions to the agency that would validate its decision unless its path truly " 'may reasonably be discerned.' " State Farm, supra, 103 S.Ct. at 2867, quoting Bowman Transportation, Inc., supra, 419 U.S. at 286, 95 S.Ct. at 442. — 150 — Second, even assuming, as the majority does, that the agency did consider the Safety Act applicable, its interpretation of the requirements of that Act appears to me to be inadequately justified and very possibly inconsistent with the Act's explicit mandate. As noted above, the Safety Act provides that motor vehicle safety standards are to be promulgated to protect the public from unreasonable risks of accident and of injury from accidents. The Safety Act also provides certain criteria that the agency must consider in prescribing such standards. Specifically, the agency must — 151 — (1) consider relevant available motor vehicle safety data, including the results of research, development, testing and evaluation activities conducted pursuant to this chapter; — 152 — (2) consult with the Vehicle Equipment Safety Commission, and such other State or interstate agencies (including legislative committees) as [it] deems appropriate; — 153 — (3) consider whether any such proposed standard is reasonable, practicable and appropriate for the particular type of motor vehicle or item of motor vehicle equipment for which it is prescribed; and — 154 — (4) consider the extent to which such standards will contribute to carrying out the purposes of this chapter. — 155 — 15 U.S.C. § 1392(f). These criteria merely specify what factors the agency is to consider in its rulemaking process; they do not add to or subtract from the statutory mandate of the agency to regulate against unreasonable risks. The legislative history, however, does give some indication of how the agency is to treat these factors. For example, both the House and Senate Reports on this bill noted specifically that safety was to be the "overriding" or "paramount" concern. See S.Rep. No. 1301, 89th Cong., 2d Sess. 6 (1966) ("safety shall be the overriding consideration in the issuance of standards under this bill"); H.R.Rep. No. 1776, 89th Cong., 2d Sess. 16 ("[m]otor vehicle safety is the paramount purpose of this bill"). — 156 — The agency, however, in its general discussion of the requirements of the Safety Act, ignored the factors the statute requires it to consider and the explicit statutory mandate regarding unreasonable risks and noted that all it would have to do would be to find that any safety differences between a 5.0 mph bumper and a 2.5 mph bumper were insignificant. See FRIA at IV-3, JA 196. It based its reliance on this single criterion, to the exclusion of any [243 U.S.App.D.C. 158] consideration of the explicit statutory factors, on a single sentence in the Senate Report on this bill. See id. In that Report the Senate Committee considering the bill noted that "[t]he Secretary is not expected to issue a standard covering every component and function of a motor vehicle, but only for those vehicle characteristics that have a significant bearing on safety." S.Rep. No. 1301, supra, at 6. The agency, relying solely on this lone sentence, evidently decided that a finding of insignificance would be enough to allow it to disregard entirely the explicit statutory requirements. Thus it is not clear to me that the agency's action, under the Safety Act, was supported by adequate findings as required by the statute or even governed by the appropriate statutory standard. — 157 — The majority defends the agency's emphasis on significance, rejecting application of the statutory standard of unreasonableness (and apparently condoning the lack of the statutorily required findings), by noting that "[t]he principle that an 'unreasonable risk' provision requires even insignificant risks to be eliminated if that can be done at (presumably) insignificant cost would turn many areas of regulation into unending pursuit of the trivial." Maj. op. at 1344 n. 5. My point is not that the agency must regulate every unreasonable risk, however small. Rather, my point is that a finding of significance properly is relevant only in the agency's initial determination regarding whether a particular vehicle characteristic should be subject to regulation. As the Senate Committee asserted, the agency is only expected to regulate "those vehicle characteristics that have a significant bearing on safety." S.Rep. No. 1301, supra, at 6. Here, that threshold finding was clearly made by the agency when it decided to issue a standard for bumper performance. And no party to this proceeding challenges this initial determination. Having decided that bumpers are significant enough to warrant regulation, the proper standard to apply in determining what specific bumper standard to promulgate is the explicit statutory standard of unreasonableness. To import into this second determination an additional significance requirement, as the agency did, is to ignore specific legislative direction regarding the proper inquiry.6 — 158 — Finally, the lack of clarity confuses the determination of whether the agency's decisionmaking was rational because it is difficult to know what the decision was. Most importantly, the majority conducts its entire review under the assumption that the agency's findings, even its findings of no reduction in safety or no measurable reduction in safety, were essentially findings of no significant reduction. This exclusive focus is problematic because it may enable approval of the agency's decision when that decision could not stand if measured against the no reduction findings. — 159 — The majority's treatment of the agency's action assumes away all of these issues. In reaching its conclusion the majority assumes that the agency's findings were findings of no significance and that significance is the sole and controlling relevant criterion under the Safety Act, and then concludes that the agency's findings were reasonably supported and adequate under that Act (although it does not assert that the agency considered the evidence under the Safety Act). I believe, at the outset, that our review is made impossible by the [243 U.S.App.D.C. 159] lack of clarity with respect to whether or how the agency confronted or resolved these issues and that remand to the agency is therefore appropriate at the outset.7 Nevertheless, because I also disagree with the majority's conclusion that, even accepting its assumptions as to the agency's actions, those actions were reasonable and not arbitrary and capricious, I will discuss these specific disagreements below. I do not, however, concede the propriety of the majority's underlying assumptions. — 160 — B. Finding of No Significant Reduction in Safety — 161 — The majority approves what it takes to be NHTSA's determinations that the reduction in bumper standards will not significantly reduce safety (1) by reducing protection of other safety features or (2) by reducing the car's ability to withstand collisions. Even assuming the agency did make these findings, I see no rational connection between them and the record upon which they are supposedly based. — 162 — 1. Effects of change in bumper standard on safety feature protection. According to the majority, the agency relied on a single study in reaching its conclusion that the proposed change in the bumper standard would have no significant effect on other safety features that might prevent accidents. Specifically, it relied on the Tri-Level Study of the Causes of Traffic Accidents prepared by the Institute for Research in Public Safety at Indiana University. See JA 2743-2809 (Executive Summary of Tri-Level Study ). The agency noted that "very few accidents occur as a result of malfunctioning of those vehicle components which are subject to the safety criteria of the bumper standard." 47 Fed.Reg. at 21827, JA 92. It stated further that it had concluded "that far fewer accidents could be attributed, and only by speculation, to a failure to repair such components after they had been damaged in the only type of collision relevant to this discussion, i.e., one which might occur at an impact speed between 2.5 and 5.0 mph." Id. — 163 — In determining whether the agency's projected conclusions here are rationally based on the record, a reviewing court must trace the agency's path from initial data to final result. In this instance the agency started with the assumption, "for purposes of analysis," FRIA at IV-11, JA 204, that a reduction in the bumper standard could result in increased damage to other safety features. The agency, in making this assumption, made no attempt to quantify the extent of this additional damage. In fact, the agency rejected totally the only data submitted on this point, a State Farm study that indicated that more damage to safety features occurred with Model Year 1973 cars (which were required to comply approximately with a 2.5 mph bumper standard) than with Model Year 1974 cars (which were required to comply with a 5.0 bumper standard). The agency dismissed these data because the MY 1973 car bumpers were not sufficiently similar to the contemplated 2.5 mph bumpers to draw any conclusions from these data because bumper heights in 1973 were not uniform (although petitioners pointed out that in 1974 only 1974 model cars had uniform bumpers) and the Fuel System requirements did not apply. See FRIA at [243 U.S.App.D.C. 160] IV-9 to -11, JA 202-204. The agency thus concluded that State Farm had not shown that the difference between the two systems was noticeable and consequently downgraded the likelihood and extent of such a difference.8 Starting with this unknown difference, the agency then proceeded to conclude that even if such a difference existed it would have little effect on actual safety. This conclusion was based on a series of discounts of the chance that reduced bumper protection of other safety features would result in decreased safety in fact. — 164 — This series of discounts began with the statistical data in the Tri-Level Study. Those data indicated that, of all accidents included in that study, 0.44-0.48% were certainly caused by problems with safety features that would be protected by the bumper standard, 0.93-1.67% were probably caused by such problems, and 1.6-5.7% were possibly caused by such problems. See Executive Summary of Tri-Level Study at 21, JA 2768. First, without any explanation the agency ignored entirely the "possible" data and stated only the certain and probable ranges indicated by the study. See FRIA at IV-12, IV-13, JA 205, 206. Next, noting that non-functioning vehicle lights and signals comprised the largest component of the causal percentages, the agency proceeded to conclude that because "the study reports that older vehicles are [243 U.S.App.D.C. 161] over-involved in accidents resulting from mechanical problems * * *, it would thus not be unreasonable to attribute natural wearing out of bulbs, or less commonly, high impact crashes, as the most typical causes of inoperation." FRIA at VI-13, JA 206. The agency's conclusion, however, does not follow from the study's findings. The study, with respect to older vehicles, notes simply that "[t]he probability of a vehicle 8 years of age or older causing an accident (by virtue of a mechanical problem) is 2.1 times greater than for vehicles in general." Id., quoting Tri-Level Study at 71; see also Executive Summary of Tri-Level Study at 23, JA 2770. This fact indicates merely that an older vehicle is more likely than a newer vehicle to cause an accident because of a mechanical problem. This fact