Pub Ctzn Inc v. Natl Hwy Traf Sfty

Notice: This opinion is subject to formal revision before publication in the Federal Reporter or U.S.App.D.C. Reports. Users are requested to notify the Clerk of any formal errors in order that corrections may be made before the bound volumes go to press. United States Court of Appeals FOR THE DISTRICT OF COLUMBIA CIRCUIT Argued May 13, 2004 Decided July 20, 2004 No. 03-1304 PUBLIC CITIZEN, INC. AND CENTER FOR AUTO SAFETY, PETITIONERS v. NATIONAL HIGHWAY TRAFFIC SAFETY ADMINISTRATION AND NORMAN Y. MINETA, SECRETARY OF TRANSPORTATION, RESPONDENTS AUTOMOTIVE OCCUPANT RESTRAINTS COUNCIL AND ALLIANCE OF AUTOMOBILE MANUFACTURERS, INTERVENORS On Petition for Review of an Order of the National Highway Traffic Safety Administration Scott L. Nelson argued the cause for petitioners. With him on the briefs was David C. Vladeck. Michael E. Tankersley entered an appearance. Bills of costs must be filed within 14 days after entry of judgment. The court looks with disfavor upon motions to file bills of costs out of time. 2 H. Thomas Byron III, Attorney, U.S. Department of Jus- tice, argued the cause for respondents. With him on the brief were Peter D. Keisler, Assistant Attorney General; Douglas N. Letter, Attorney; Jeffrey A. Rosen, General Counsel, U.S. Department of Transportation; Paul M. Geier, Assistant General Counsel; Lloyd S. Guerci, Assistant Chief Counsel, National Highway Traffic Safety Administration; and Enid Rubenstein, Attorney. Erika Z. Jones argued the cause for intervenors in support of respondents. With her on the brief were Adam Sloane and David M. Gossett. Shari T. Kendall, Stephen L. Oesch, and Michele M. Fields were on the brief for amicus curiae Insurance Institute for Highway Safety in support of respondents. Before: EDWARDS, RANDOLPH, and TATEL, Circuit Judges. Opinion for the Court filed by Circuit Judge TATEL. TATEL, Circuit Judge: Air bags have saved thousands of lives since first appearing in passenger vehicles approximate- ly three decades ago. Their ability to save some occupants, however, has proven fatal to others, especially children and small women. Responding to this problem and to a new congressional directive, the National Highway Traffic Safety Administration revised one of its auto safety standards to improve air bags’ life-saving benefits while reducing their potentially deadly risks. In this case, we consider a challenge to one aspect of that new standard: the agency’s decision to set the speed for unbelted vehicle crash testing at twenty-five rather than thirty miles per hour. Because in doing so the agency acted consistently with Congress’s directive and nei- ther arbitrarily nor capriciously, we deny the petition for review. I. In 1993, the National Highway Traffic Safety Administra- tion (NHTSA) began requiring manufacturers to install air bags in new cars and light trucks. See Federal Motor Vehicle Safety Standards; Occupant Crash Protection, 58 3 Fed. Reg. 46,551, 46,553 (Sept. 2, 1993) (codified at 49 C.F.R. § 571.208 (2003)). Under NHTSA’s Federal Motor Vehicle Safety Standard No. 208, manufacturers had to certify that their air-bag equipped vehicles would protect occupants in the event of a crash. Specifically, auto makers had to show that their vehicles satisfied certain injury criteria limits in simulat- ed rigid barrier crashes at speeds up to and including thirty miles per hour, using both belted and unbelted fiftieth- percentile adult male dummies. See id. at 46,552; Federal Motor Vehicle Safety Standards; Occupant Crash Protection, 65 Fed. Reg. 30,680, 30,741 (May 12, 2000). In plain English, Standard No. 208 required auto makers to show that a test dummy representing an average-sized man—both with and without a seat belt—would avoid serious injury (as defined by the standard) when his car slams into a fixed barrier at thirty miles per hour. Air bags installed in response to this mandate saved thou- sands of lives. Because air bags are designed to inflate almost instantly upon impact, however, the force of the inflation can injure, even kill, smaller occupants sitting too close to the deploying bag. See Federal Motor Vehicle Safety Standards; Occupant Crash Protection, 62 Fed. Reg. 12,960, 12,960–61 (Mar. 19, 1997). As of February 1997, NHTSA had documented thirty-eight crashes in which the force of a deploying air bag had killed a child. Id. at 12,960. Twenty- one drivers and two adult passengers had also died from air- bag induced injuries. Id. at 12,960–61. Reacting to these fatalities and to a growing public outcry, NHTSA amended Standard No. 208 in March 1997 to encour- age manufacturers to redesign air bags quickly to make them inflate with less force. Id. at 12,961–62. Under the revised rule, manufacturers no longer had to test vehicles using the thirty mile per hour unbelted crash test. Instead, they could use a thirty mile per hour ‘‘sled test’’—a test roughly equiva- lent to a twenty-two mile per hour crash test—to measure vehicle ability to protect fiftieth-percentile male dummies. See id. at 12,974; 65 Fed. Reg. at 30,689. In a sled test, a vehicle placed on a sled is accelerated rapidly backward, but never actually crashed into a barrier. 