does not in any way indicate what percentage of the accidents that are in fact caused by mechanical problems are caused by older vehicles: Only with data regarding the mix of vehicles involved in these accidents or regarding the mix of vehicles on the road overall could the inference drawn by the agency logically be drawn. Thus, even if the agency were correct in its (wholly unsupported) assumption that mechanical defects in older vehicles are more likely to be caused by natural wearing out of parts or by high speed crashes than by failure to repair damage from low-speed crashes, it still did not show any reasonable basis for concluding that these are "the most typical causes of inoperation" of those mechanical components. FRIA at IV-13, JA 206. — 165 — The agency then further discounted the safety effects of those accidents that may result from mechanical problems by noting that "the contribution of such accidents to injury was extremely rare." FRIA at IV-14, JA 207. This conclusion was based on its observation of the study's conclusion that other factors were overrepresented in personal injury accidents. Although the Tri-Level Study does indicate that other factors are more likely to correspond to personal injury accidents, the percentage of accidents caused by problems with malfunctioning bumper-related safety features that resulted in some personal injury was estimated to be perhaps 20%, surely not an "insignificant" percentage. See Executive Summary of Tri-Level Study at 25, JA 2772. — 166 — Cumulating all these discounts--quantified and unquantified, proven and speculative--the agency concluded, in the majority's view, that any safety effects of this type from changing the bumper standard to 2.5 mph would not be significant. Although the reasoning underlying this conclusion is quite opaque, it appears that the agency's logic was along the following lines: Because, by its calculations, the accidents potentially affected by any change in the bumper standard would form only a small percentage of all personal injury accidents, those accidents were insignificant, even though the magnitude of the differences in safety protection provided by various bumper standards was unknown and might be quite large. — 167 — This conclusion is irrational for two reasons. First, the "insignificant" percentage was very probably not as small as its erroneous assumptions indicated. All of the unsupported assumptions or conclusions made by the agency in reaching its result had the effect of understating the percentages at issue. Thus it is not clear that the correct percentage would be insignificant even according to whatever (undisclosed) criteria of insignificance the agency was using. — 168 — Second, as petitioners point out, the assumption that a safety standard is necessarily insignificant if it affects only a small percentage of accidents runs directly counter to the statutory mandate of the Safety Act. In that Act Congress dealt specifically with vehicular features that might impair or improve highway safety. That vehicular factors are a relatively small part of all safety risks does not mean that any vehicular factor alone is prima facie insignificant (most vehicular accident causes, by definition, will affect only a small percentage of even vehicular causes generally, let alone all accident causes). Thus the fact that a particular risk is merely a small [243 U.S.App.D.C. 162] component of a larger risk does not mean that the smaller risk itself is insignificant. This conclusion gains support from this court's decisions under the Safety Act regarding defects. See United States v. General Motors Corp., 518 F.2d 420, 438 n. 84 (D.C.Cir.1975) ("We use the term 'significant' to indicate that there must be a non-de minimus [sic ] number of failures. * * * The number of failures need not be and normally will not be a substantial percentage of the total number of components produced." (emphasis added)). Thus the agency's conclusion that the fact that any potential safety effects would occur in only a small percentage of cases meant that those safety effects were insignificant is not, without more, consistent with its statutory mandate. Nor did the agency make any attempt to explain why this particular small percentage was insignificant. That is, it formulated no criteria for defining what an insignificant safety risk might be. It simply concluded that this particular small percentage was insignificant. This is the essence of unreasoned and unsupported decisionmaking. 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) (plurality opinion) (noting that the agency may ascertain what risk levels are significant but requiring that it "determine, in the first instance, what it considers to be a 'significant' risk").9 — 169 — The majority's review of the agency's action here is completely unconvincing. First, the majority accepts in its entirety the asserted assumption that any risk that is a small percentage of total risk is insignificant. Second, the majority approves each and every assertion made by the agency whether or not that assertion is supported in the record. The majority notes with acceptance each discounting of the percentage figure of accidents potentially caused by bumper-related safety features. Then, noting that these "massive" discounts would "obviously" apply, maj. op. at 1345, the majority concludes that it was reasonable for the agency to conclude that the difference between the 5.0 and the 2.5 mph bumper systems was not measurable and not a significant safety consideration. The majority makes no attempt to evaluate in any way the intermediate steps taken by the agency or the reasonableness of those steps. It simply notes those steps and accepts them. I cannot agree that this superficial summary comprises the searching and careful inquiry to which we are bound. Nor can I agree with the majority's conclusion that the agency's action on this point was reasonable. — 170 — 2. Effects of change in bumper standard on vehicle crashworthiness. NHTSA also, according to the majority, concluded that the change in the bumper standard would have no significant effect on the crashworthiness of motor vehicles in crashes at speeds higher than the design speeds. The agency based this conclusion on estimates submitted by car manufacturers and by State Farm. The manufacturers submitted comments to the effect that a bumper system of either 5.0 or 2.5 mph design would contribute only about 5% of the total energy management in 30-mph [243 U.S.App.D.C. 163] crash tests. State Farm submitted evidence to the effect that a 2.5 mph bumper would be only 25% as effective as a 5.0 mph bumper in managing crash energy. See FRIA at IV-4 to -5, JA 197-198. — 171 — Several manufacturers also submitted comments to the effect that "in 30 mph perpendicular barrier crashes * * * in certain cases it may be possible to have a better optimized 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." Volvo of America Corporation, Docket Comments at 3, JA 1655 (quoted in FRIA at IV-5 to -6, JA 198-199). See also General Motors Corporation, Docket Comments at A-5, JA 1773; Ford Motor Company, Docket Comments at 24, JA 1835; Volkswagen of America, Inc., Docket Comments at II, JA 1722. — 172 — Finally, one car manufacturer (Volvo) indicated that a 5.0 mph bumper might in fact be more "aggressive" than a 2.5 mph bumper and that it thus might actually be more dangerous, especially in side impacts. See FRIA at IV-4 to -5, JA 198-199. Another manufacturer (General Motors), however, noted that there would be no difference in aggressiveness between the two types of bumpers. Id. at IV-5, JA 199. — 173 — From these comments the agency concluded that "[t]he difference in the energy management capability of 5.0 mph bumpers and 2.5-mph bumpers is negligible at crash speeds such as those [30 mph] specified in the safety standards regulating the crashworthiness of new cars." 47 Fed.Reg. at 21827, JA 92. — 174 — Although not disputing this particular conclusion, the Insurance Institute for Highway Safety, in its comments submitted in support of the petition for reconsideration of the agency decision, noted that "the agency did not address * * * the contribution of 5 mph bumpers to energy management in crashes at speeds of less than 30 mph, which produce the vast majority of occupant injuries. * * * At these speeds, 5 mph bumpers with energy absorbers reduce the violence of the impact forces * * * far better than at 30 mph." Comments of IIHS at 10, JA 22. — 175 — The agency, in its statement accompanying the denial of the petition for reconsideration, stated that — 176 — the agency fully considered that issue as it related to all speeds, notwithstanding the fact that certain of the illustrations cited by the agency involved 30 mph crashes. As IIHS pointed out, it is true that severe injuries can and do occur in crashes at speeds as low as 10-15 mph. However, at those speeds as well as at higher speeds, 5 mph bumpers contribute insignificantly to changes in the management of the crash energies to which vehicle occupants are subjected. — 177 — 47 Fed.Reg. at 56643, JA 4. The agency went on to explain this conclusion by noting that 5.0 mph bumpers could only have significant effects on crashworthiness at these lower speeds if they had the effect of slowing the decrease in velocity of a car upon crashing. And it noted that it had determined that 5.0 mph bumpers would not have this effect because only differences in total energy absorption capability can have this effect and these differences are small percentages of the total energy at crashes at 10-15 mph and above. The agency also commented that the potentially more aggressive nature of the 5 mph bumpers could actually adversely affect crashworthiness. — 178 — For all but the last of these apparently factual statements, the agency produced no evidentiary support. It produced no support for its statement that it had considered crashes at lower speeds in its final rulemaking (or in the preliminary rulemaking assessment). And, according to the administrative record, the only facts considered in the rulemaking concerned the 30-mph crashes. The agency also produced no support for its conclusion that 5 mph bumpers have only a minimal effect on total energy management at speeds above 10-15 mph. The comment of the agency with respect to bumper aggressiveness was the only statement supported by the record. That support, however, was of dubious [243 U.S.App.D.C. 164] validity. The only car manufacturer that noted that 5 mph bumpers might increase the aggressiveness of the car in collisions was Volvo, a manufacturer whose estimate of primary weight effects the agency excluded in another part of the rule-making because, as the majority notes, it had "very small sales in the United States." Maj. op. at 1353. The agency, however, neglected to mention a comment to precisely the opposite effect by General Motors, see FRIA at IV-6, JA 199, clearly not an insignificant contributor to the entire American car market. Thus the agency's finding of "no significant effect" at all speeds above 10 mph also has no basis in the record and is therefore arbitrary and capricious. — 179 — The majority dismisses any objections to the agency's conclusions with the comment that "[p]etitioners do not advance any data showing that there is better crash energy management with a 5.0 mph bumper system." Maj. op. at 1350. With all due respect, this is just not the point. The point is not whether petitioners can mount a conclusive evidentiary showing that their (common sense) view--that stronger bumpers will provide greater protection in automobile collisions--is correct. The question is whether the agency's ultimate conclusions and subsidiary findings to the contrary are adequately reasoned from the available evidence. — 180 — The majority chalks up the agency's conclusions to its expertise in generalizing from field studies and defers to those conclusions on that basis. See maj. op. at 1350 n. 10. Although a reviewing court, as discussed above, should defer to an agency in such matters, such deference should not equal total capitulation. The agency here started with manufacturer estimates for 30 mph crashworthiness. It then, in considering the petition for reconsideration and not before, concluded that it could extrapolate these estimates to crashworthiness at all injury-producing lower speeds as well. It based these extrapolations on factual assertions that are supported in the record by only one piece of evidence, and a dubious piece of evidence at that. I do not believe that deference extends this far. — 181 — In sum, therefore, to the extent that it considered the Safety Act at all, the agency first substituted the questionable criterion of "significance" for the explicit statutory criteria of the Act. Then, it mistakenly interpreted the meaning of the term "significant." Finally, the agency formulated a chain of calculations, some of which were erroneous, to show that, under that mistaken interpretation of a questionable standard, the safety effects of changing the standard would be acceptable under the statute. This comprised its entire treatment of safety--the factor which legislative direction clearly indicated should be given conclusive weight. Considering these layers of error on the part of the agency, I would hold, under the Safety Act, that the agency has acted arbitrarily and capriciously and not in accordance with law in determining that there would be no unacceptable safety effects from the change in the bumper standard. — 182 — III. REVIEW OF THE AGENCY'S DECISION UNDER THE COST SAVINGS ACT — 183 — Although, as detailed above, I would find that the agency's inadequate and irrational consideration of the safety ramifications of its rule under the Safety Act preclude affirmance, I also disagree with the majority's conclusion that the agency's decision was sufficiently reasonable to withstand judicial scrutiny under the Cost Savings Act. My disagreement with the majority's wholesale acceptance of the agency's cost-benefit analysis under that Act ranges from the basic approach to the specific choices. I will detail each area in which my review has resulted in my concluding that the agency acted unreasonably: the treatment of health and safety costs, the calculation of benefits from changing the bumper standard, the calculation of the costs of changing the standard, and the agency comparison of the costs and benefits of the alternatives it considered.[243 U.S.App.D.C. 165] A. Treatment of Health and Safety Costs — 184 — My first, and perhaps most major, area of concern is with the treatment of health and safety costs under the Cost Savings Act. Both the agency's and the majority's safety analyses go only to their conclusion that the Safety Act does not prevent the proposed change in the standard. Thus the agency prefaced the only discussion of safety in its Final Regulatory Impact Analysis with a discussion of how the Safety Act required only the maintenance of "standards on significant problems, not all problems." FRIA at IV-3, JA 196. It then went on to find that the Safety Act did not preclude its chosen action. The majority also notes, in reviewing the agency's findings with respect to safety, that "[t]he Safety Act's mandate is not * * * categorical. Not all risks of accident or injury are to be eliminated, but only those that are 'unreasonable[.]' " Maj. op. at 1343. — 185 — Although the Safety Act mandates a reasonableness analysis when the agency seeks to revoke or amend the bumper standard as a motor vehicle safety standard, such an analysis does not conclude the agency's responsibility in this case. For the Cost Savings Act explicitly requires that the agency, in its determination of what bumper standard will reduce costs to the maximum feasible extent, take into account "considerations of health and safety." 15 U.S.C. § 1912(b)(1)(D).10 This requirement is entirely separate from the provision that bumper standards promulgated under the Cost Savings Act cannot conflict with motor vehicle safety standards under the Safety Act. See 15 U.S.C. § 1912(b)(2). Taking into account the settled rule of statutory interpretation that a statute is to be read, where possible, to give effect to every provision of that statute, I conclude that 15 U.S.C. § 1912(b)(1) requires that, under the Cost Savings Act, NHTSA not simply determine whether a safety risk is unreasonable but also incorporate any safety effects from a change in the bumper standard into the cost-benefit analysis mandated by the Cost Savings Act itself.11 — 186 — It is possible that the agency originally intended its safety findings of no effect on safety to obviate any need for consideration of the costs of any reduction in safety under its Cost Savings Act cost-benefit analysis. Under the agency's (and the majority's) current interpretation of its safety findings, however, it has not fulfilled the Cost Savings Act requirements. The agency (and the majority) should not be able to have it both ways. If the agency found no safety effects, those findings should be reviewed as such under the Safety Act analysis. If it found no significant effects, the lack of inclusion of the costs of those effects in its cost-benefit analysis constitutes the "fail[ure] to consider an important aspect of the problem" noted by the Supreme Court to be a primary indicium of arbitrariness and caprice. See State Farm, supra, 103 S.Ct. at 2867. It is also not in accordance with the law. See 5 U.S.C. § 706(2)(A) (1982). Thus the agency decision should be overturned on this basis alone.[243 U.S.App.D.C. 166] B. Calculation of Benefits of Changing Bumper Standard — 187 — In its analysis NHTSA stressed the benefits associated with degradation of the bumper standard--benefits that it estimated would consist mainly of cost savings to the industry (which savings the agency assumed would be passed through in their entirety to consumers due to the competitive nature of the automobile manufacturing industry). In particular, the savings would stem from the reduced weight of the bumper systems themselves ("primary weight" savings) and the reduced weight of the vehicle components that support the bumper systems ("secondary weight" savings). The cost savings estimated to result from the change in the bumper standard would, the agency found, come about in two ways: Some would occur directly as cost savings due to lesser materials requirements and some would occur indirectly as fuel savings. The majority traces through petitioners' arguments that these estimates were unacceptably biased and that certain factors were inadequately considered and resolved by the agency. With respect to most of these issues, I believe those decisions are either reasonably supported by the record and within the agency's field of expertise or sufficiently insignificant not to warrant reversal alone. I will just say that, with respect to most of these cost savings issues, the agency's conclusions were based on specific evidence before it and the decisions that it made with respect to that evidence were within its area of expertise, even though it consistently chose to credit those data that tended to support the ultimate conclusion it finally reached. — 188 — With respect to one issue raised by petitioners, however, I cannot accept the majority's approval of the agency's decision. Specifically, petitioners note that secondary weight reductions, according to industry statements, would not occur immediately and that those weight reductions should, therefore, have been discounted to their present value in the benefit calculations. In the face of this comment the agency acknowledged that it was true that production of the 2.5 mph bumper vehicles would not begin immediately. See 47 Fed.Reg. at 56647, JA 8 ("Production requirements and market pressures will constrain any move immediately to 2.5 mph bumpers only."). But, having acknowledged this, the agency totally ignored the criticism of its discounting methodology. — 189 — The majority takes the agency's bald statement acknowledging that the 2.5 mph bumpers would not be produced immediately and interprets it as a statement that the criticism of the methodology with respect to the secondary weight savings is invalid because "the same could be said of primary weight reductions." Maj. op. at 1354. The majority then proceeds to refine this creative interpretation as follows: The agency's admittedly unreal assumption for the purposes of its cost-benefit analysis does distort the magnitude of the costs and benefits that the new standard would produce but it does not distort the relevant calculation, i.e., the "relative net benefit of one standard over the other." Id. (emphasis in original). — 190 — I cannot agree with either step of the majority's reinterpretation of the agency statement. First, I do not agree that an admission that the 2.5 mph bumpers would not begin to be produced immediately can fairly be construed by this reviewing court to carry with it the relatively complete explanation attributed to the agency by the majority. All the agency said was that the IIHS comment was factually accurate, at least in part. Second, the explanation extracted by the majority from the agency's unelaborated statement is flawed in and of itself. As I understand the majority's explanation, it means that even though the costs and benefits of the 2.5/2.5 mph standard would change if discounted from, say, five years in the future to the present, the 5.0/5.0 standard's or the 5.0/2.5 standard's costs and benefits would change as well. Thus the discounted present values of the future costs and benefits of all the standards will bear the same relation to each other as do those future costs and benefits [243 U.S.App.D.C. 167] without discounting.12 This explanation, although perhaps superficially attractive, makes two assumptions that are nowhere supported in the record--and one of which is, in fact, contradicted by a statement of the agency itself. Specifically, this explanation assumes that the calculated costs and benefits of each standard will accrue at the same time. It also assumes that implementation of whatever standard was chosen (except the 5.0/5.0 standard) would occur at the same point in the future. — 191 — The first assumption is explicitly contradicted by a statement that the agency made in the same place as its generalized statement that the majority is "interpreting": The agency, in its denial of the petition for reconsideration, notes that, although — 192 — the agency does not expect many vehicles to be produced with 2.5 mph bumpers that do not have some corresponding changes in secondary weight[, t]o the extent that any such vehicles are produced in the short term, the associated costs are normal transition costs associated with any change in requirements. The possibility that the rulemaking may have such short-term costs is outweighed by the significant benefits that owners of future cars will experience as a result of the new standard. — 193 — 47 Fed.Reg. at 56647, JA 8. Thus the agency admits that the cost savings associated with the secondary