65 Fed. Reg. at 30,738. 4 As the car moves backward, the test dummies lurch forward, simulating an actual crash; the cars’ air bags are manually deployed at a pre-selected time. Id. By permitting auto makers to certify their vehicles’ crash-protection systems through this less stringent safety test, NHTSA made it easier for them to maintain compliance with Standard No. 208 while ‘‘depowering’’—i.e., reducing the force of—air bag inflation. Although recognizing that depowered air bags could reduce protection for unbelted adults and teenagers, NHTSA con- cluded that ‘‘the opportunity to avoid the deaths of a signifi- cant number of children who would otherwise be fatally injured by air bags’’ justified its depowering rule. 62 Fed. Reg. at 12,964. The agency maintained that ‘‘it is not accept- able that a safety device cause a significant number of fatalities in circumstances in which fatal or serious injuries would not otherwise occur.’’ Id. Indeed, NHTSA found it ‘‘particularly unacceptable that the vehicle occupants being fatally injured are young children[ ] and that the number of those deaths is steadily growing.’’ Id. Nevertheless, be- cause of the possible safety trade-offs associated with bag depowering and because expected technological advances could reduce risks to children and small women without diminishing protection for unbelted adults and teens, NHTSA’s 1997 rule provided that the sled test option would terminate in September 2001. See id. at 12,967–69. The agency explained: [T]here is no need to permanently reduce Standard No. 208’s performance requirements to enable man- ufacturers to fully address the adverse effects of air bags. This is because there are various alternatives, albeit with longer technological development and implementation leadtimes than depowering, that are already allowed by the standard and that appear likely to result in equal or greater benefits than depowering without creating adverse safety trade- offs. Thus, the agency views depowering as an interim approach, while the vehicle manufacturers develop and implement better solutions. 5 Id. at 12,968. Responding to the increased flexibility provid- ed by the sled test option, manufacturers began installing what are known as ‘‘redesigned air bags’’—air bags designed to pass the sled test. See 65 Fed. Reg. at 30,738. Although many redesigned air bags deployed with less force than their predecessors, they still inflated with more power than needed to comply with the sled test. Id.; id. at 30,689. In June 1998, Congress stepped in and directed NHTSA to require manufacturers to install a new generation of air bags known as ‘‘advanced air bags.’’ In language central to the issue before us, the Transportation Equity Act for the 21st Century (TEA 21) provides: [T]he Secretary of Transportation shall issue a no- tice of proposed rulemaking to improve occupant protection for occupants of different sizes, belted and unbelted, under Federal Motor Vehicle Safety Standard No. 208, while minimizing the risk to in- fants, children, and other occupants from injuries and deaths caused by air bags, by means that in- clude advanced air bags. Pub. L. No. 105–178, § 7103(a)(1), 112 Stat. 107, 466 (1998) (codified at 49 U.S.C. § 30127 note (2000)). ‘‘Advanced air bags’’ incorporate new technologies or designs that either prevent air bags from deploying in inappropriate circum- stances or ensure that they inflate in low-risk ways. See 65 Fed. Reg. at 30,738. For example, advanced air bags could include a device that senses the weight of the occupant and then prevents the air bag from activating if the occupant is a child. See id. (defining ‘‘occupant weight sensors’’). Also significant to the challenge before us, TEA 21 super- seded NHTSA’s 1997 decision to sunset the sled test by September 2001, providing instead that the ‘‘requirements of S13 of Standard No. 208 [prescribing the thirty mile per hour unbelted vehicle sled test] shall remain in effect unless and until changed by the rule required by this subsection.’’ TEA 21 § 7103(a)(4). 6 Responding to TEA 21, NHTSA issued a Notice of Pro- posed Rulemaking (NPRM) in September 1998, proposing a battery of new, complex vehicle-safety performance tests. See Federal Motor Vehicle Safety Standards: Occupant Crash Protection, 63 Fed. Reg. 49,958 (Sept. 18, 1998). To satisfy TEA 21’s mandate to ‘‘improve occupant protection,’’ NHTSA proposed phasing out the prevailing sled test and reinstating the thirty mile per hour rigid barrier crash test for the fiftieth-percentile male dummy. See id. at 49,970–71. NHTSA also proposed (among other things) requiring manu- facturers to satisfy the unbelted and belted rigid barrier crash test with a fifth-percentile female dummy—a dummy representing a small woman. Id. at 49,972. With respect to TEA 21’s requirement that the rule minimize risk to infants and children, NHTSA proposed various crash test alterna- tives. See id. at 49,973–75. For example, to protect