weight saving of a 2.5 mph standard might not accrue for some time after the benefit losses associated with that standard have begun. To airily dismiss this discrepancy as a "normal transition cost" and to assume, entirely without basis, that such costs will in the long run be outweighed by the eventual cost savings ignores basic principles of cost-benefit analysis.13 — 194 — The majority's second assumption, although not directly contradicted in the record, is not supported by the record either. Nowhere does the record note that the cost savings associated with the 2.5/2.5 standard will accrue at the same time as those associated with, for example, the 5.0/2.5 standard. And common sense would indicate that the changes associated with the 5.0/2.5 standard, requiring fewer changes from the status quo, might actually occur sooner. Particularly where these two standards were close in terms of the per-vehicle savings calculated by the agency, the assumption of no time difference in implementation is one that must be at least considered by the agency.14 — 195 — In sum, the agency's patent failure to consider even the question raised by petitioners is yet another indication of its arbitrariness and caprice. The majority's attempt to avoid facing this fact by supplying an explanation that is nowhere apparent in the agency's statements and that may itself be wholly invalid is not a defensible exercise of judicial review. As the Supreme Court noted in State Farm, supra, 103 S.Ct. at 2867 (citing SEC v. Chenery Corp., 332 U.S. 194, 196, 67 S.Ct. 1575, 1577, 91 L.Ed. 1995 (1947)), "We may not supply a reasoned basis for the agency's action that the agency itself has not given." The reasons for this general rule are [243 U.S.App.D.C. 168] clearly apparent in the instant situation. Although the agency may have intended to explain its action with the (potentially flawed) analysis offered by the majority, it is equally as likely that the agency either simply ignored the point raised by the comment or resolved it in an entirely different, but unfortunately undisclosed, fashion. Thus, although I conclude that the agency's treatment of the cost savings associated with movement away from the 5.0/5.0 mph standard was largely reasonable, I cannot agree with the majority that its treatment of the potential time effects of any delay in secondary weight effects are adequate, reasonable, or, without further explanation by the agency, affirmable by this court. — 196 — C. Calculation of Costs of Changing Bumper Standard — 197 — The costs calculated by the agency to be associated with the proposed movement away from the 5.0/5.0 mph bumper standard are those increased repair and associated costs that will result from increased damage to motor vehicles in low-speed crashes. Petitioners' main disagreement with the agency cost estimates stems, as the majority notes, from a disagreement with the particular intermediate values used by the agency, not with the methodology used by the agency. — 198 — The primary focus of petitioners' argument is the curves relied upon by the agency to determine the effectiveness of a 2.5 mph bumper relative to a 5.0 mph bumper (effectiveness curves). In drawing these theoretical curves the agency made certain assumptions about all the bumpers it was comparing. It assumed: — 199 — (1) No bumper system of any design speed that is in the range considered (up to 10 mph BEV) is 100 percent effective in any incremental speed range. — 200 — (2) All bumper systems reach 0 percent effectiveness at replacement speed. Bumpers are not fully effective in all crash configurations up to design speed because, for example, of override/underride problems. Hence full effectiveness cannot be assumed. (3) At design speed, aluminum and steel bumpers are 60 percent effective and soft facia correspondingly higher. * * * (4) Bumper Effectiveness curves are of * * * approximate [S-shape]. — 201 — Calculations and Supporting Material for the Preliminary Analysis of the Bumper Standard at B-1, JA 1218. Petitioners now contend that these assumptions are unsupported and invalid. — 202 — The majority dismisses these points for reasons which are entirely unrelated to their potential merit. The majority notes that these assumptions, and the resulting effectiveness curves, first appeared in the 1979 Assessments of the Bumper Standard prepared by the agency (both Preliminary and Final). The majority then concludes that the fact that petitioners first raised this point on this appeal means that it cannot be critical to the agency's rulemaking: — 203 — It is impossible to think so, since if they were there would have been strenuous objection much before this. Both the assumption of 60 percent effectiveness at design speed and the shape of the effectiveness curves have been a part of the agency's methodology since the January 1979 Preliminary Assessment and were carried forward (with minor modifications in light of IIHS comments) to the June 1979 Final Assessment, which the insurance industry found not only acceptable, but congenial. — 204 — Maj. op. at 1360. — 205 — The majority's assumption that inaction by petitioners can be taken as conclusive of the unimportance of the agency's actions here is utterly unsupported and, in addition, not indicative of the type of searching judicial review we are obligated to perform. For one thing, there might have been other reasons for petitioners not to object to the figures in the 1979 Assessments. For example, those Assessments supported retention of the 5.0/5.0 standard, a result which petitioners have supported all along. See, e.g., Final Assessment at 88, JA 1069. And those Assessments specifically stated that [243 U.S.App.D.C. 169] the disputed effectiveness curves were conservative assumptions: There were other data, data that would have indicated a greater relative effectiveness on the part of the 5.0 bumpers, that the agency did not use. See, e.g., Final Assessment at 31, F-18, F-19, JA 1012, 1167, 1168. Perhaps petitioners' failure to object can be attributed to the fact that even with these conservative assumptions the 5.0 standard was found to be the optimal one. Further, with respect to the use of these estimates in the current rulemaking, the main alternatives presented in the PRIA were these estimates or the use of actual production data, data which the agency eventually rejected (they would have made the 5.0 standard look more favorable) as being based on inapplicably limited data. Perhaps petitioners focused on the differences between the alternatives offered in the PRIA rather than on the weaknesses of each alternative. These factors lend support to the hypothesis that there may have been alternative reasons for petitioners' failure to object before this appeal and undercut the majority's superficial presumption of insubstantiality.15 — 206 — More importantly, whatever the reasons for the delay in making the arguments (and assuming that they were not, in fact, raised in their precise form below16 ), that delay is irrelevant. We review the agency's action, not petitioners'. And I consider petitioners' argument on this point, whenever raised, to be compelling evidence of the hasty and arbitrary path followed by the agency in reaching its foregone conclusion. That is, I believe this issue to be sufficiently critical to require more explanation than that given by the agency and endorsed by the majority ("engineering judgment"). The curves in question are the sole and entire basis for all of the calculations regarding the costs of the increased damage that could be expected from the rollback of the standard--the major component of all the costs of the rollback. Given this, I do not see how the majority can dismiss them as narrow points. — 207 — If one looks beyond the majority's facile conclusion that the absence of prior complaints aimed at the precise points raised here conclusively demonstrates their uncritical nature to the underlying merits of petitioners' claim, one sees more evidence of unsupported and unreasoned decisionmaking on the part of the agency. For example, the first assumption contested by petitioners is that of when the bumpers reach 0% effectiveness.17 After the 1979 Preliminary Assessment IIHS commented that the assumed replacement speeds were not valid. See Comments of IIHS at 3-5, JA 2240-2242. In the 1979 Final Assessment the agency did not portray revised effectiveness curves, but it did change some of the effectiveness numbers (shown in tabular form). One change was that the relative effectiveness at low speeds of 2.5 bumpers was increased, although 0% effectiveness, according to the IIHS comments, should have occurred at a lower speed. The majority notes that the fact that other [243 U.S.App.D.C. 170] changes occurred as well indicates that "[t]he logical inference is that the agency did in fact rethink and redraw its curves following IIHS's suggestion." Maj. op. at 53. It may indeed be a logical inference that the agency redrew the curves, but the important point for our review here is not that it redrew the curves but how it redrew them. And nowhere in the 1979 Final Assessment are the redrawn curves shown (although values different from the original ones are used); nor is the agency's methodology for redrawing them described or explained. Yet the relative effectiveness values tabulated in the Final Assessment were lifted without change and used in the current rulemaking. — 208 — The second assumption at issue is the selection of the 60% effectiveness number at design speeds. Although the reasoning behind this assumption is quite opaque, the statement by the agency to the effect that "[b]umpers are not fully effective in all crash configurations up to design speed because, for example, of override/underride problems," Calculations and Supporting Material at B-1, JA 1218, indicates that at least one concern leading to the assumption was that bumpers of unequal heights would lock and create additional damage. Yet, as the agency has pointed out in the numerous contexts in which it has rejected 1973 real world data (the only such data) with respect to several of the variables estimated by the agency, such data cannot be applied to the proposed 2.5 bumpers, which will all be of uniform height. Where the agency rejects some data based on one rationale and bases other estimates on that same rationale, those decisions appear to me to be unreasoned, at least without some explanation of this inconsistency on the part of the agency. Further, even without that inconsistency, the question of what other factors prompted the agency to arrive at this figure remains unanswered. — 209 — Finally, the question as to the shape of the effectiveness curves remains. Although the general shape of the curves appears to be invoked by several of the parties here involved, it seems that an agency should be required to at least explain the basis for that shape. The majority notes that the agency explained that these were based on the agency's engineering judgment and that deference to its result is therefore appropriate. I agree wholeheartedly that deference to an agency's engineering expertise is appropriate, but I cannot agree that, where the agency merely refers to its expertise and does not in any way explain the derivation of the curves at issue, it has adequately shown that deference is warranted. Agencies cannot just cite their expertise and magically unveil a number. They must state the underlying assumptions, uncertainties, and policy decisions with respect to those uncertainties. In consequence, I would find the agency's action in formulating these curves to be inadequately justified. Consequently, reliance on those assumptions, without more, is arbitrary and capricious. — 210 — D. Agency Comparison of Costs and Benefits of Alternatives — 211 — Having formulated estimates of the costs and benefits of the alternatives before it, the agency, rather than arriving at one figure for each variable, formulated a range of values for each. From these ranges, it then extracted a high and low range for the costs and the benefits of each alternative. Using these ranges, the agency then compared the cost-effectiveness of each alternative. To accomplish this, it evaluated the ranges it had before it in four separate ways. See maj. op. at 1364-65. With three of these, the 5.0/2.5 alternative came out to be the most cost-effective. With the fourth, the 2.5/2.5 alternative prevailed. The agency then determined that this fourth method was the most accurate and proceeded to declare that the 2.5/2.5 alternative had been shown to be the best of those considered. — 212 — Under this fourth alternative the agency arrived at four possible net benefit (or net cost) figures for each alternative by combining the high and low range cost values for each alternative with the high and low range benefit values for each alternative. Thus, it had a high cost/high benefit, a [243 U.S.App.D.C. 171] high cost/low benefit, a low cost/low benefit, and a low cost/high benefit net value estimate for each alternative. In the fourth calculation of what these values meant, the calculation that the agency eventually relied upon, the agency excluded entirely the high cost/low benefit number and simply averaged the other three. It explained this move as being a result of its conclusion that it was "virtually impossible that the factual elements of that combination of assumptions could occur in reality, in large part because of inherent contradictions in economic or behavioral results that would be associated with such alignment." 47 Fed.Reg. at 21825, JA 90. See also 47 Fed.Reg. at 56650, JA 11. As the majority points out, this assertion on the part of the agency "bears every evidence of having been inserted as a make-weight by someone who had not the slightest idea what he was talking about." Maj. op. at 1366. The majority concludes, however, that the agency formulated another reason for excluding this fourth number: that "although each of these extreme assumptions may separately have some degree of probability, the probability that they will be accurate in combination is virtually zero." Id. at 1365. As with the majority's discussion of the discounting problem of secondary weights, however, this has two problems: First, the agency nowhere expressed this justification, and second, the justification itself is incoherent as applied to what the agency actually did. — 213 — With respect to the first point, the agency's conclusions with respect to the inherent improbability of the confluence of these two unlikely sets of assumptions are, wherever expressed, justified solely, or at least largely, on the basis of contradictory underlying assumptions. See 47 Fed.Reg. at 21825, JA 90; 47 Fed.Reg. at 56650, JA 11; FRIA XI-15 to -17, JA 345-347. The majority's interpolation of an additional independent ground is simply not based on the agency's own statements. — 214 — Second, the majority's interpretation not only does not support the action taken by the agency, it further draws that action into question. For if there is no causal connection between the low cost/high benefit values, there is not greater probability that either will occur with or without the other. Thus the agency, if it had decided that these values were too inherently improbable to be included at all in the analysis, would have had to exclude or at least discount the low cost/low benefit and high cost/high benefit numbers. Yet it did not do so. This indicates either that the agency was indeed relying on the "make-weight" argument rejected by the majority or that the agency was incorrectly following the majority's substitute rationale. Further, although the evidence before the agency may have supported some discounting of certain values based on their inherent probability, the rational way of dealing with this would have been to formulate probability estimates for all four values. No justification for the total exclusion of one value was given. Nor was a justification for discounting that value and not the others. Thus under neither view of the agency's rationale does the agency's action appear to have been reasoned decisionmaking. In sum, the majority's explanation is no more coherent than that it rejected, and the selection of the 2.5/2.5 alternative through the methodology used was arbitrary and capricious. — 215 — Having gone through these contortions just to reach the result that the 2.5/2.5 standard was the superior one, the agency reached back to justify that result even with an analysis based on the straight averaging of all four value combinations. It noted that even using all four value combinations the 5.0/2.5 standard was only marginally superior to the 2.5/2.5 standard. (This marginal difference was $1 net additional benefits per car; the agency did not bother to translate this difference into a total (and perhaps not so insignificant) yearly cost of approximately $11,000,000.18 ) The agency then concluded [243 U.S.App.D.C. 172] that, where the two standards were essentially equivalent, it would choose the 2.5/2.5 standard anyway. The reasons for this choice it enumerated as follows: First, the 2.5/2.5 standard "imposes the least direct, immediate costs on the consumer, i.e., the least increase in the cost of a new car. To illustrate this point, if the unregulated bumper is considered the baseline," more immediate costs would be required to comply with the bumper requirements. 47 Fed.Reg. at 21826, JA 91. The problem with this justification is that the unregulated bumper is not the baseline. The 5.0/5.0 standard is the baseline. Agencies are to start from the current rule in evaluating changes, not from some theoretical or historical unregulated state. See State Farm, supra, 103 S.Ct. at 2866. The majority notes that the agency is entitled to discount more speculative costs in favor of more certain costs. Even assuming that these costs are more certain (an assumption which does not appear to be based on anything in the record), the place to incorporate such a discount is in the cost-benefit calculation, not as a post hoc justification for choosing the alternative the agency wanted in the first place. — 216 — Second, the agency notes that the lower standard would permit innovation, which "could result in more effective bumpers at lower cost to the public than would otherwise be available." 47 Fed.Reg. at 21827, JA 92. The agency also notes that the same standard for front and rear will promote commonality (common components) and thus reduce prices of bumpers. Assuming that such costs and/or cost savings should be considered and that they occur in actuality, the proper way to consider them is to estimate them and to incorporate them into the cost-benefit calculation. Trotting out auxiliary hypothetical and unquantified costs to rationalize one choice or another after having supposedly performed the cost analysis mandated by the statute is manifestly inconsistent with reasoned decisionmaking. Thus I conclude that this aspect of the agency's analysis was arbitrary and capricious and should not be upheld by this court. IV. CONCLUSION — 217 — In sum, I have looked carefully at the steps taken by the agency in promulgating this rule and have come to the conclusion that this court should properly reject it as an arbitrary and capricious exercise of agency power. The agency, in this case, stepped beyond the bounds of legitimate policy choice and moved into the field of unreasoned ratification of preordained conclusions. The majority, unfortunately, has exercised such deference to the agency's statements, however opaque, unsupported, or unreasoned they might be, that it has abandoned entirely its mandate of rigorous judicial review. So, too, has it failed to ensure that the agency acted in consonance with its legislative mandate. This I cannot accept, and I therefore respectfully dissent. Footnotes [1] In the perpendicular barrier test, the vehicle collides with a fixed barrier, such as a wall, while travelling at the test speed [2-1] 2 The pendulum test consists of a series of blows to the inert vehicle at the test speed by a specially designed block whose effective impacting mass equals that of the tested vehicle [3] The nine new alternatives were: Alternative IA: 5.0 mph front/2.5 mph rear/Phase II damage criteria. Alternative IB: 5.0 mph front/2.5 mph rear/Phase I damage criteria. Alternative IIA: 5.0 mph front/height only rear/Phase II damage criteria. Alternative IIB: 5.0 mph front/height only rear/Phase I damage criteria. Alternative IIIA: 2.5 mph front/2.5 mph rear/Phase II damage criteria. Alternative IIIB: 2.5 mph front/2.5 mph rear/Phase I damage criteria. Alternative IVA: 2.5 mph front/height only rear/Phase II damage criteria. Alternative IVB: 2.5 mph front/height only rear/Phase I damage criteria. Alternative V: height only front/height only rear. See Notice of Rulemaking, 46 Fed.Reg. 48,262, 48,266-67 (1981) (proposed Sept. 25, 1981). [4] In the 1976 rulemaking NHTSA invoked only the authority of the Cost Savings Act, 41 Fed.Reg. 9346, and safety considerations were not discussed. In issuing the present rule, however, NHTSA recognized that the 1976 rulemaking was conducted "[p]ursuant to both the new authority of the [Cost Savings] Act and that of the Safety Act." 47 Fed.Reg. 21,821 (1982) [5] The dissent raises three points regarding this safety finding, which it asserts render it "difficult to know what the decision was," Dissent at 1377, and make judicial review "impossible," id. It is significant that none of these alleged massive ambiguities was noticed by petitioners during the course of this proceeding, or even raised before us here. That alone, especially in a proceeding of this length and complexity, is reason enough to reject them. See Carducci v. Regan, 714 F.2d 171, 177 (D.C.Cir.1983). Moreover, the reason they escaped notice is that they do not exist (1) The dissent thinks it unclear whether the agency found that the change in bumper standards would have no effect or no significant effect on safety. Dissent at 1374-75. This alleged ambiguity is based solely on the statement in the summary of the rule that "this action will cause no reduction in vehicle safety," 47 Fed.Reg. 21,820. All other references to the agency's safety conclusions--in the rule, see id. at 21,824, 21,827, 21,835-36, in the FRIA incorporated by reference in the rule, see FRIA at IV-3, IV-15, and in the denial of petitions for reconsideration, see 47 Fed.Reg. 56,640, 56,643 (1982)--are couched in terms of no "significant," "discernible," or "measurable" effect. (The latter two adjectives can obviously mean, in addition to capable of perception or measurement--which, as we shall see, was also apt here--significant. See Webster's Third New International Dictionary, defining "measurable" as "great enough to be worth consideration: SIGNIFICANT"; and defining "indiscernible" as "not ... perceptible," and "imperceptible" as "extremely slight." The agency was obviously using the terms in both senses.) Since the summary only purported to be a summary, its statement of "no reduction" rather than "no significant reduction" obviously represents an imprecision rather than an alternate conclusion; and an imprecision of utterly no consequence. (2) The dissent questions whether the agency thought it necessary to consider the relevant Safety Act criteria, or even thought the Safety Act applicable at all. Dissent at 1375-76. However, the very passage seized on by the dissent, id. at 1375 n. 5, made clear that the rulemaking was explicitly conducted under concurrent authority of both the Cost Savings Act and the Safety Act and that the agency found the Safety Act criteria to be satisfied. See 47 Fed.Reg. 21,830. The agency's unwillingness to concede that the specific Safety Act criteria of objectivity, practicability and safety, see 15 U.S.C. § 1392(f), must be applied "in all cases under the [Cost Savings] Act where any safety relationship can be asserted," 47 Fed.Reg. 21,830, quoted by Dissent at 1375 n. 5 (emphasis added), in no way impairs its explicit finding that in this case "the 5.0-mph safety criteria of Part 581 have been determined to be unsupported, even under the Safety Act criteria, and are amended by this notice," 47 Fed.Reg. 21,830. (3) The dissent suggests that we remand to the agency because it did not state that the insignificant safety risk resulting from a reduction in the bumper standard could be prevented only at a cost that made such prevention unreasonable. Dissent at 1376-77. The principle that an "unreasonable risk" provision requires even insignificant risks to be eliminated if that can be done at (presumably) insignificant cost would turn many areas of regulation into unending pursuit of the trivial. The dissent thinks to avoid this conclusion by making a distinction between insignificance of the safety effects of (a) an automobile component as a whole (which it acknowledges can be the basis for a decision not to regulate) and (b) particular characteristics of a safety-significant component, such as the 2.5 vs. 5.0 mph bumper test impact speed (which it asserts cannot be a justification for nonregulation, presumably so long as the cost of the regulation is reasonable in light of the insignificant benefits derived). This distinction hardly eliminates the pursuit of the trivial. It produces the conclusion that NHTSA need not regulate automobile seat covers, but must enter into lengthy and expensive calculations of whether automobile manufacturers should be required to raise the decibel level of automotive horns to degrees that have insignificant effect upon safety. In any case, the cost at issue for purposes of determining whether the insignificant safety risk poses an "unreasonable" risk of injury would be the production cost for the manufacturer--and here no one contends that is insignificant. In sum, the agency's failure to label the risk "unreasonable" instead of "insignificant" scarcely qualifies as imprecision, much less reversible error. See Bowman Transportation, Inc. v. Arkansas-Best Freight System, Inc., 419 U.S. 281, 286, 95 S.Ct. 438, 442, 42 L.Ed.2d 447 (1974). [6] In their Brief in Support of State Farm's and Allstate's Motion in the Alternative for a Limited Remand to Consider New Evidence, Insurance Company Petitioners also called the court's attention to a NHTSA study of side marker lamps, released after the conclusion of the rulemaking proceeding. For reasons discussed in Part V, infra, we do not consider that study 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% [7] If State Farm had expanded its sample of post-MY 1973 statistics to include MYs 1975-78, data for which was included in raw form in its submission to the agency, a quite different picture would have emerged. Table II below shows the results of such an exercise for tail lamp damage. Here, MY 1973 incidence is compared, not with MY 1974, as petitioners had done in their presentation to the agency and in their brief, but with the average of MYs 1974-78 TABLE II Number of Safety Items Damaged Per 100 Estimates Written 2.5 mph 5 mph Percentage rear bumper rear bumper Improvement of standard standard 5 mph over Component MY 1973 Avg: MYs 1974-78 2.5 mph --------- Tail Lamp --------------- Subcompact 10.34 8.12 21% Compact 8.19 8.88 -8% Intermediate 8.64 8.29 4% Full Size 7.49 8.11 -8% Source: The court's own analysis based upon data submitted in Comments of State Farm, Docket No. 73-19-N.27-088 at Table 3 (Dec. 21, 1981), J.A. 1621-23. This modest exercise in data massage is not intended to show, as Table II apparently does show, that the change from Standard 215 2.5 mph to Standard 215 5.0 mph bumpers resulted in insignificant and spurious improvement; nor that courts should engage in such detailed reexamination of technical data before the agency; but only the wisdom of the Supreme Court's injunction that "the generalizability of ... field studies ... is precisely the type of issue which rests within the expertise of NHTSA, and upon which a reviewing court must be most hesitant to intrude." State Farm, 103 S.Ct. at 2872. [8] State Farm also argued that even if damaged safety systems do not cause subsequent accidents, they increase the "risk of death and injury in accidents caused by other factors." Comments of State Farm, Docket No. 73-19-N.27-061 at 7 (Nov. 30, 1981), J.A. 1668. The agency correctly noted that the only safety system protected by the bumper standard that might, if damaged, increase the severity of injury in subsequent accidents is the fuel system. The State Farm data pertaining to fuel system damage predated the separate safety standard for fuel systems, 49 C.F.R. § 571.301 (1983), which "provides protection, independent of and substantially superior to that of the bumper standard, against the risk that fuel leaks will create a safety hazard in an accident." 47 Fed.Reg. 21,827. Petitioners have conceded this point [9] The dissent's courageous attempt, Dissent at 1381 n. 9, to come up with some numerical figure--the figure of 35,000 accidents involving personal injury attributable to failure of the relevant safety systems--succeeds only in demonstrating the impossibility of quantification. This figure not only must be reduced to a presumably massive extent by the accidents preventable by a 5.0 mph bumper that would be prevented by a 2.5 mph bumper as well, as the dissent acknowledges; but it must also be reduced by other factors which the dissent seems to disregard--notably (and again massively) by the unknown percentage of safety-system failures that are caused by prior accidents not protectable by either bumper, or that are not caused by prior accidents at all. We may note in addition that it is not clear that the figure of 20 percent of all accidents constituting injury-producing accidents can properly be applied to the 17,900,000 figure used by the dissent as representing "automobile accidents in this country." Id. The 20 percent figure, derived from the Tri-Level Study, represented the percentage of injury-producing accidents in a universe consisting of only those accidents coming to the researchers' attention "by radio monitoring of police frequencies or by a direct hot-line telephone call from police agencies." Executive Summary at A-2, J.A. 2802. That universe would omit the ordinary "fender-bender," and would be heavily weighted in favor of relatively serious collisions. See Comments of IIHS, Docket No. 73-19-N.25-097, App. A at 4 (Apr. 30, 1979), J.A. 2216 ("Police accident reports generally are not required in a crash unless the amount of damage exceeds some given threshold or a personal injury has occurred, and as a result most low speed crashes are not reported to the police") [10] The dissent's objection that the agency failed to provide evidentiary support for its findings regarding crash energy management, Dissent at 1382, is difficult to fathom. It is clear from the record that the agency's conclusions were based on manufacturer estimates and on the agency's own engineering judgment and analysis, certainly a permissible ground. The dissent also objects to the agency's use of data from 30 mph tests to evaluate crash energy management at speeds below 30 mph, id., but the generalizability of field studies is a question for the agency, not the court. See State Farm, 103 S.Ct. at 2872. Finally, the dissent's focus on the petitioners' "common sense" view that stronger bumpers better promote crash energy management, Dissent at 1383, misses the point. NHTSA did not dispute that 5.0 mph bumpers absorb more crash energy than 2.5 mph bumpers; it simply noted that the contribution of both bumpers to crash energy management at injury-producing speeds was minimal, and that any safety effects of the differential between the bumpers were of no consequence or perhaps more than offset by the more rigid frames that the 5.0 mph bumper required. See 47 Fed.Reg. 56,643 [11] Petitioner CFAS protests that in fact NHTSA analyzed only the cost savings to the auto manufacturers, and that to assume these will be passed through to consumers "overlooks the very elementary fact that automobile manufacturers, like other manufacturers of goods, are in business to make money." Brief for Petitioner CFAS at 35. Of course, far from being overlooked, this consideration clearly underlies the agency's reasoning. Because automobile manufacturers seek to make money, and because their industry is a highly competitive one, it is reasonable to assume that they will view cost savings as an opportunity to compete for profitable business by lowering their prices, or will be forced to follow the lead of those who do. See 47 Fed.Reg. 56,644 ("[t]he agency believes that cost savings will be passed through to consumers as a result of competition"). Cf. A. Smith, The Wealth of Nations 14 (1776) ("[i]t is not from the benevolence of the butcher, the brewer, or the baker that we expect our dinner, but from their regard to their own interest") [12] A number of petitioners and amici, however, assert that the agency applied this methodology to an unduly limited set of alternatives. They claim that the agency failed to give reasoned consideration to the simple expedient of fixing maximum allowable weights for 5.0 mph bumpers. This claim is baseless. NHTSA devoted substantial attention to this alternative, rejecting it on both practical and jurisdictional grounds. The agency reasoned that, for a given level of protection, lighter bumpers have higher material and design costs than heavier bumpers. Hence, any fuel savings resulting from agency-mandated lightweight 5.0 mph bumpers would be offset by increased purchase and replacement costs. NHTSA also doubted, quite correctly, whether a weight requirement would be a "property loss reduction standard" within the definition of a "bumper standard" provided by 15 U.S.C. § 1901(6). See 47 Fed.Reg. 56,642-43 In a related argument, Insurance Company Petitioners allege that, by taking existing bumper designs as given, NHTSA ignored its statutory mandate under the Cost Savings Act to "force" technological innovation. There is no such mandate. "Technology forcing," which prescribes a particular result despite its cost (i.e., the unknown cost of developing and implementing an unknown technology) is fundamentally incompatible with the cost-benefit analysis that the Act requires. [13] Insurance Company Petitioners suggest in passing that the agency's high-range figure of $35 in primary cost savings is unreasonable because AMC and General Motors estimated savings of only $17 and $18, respectively. Brief for Insurance Company Petitioners at 48 n. 54. See FRIA at VII-8, -19. However, Ford and Chrysler both projected savings of $35, id. at VII-8, and the agency's own savings estimate was $49, which is 40 percent higher than the figure questioned by petitioners. Id. at VII-25. Since this was intended to be an upper-range projection rather than a best estimate, it seems to us that the $35 figure was, if anything, too conservative [14] Petitioners and amici insist that NHTSA greatly understated the increases in insurance costs that can be expected from a 2.5 mph standard by ignoring real-world insurance industry evidence that rates are likely to rise 10-20 percent as a result of the new standard. However, these industry calculations were based on reductions in premiums made after the first federal bumper standard was promulgated. Particularly in light of the significant effect of height standardization, discussed above, introduced in the first standard and of course retained under the present revision, NHTSA could reasonably conclude that the move from a 5.0 mph to a 2.5 mph standard is not the converse of the move from no regulation to the original standard. See 47 Fed.Reg. 21,834; FRIA at VI-14 to -17 [15] The dissent asserts that in addition to these factors the agency was obliged to consider under the Cost Savings Act the savings that the more protective bumpers would produce by reducing physical injuries--even if that reduction does not require action under the Safety Act. Dissent at 1384. See 15 U.S.C. § 1912(b)(1)(D). Granting the theoretical point, we cannot agree that the failure to factor insignificant health and safety effects into the agency's cost-benefit analysis constituted a failure "to consider an important aspect of the problem." State Farm, 103 S.Ct. at 2867 (emphasis added). A finding that the agency's action "would not have any significant effect on safety," 47 Fed.Reg. 56,643, renders formal inclusion of that effect in the cost-benefit model an academic exercise, especially when combined with the agency's determination that the effect cannot even be validly quantified. An agency need not address every conceivable issue or alternative, no matter how remote or insignificant. See Vermont Yankee Nuclear Power Corp. v. Natural Resources Defense Council, Inc., 435 U.S. 519, 551, 98 S.Ct. 1197, 1215, 55 L.Ed.2d 460 (1978); Friends of the River v. FERC, 720 F.2d 93, 101-05 (D.C.Cir.1983) [16] NHTSA also had available crash test data relating repair cost to speed, but it chose not to conform its cost-speed curve to these data because of the limited scope of the tests. Final Assessment at F-18, J.A. 1167 [17] The outdated figures used in the 1979 study would not significantly distort the relative effectiveness calculation in Table III but would substantially affect the calculation of total dollar value of damage prevented [18] Table III of course only shows the percentage of lifetime accidents in each speed cell, and not the absolute number of accidents [19] Test data showed the replacement speed of 5.0 mph bumpers to be 10 mph. NHTSA Calculations at B-4, J.A. 1222. Initially, NHTSA estimated the replacement speed for other bumpers by applying to them a comparable ratio of design speed to collision energy at replacement speed, yielding approximately a 7.0 mph replacement speed for the 2.5 mph bumper and a 12 mph replacement speed for a 7.5 mph bumper. Id. IIHS comments on this methodology asserted that the replacement speed for all bumpers should reflect the same energy ratio between replacement speed and design speed shown by the test data for the 5.0 mph bumper--i.e., that the replacement speed for all bumpers should be twice design speed, Comments of IIHS, Docket No. 73-19-N.25-097, App. C at 4-5 (Apr. 30, 1979), J.A. 2240-42; NHTSA ultimately concurred in this analysis. Final Assessment at 29, J.A. 1010 [20] For example, the total consumer cost savings of the 2.5/2.5 system range from $54-$131. The low figure was derived by computing relative cost savings using the lowest value for each of the three components of bumper system costs formulated as ranges. Thus, the $54 figure assumes $18 in primary cost reduction (from a range of $18-$35), $6 in secondary cost reduction (from a range of $6-$20), $28 in fuel savings (from a range of $28-$70), and $2 in finance cost reductions (from a range of $2-$6). The high value was computed using the high values for each of the three components. High and low values were also assigned to the following factors: (1) the economic value of increased delay and inconvenience caused by damage-producing accidents ($26-$50), (2) the number of bumper-related collisions an automobile is likely to experience over its lifetime (2.45-3.20), and (3) the value of unrepaired damage (50-75 percent of the cost of repair). See FRIA at XI-12. The effects of the ranges on the outcome are set forth in Table IV, infra TABLE IV --COST/BENEFIT SUMMARY-- 5 mph and 2.5 mph Bumpers PHASE II BUMPERS (5/5 AS BASELINE) BENEFITS COSTS NET CONSUMER BENEFITS (Reduced (Increased _______Alternatives__Bumper _____ Table Consumer Damage 5/2.5 5/U 2.5/2.5 2.5/U U/U Costs) Costs) ------------------------------------------------------------------------ X-5 High Range Low Range $51 $62 $83 $91 $66 ($35 @ 1.0) (2.45 @ 50%) * X-6 High Range High Range $33 $3 $37 $6 ($82) ($35 @ 1.0) (3.20 @ 75%) ** X-7 Low Range Low Range $14 $33 $6 $22 $3 ($18 @ 0.7) (2.45 @ 50%) * X-8 Low Range High Range ($4) ($26) ($40) ($63) ($145) ($18 @ 0.7) (3.20 @ 75%) ** (*) denote negative net consumer benefits. * $26 Delay and Inconvenience **$50 Delay and Inconvenience ------------------------------------------------------------------------ MEMO: Average of Tables X-5, X-6 $33 $33 $42 $40 ($4) and X-7 Average of all 4 Tables, including the most unlikely combination of extremes (Tables X-8) $23 $18 $22 $14 ($40) Note by the court: The figures in brackets in the second and third columns represent, from left to right, primary production cost reduction, ratio of secondary to primary weight reduction, lifetime frequency of bumper-related accidents, and value of unrepaired damage. Some constituent elements of the cost-benefit calculus--e.g., the range for primary weight reductions--are not shown on the chart. [21] NHTSA also noted that an equal number of sets of extreme assumptions (displayed in Table IV, supra ) supported each of the two alternatives, 47 Fed.Reg. 21,826, and it averaged the results in the first, third, and fourth methods of comparison, producing a $1.33 net benefit advantage for the 2.5/2.5 alternative. See FRIA at XI-18 [22] The effects of the improbability of these assumptions are cumulative. If three (independent) assumptions each has a 10 percent probability of being correct, the probability that all will be correct in combination is 1/10 of 1 percent. By contrast, if two of the assumptions have a 50 percent probability and the third a 10 percent probability of being correct, the probability that all three will be simultaneously true leaps to 2.5 percent, a 25-fold increase [23] The statute reads: If the petitioner applies to the court for leave to adduce additional evidence, and shows to the satisfaction of the court that such additional evidence is material and that there were reasonable grounds for the failure to adduce such evidence in the proceeding before the Secretary, the court may order such additional evidence (and evidence in rebuttal thereof) to be taken before the Secretary, and to be adduced in a hearing, in such manner and upon such terms and conditions as the court may deem proper. The Secretary may modify his findings as to the facts, or make new findings, by reason of the additional evidence so taken, and he shall file such modified or new findings, and his recommendation, if any, for the modification or setting aside of his rule, with the return of such additional evidence. [24] It appears to us that Insurance Company Petitioners' argument that NHTSA's projections of gasoline prices were wrong in retrospect should also be part of this motion, since the evidence is arguably material and could not be adduced at the rulemaking proceeding. Our response to the present arguments thus also applies to the issue of gasoline price forecasts [25] The market incentive for such behavior is suggested by State Farm's allegation in its Brief in Support of State Farm's and Allstate's Motion in the Alternative for a Limited Remand to Consider New Evidence that the United States Government, through the General Services Administration, will continue to demand vehicles with 5.0 mph bumpers in 1984 [1-1] 1 The majority notes that, since its promulgation, there has been considerable discussion over the validity of this standard and implies that there was inadequate support for its promulgation initially. See majority opinion (maj. op.) at 1338-42. I note first that agency discussion of a standard, especially where it has not previously revoked that standard, in no way diminishes the standard of review of an amendment or revocation of that standard. See Motor Vehicle Mfrs Ass'n v. State Farm Mutual Auto. Ins. Co., 463 U.S. 29, 103 S.Ct. 2856, 77 L.Ed.2d 443 (1983) (hereinafter cited as State Farm to 103 S.Ct. only) (applying arbitrary and capricious standard of review even where agency had previously suspended and then reinstated requirement). I note second that the hypothesized invalidity of a standard initially has no bearing on its revocation or amendment. It may make the agency's job easier in showing that revocation or amendment is rational, but it does not alter the standard of review: "If Congress established a presumption from which judicial review should start, that presumption * * * is not against safety regulation, but against changes in current policy that are not justified by the rulemaking record." Id. at 2866 (emphasis in original) [2-2] 2 The Supreme Court noted in State Farm that Congress required compilation of a rulemaking record for the reviewing court, see 15 U.S.C. § 1394, State Farm, supra note 1, 103 S.Ct. at 2867, and intended that agency findings under the Safety Act be supported by "substantial evidence on the record as a whole." See id. at 2867, citing S.Rep. No. 1301, 89th Cong., 2d Sess. 8 (1966); H.R.Rep. No. 1776, 89th Cong., 2d Sess. 21 (1966). In its review of the agency action at issue in State Farm, however, the Court focused on an arbitrary and capricious standard This focus accords with our decision in Pacific Legal Foundation v. Dep't of Transportation, 593 F.2d 1338, 1343 n. 35 (D.C.Cir.), cert. denied, 444 U.S. 830, 100 S.Ct. 57, 62 L.Ed.2d 38 (1979), in which we noted that the distinction between the arbitrary and capricious standard and substantial evidence review is largely semantic * * *. Since our review in this case, under Overton Park, involves a complete examination of the record, we agree with Judge Lumbard that "when an agency engages in substantive rulemaking, it abuses its discretion (or acts arbitrarily and capriciously) if its actions are not supported by substantial evidence." Nat'l Nutritional Foods Ass'n v. Weinberger, 512 F.2d 688, 705 (2d Cir.1975) (Lumbard, J., concurring in the result) [cert. denied, 423 U.S. 827, 96 S.Ct. 44, 46 L.Ed.2d 44 (1975) ]. [3-1] 3 Although my review would indicate reversal on Safety Act grounds alone, I will review the Cost Savings Act analysis because the majority reaches that issue and because I differ from the majority there as well [4-1] 4 The majority attempts to evade the normal and plain meaning of the terms used by the agency by concluding that the agency's summary statement that the rule would have no effect on safety should for some reason be ignored. See maj. op. at 1344-45 n. 5. The majority further attempts to reconcile the agency's various qualified statements ("no measurable or discernible effect") by asserting that they all meant one thing only--that the rule would have no significant safety effects. Unlike the majority, I believe that the variety of terms used by the agency simply corroborates my concern regarding the lack of clarity as to the agency's conclusion. Further, I do not believe that the disparate terms used by the agency can, by intricate dictionary explication, be reasonably made to be synonymous with "significant." [5-1] 5 These paragraphs read: Given the hybrid nature of the part 581 Standard, this rulemaking action was initiated under the concurrent authority of the [Cost Savings] Act and the Safety Act. Without deciding whether the criteria established for safety standards under section 103 [of the Safety Act] must necessarily be applied in all cases under the [Cost Savings] Act where any safety relationship can be asserted, the agency has concluded, based on the discussion in this notice and the FRIA, that its actions in this proceeding are in all respects in accordance with the applicable criteria of the Safety Act itself. By the same token, this action does not conflict with safety standards promulgated under the Safety Act. To the extent that bumper standards may be considered to be safety standards, the 5.0-mph safety criteria of Part 581 have been determined to be unsupported, even under the Safety Act criteria, and are amended by this notice: Reducing the test speed does not make compliance with any safety standard more difficult. The changes made by this rulemaking action do not necessitate any change in efforts to comply with the existing safety standards. * * * [31] 47 Fed.Reg. 21830 (1982), JA 95 [6-1] 6 This point is elucidated by the Supreme Court's analogous treatment of the Occupational Safety and Health Act of 1970. With respect to that Act, the Court held that the agency could regulate workplaces as unsafe only if it first found significant risks. See 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) (plurality opinion); id. at 664-665, 100 S.Ct. at 2875-2876 (Powell, J., concurring in the judgment). This holding is analogous to the idea here that not every vehicle characteristic should be regulated under the Safety Act Later, the Court also interpreted the Occupational Safety and Health Act's standard for setting safe levels for regulated substances. See American Textile Manufacturers Institute v. Donovan, 452 U.S. 490, 101 S.Ct. 2478, 69 L.Ed.2d 185 (1981). In that case the Court did not interject into its discussion an additional or substitute requirement of significance. I believe that the agency in this case should also have confined itself to the statutory standard. [7-1] 7 It is true that petitioners do not raise these concerns specifically. (They do raise the question of what findings the record must support in the sense that their opening brief focuses on the no reduction findings and asserts that safety should be placed first and their reply brief focuses on the no significant reduction findings asserted by the agency in its brief.) I do not believe this should interfere with our consideration of these issues, however. First, to the extent that the agency's lack of clarity prevents meaningful judicial review we have an independent obligation to focus on it. Second, the agency's lack of clarity may actually have caused petitioners to interpret statements in one way and focus their appeal on that interpretation. (For example, the agency's statements can readily be understood to say that the Safety Act does not even apply because no reduction in safety would occur, rather than that the Safety Act is complied with by virtue of that finding.) Where the agency's own actions may have misled those appealing its decision, it should not benefit from the ambiguous nature of those actions [8-1] 8 I do not assert here that the agency's conclusion that the data were inapplicable was wrong. As the majority notes, such decisions normally fall within the scope of the agency's expertise. Rather, my point is that this conclusion was reached without supporting data and in contradiction to common sense. (My common sense indicates that increased protection of safety features would effect increased safety.) The majority notes that NHTSA merely concluded that State Farm's data "falls short of demonstrating that [the difference] is noticeable." Maj. op. at 1346, citing FRIA at IV-9, JA 202. As I note below, see text at p. 1349 infra, State Farm does not have to prove that the agency's conclusions are false. The agency itself must show that those conclusions are reasonable. That the proffered data were not conclusive does not prove their reverse The majority, in its eagerness to discredit the State Farm data, performs a little analysis of its own. See maj. op. at 1347 n. 7. There, it purports to show why deference to the agency is appropriate by noting that an alternative to State Farm's interpretation of those data shows uncertain improvement for one safety component for Model Years 1974-1978 over Model Year 1973. Unfortunately, the majority fails to address how the relative percentages in its table translate into absolute numbers. (It might be that overall safety increased even where safety in one or two categories decreased.) The majority also fails to note that the MY 1974-1978 data submitted by State Farm for two other safety features show consistent and marked improvement in the average over MY 1973. Number of Safety Items Damaged per 100 Estimates Written [35] 2 5 mph 5.0 mph rear bumper rear bumper % improvement standard standard of 5.0 mph Component MY 1973 MY's 1974-1978 over 2(.5) mph -------------------------------------------------------------- Trunk Lid --------- Subcompact 14.51 5.91 59% Compact 11.55 7.27 37% Intermediate 11.64 8.42 28% Full Size 10.97 7.55 31% Gas Tank -------- Subcompact .73 .35 52% Compact .82 .35 57% Intermediate .70 .23 67% Full Size .54 .18 67% [9-1] 9 One way to put the agency's conclusions in perspective is to make some tentative calculations as to the magnitude of safety effects potentially at issue here. As petitioners note, even where the percentages are small the absolute numbers might not be "insignificant." In this context, applying the percentages hypothesized above, we know that in 1980 50,000 people died and 2,850,000 were injured in 17,900,000 automobile accidents in this country. See reply brief for petitioners State Farm Mutual Automobile Insurance Company and Allstate Insurance Companies at 10. Assuming that 1% of these accidents were caused by potentially bumper-related safety feature problems, approximately 179,000 accidents were caused by such problems. Assuming 80% of these were property damage only and that 20% were thus personal injury accidents, perhaps 35,000 of these accidents involved personal injury. Even assuming that this number might be discounted further due to the fact that not all of these safety feature malfunctions resulted from crashes that would have caused negligible damage with 5.0 mph bumpers but significant injury with 2.5 mph bumpers, to totally discount this figure because of those factors--factors on which no data were cited--seems to me to define arbitrary and capricious behavior on the part of the agency [10-1] 10 Even without this specific mention of health and safety costs, I believe that consideration of such costs would be mandated under § 1912(b)(1)(A), the general requirement that the costs and benefits of implementing a standard are to be taken into account [11-1] 11 The majority asserts that insignificant safety effects need not be factored into the Cost Savings Act analysis. See maj. op. at 1355 n. 15. Even assuming, with the agency and the majority, that safety effects were supportably found to be insignificant and thus unimportant under the Safety Act, I do not see that that finding automatically proves their insignificance for Cost Savings Act purposes. (As noted previously, the lack of any guidance as to what the agency considers significant makes this review additionally difficult.) For example, it seems not inconceivable to me that a safety problem could be considered not sufficiently significant to regulate under the Safety Act while the costs of that problem were at the same time significant enough to factor into a Cost Savings Act cost-benefit analysis--especially where, as here, two alternative standards under the latter Act were close under that analysis. See text at p. 1357 infra [12-1] 12 To exemplify this point, if one assumes that the net benefit of Alternative A, if implemented immediately, would be $100 and that the net benefit of Alternative B, if implemented immediately, would be only $75, A will still be preferable to B if they would only be implemented 5 years from now: Assuming a discount rate of 10%, the present value of A would be $62.09 and the present value of B would be $46.57. Thus, although the absolute net benefits have changed because of the discounting, their relation to each other has not [13-1] 13 Taking the alternatives set forth in note 12 supra, if the costs and benefits of B accrue at the same time, 5 years in the future, its net benefit present value will remain $46.57. But if A's costs accrue immediately (say $50) and its benefits accrue in 5 years (say $150), although its non-time-weighted net benefit will remain the same, the present value of its net benefit will be only $43.14, less than the present value of the net benefit of B [14-1] 14 Again taking the alternatives set forth in note 12 supra, if B can be implemented in 5 years and A cannot be implemented until 10 years, B's net benefit present value will again be $46.57, but A's net benefit present value will decrease to $38.55, even assuming its costs and benefits accrue at the same time [15-1] 15 The majority's emphasis on petitioners' failure to raise their asserted objections during the 1979 proceedings would effectively undercut the general rule that when an agency reproposes a rule, even a rule that is already in force, all aspects of its decision to promulgate that rule after reproposing it are subject to comment and, subsequently, judicial review. Failure of a party to have challenged a specific aspect of the rule originally in no way precludes this. See, e.g., State of Montana v. Clark, 749 F.2d 740, 744 (D.C.Cir.1984). There are many reasons, as noted in text, why parties may choose not to address a specific aspect of a complex rulemaking. To conclude, in a later proceeding, that this choice proves the unimportance of that aspect unsupportably limits those parties' right to participate fully in rulemaking proceedings [16-1] 16 Petitioners repeatedly raised the question of the validity of the effectiveness curves by noting that what real world data there are contradict those theoretical conclusions. It is unclear to me that one must raise the precise point below that one raises on appeal; I would assume that raising a basic issue would suffice in many circumstances. Further, as the majority notes, IIHS did object to the assumption of 0% effectiveness at replacement speed when the curves were first drawn. As noted infra, the majority's dismissal of that point is as unconvincing as its conclusion that the curves were not critical to the agency's analysis [17-1] 17 Assumption (1) is not contested [18-1] 18 The $11,000,000 figure comes from multiplying the $1 per car difference by the 11 million estimate of cars per year. See PRIA at VI-25, JA 502