infants in rear-facing child seats, the NPRM included two alternative series of tests from which manufacturers could choose: tests proving that air bags could deactivate themselves or tests demonstrating that air bags would deploy in a low-risk way. Id. at 49,973. Although vehicle manufacturers initially re- sponded to the proposed rule by supporting retention of the sled test, they later expressed their consensus view that the revised Standard No. 208 should include an unbelted crash test with a maximum speed of twenty-five miles per hour, not thirty. See Federal Motor Vehicle Safety Standards; Occu- pant Crash Protection, 64 Fed. Reg. 60,556, 60,560 (Nov. 5, 1999). In November 1999, NHTSA issued a Supplemental Notice of Proposed Rulemaking (SNPRM), explaining that ‘‘[t]he public comments and the agency research and analysis since our 1998 NPRM have enabled us to refine and in some cases simplify the proposed amendments that we are considering.’’ Id. at 60,557. Of particular significance, NHTSA proposed two alternative unbelted crash tests instead of the single thirty mile per hour rigid barrier test included in its initial NPRM. Only one of these alternatives is relevant here: an unbelted rigid barrier test with a maximum speed ‘‘to be established in the final rule within the range of’’ twenty-five 7 to thirty miles per hour. Id. at 60,558–59. As to this alternative, the agency explained that ‘‘the potential exists for a phase-in sequence in which the maximum speed would initially be set at TTT 25 mph[ ] to provide vehicle manufactur- ers additional flexibility when they are introducing advanced air bags,’’ and then increased to thirty miles per hour ‘‘after a reasonable period of time.’’ Id. at 60,563. Following issu- ance of the SNPRM, several transportation safety organiza- tions, including the National Transportation Safety Board, the Insurance Institute for Highway Safety, the National Safety Council, the American Trauma Society, and the National Association of Governors’ Highway Safety Representatives expressed to NHTSA their ‘‘strong[ ]’’ opposition to returning to the thirty mile per hour standard. See Letter from James E. Hall, Chairman, National Transportation Safety Board, et al., to the Honorable Rodney E. Slater, Secretary, United States Department of Transportation 2 (Feb. 16, 2000). That standard, they explained, ‘‘would result in some vehicles’ airbag inflator power returning to levels that have caused occupant deaths and injuries without increasing the benefits to unbelted occupants in crashes at higher speeds.’’ Id. at 1. NHTSA then submitted a draft final regulation to the Office of Management and Budget for review. See Memoran- dum from Stephen P. Wood, Assistant Chief Counsel for Rulemaking, National Highway Traffic Safety Administration, to Docket No. NHTSA 00–7013; Notice 1 at 1 (May 10, 2000) [hereinafter ‘‘Wood Memorandum’’]; see also Exec. Order No. 12,866 § 6(a)(3)(B)(i), 58 Fed. Reg. 51,735 (Sept. 30, 1993) (requiring agencies to submit certain draft regulations to OMB’s Office of Information and Regulatory Affairs). The draft proposed an unbelted rigid barrier test to be imple- mented in two stages. During the first stage, from Septem- ber 2003 through August 2006, manufacturers would have to test vehicles using a maximum test speed of twenty-five miles per hour, while in the second stage, from September 2007 through August 2010, the maximum test speed would increase to thirty miles per hour. See Wood Memorandum at 1. Two months later, NHTSA published an ‘‘interim final rule,’’ establishing an unbelted rigid barrier crash test with a 8 maximum test speed of twenty-five miles per hour for the period extending from September 2003 through August 2006. See 49 C.F.R. § 571.208 S5.1.2; 65 Fed. Reg. at 30,685. NHTSA explained: The provisions of this rule, particularly the maxi- mum test speed for the unbelted rigid barrier test, reflect the uncertainty associated with simultaneous- ly achieving the twin goals of TEA 21. This uncer- tainty leads us to take an approach that best assures improved air bag protection for occupants of all sizes, without compromising efforts to reduce the risks of injury to vulnerable occupants, including children and short women seated very close to air bags and out-of-position occupants. Such an ap- proach is one that involves the least uncertainty for the occupants who have been most at risk. As long as the manufacturers improve the already substan- tial overall level of real world protection provided by current redesigned air bags, the uncertainty associ- ated with the challenge of simultaneously achieving the twin goals of TEA 21 is best resolved at this point in favor of minimizing risk. This is especially true in the early stages of the introduction of ad- vanced air bag technologies. 65 Fed. Reg. at 30,680. NHTSA’s final rule also added a wide range of new safety performance tests. Among other things, the revised Stan- dard No. 208 requires auto makers to use an entire family of test dummies—not just the fiftieth-percentile adult male dum- my, but also new dummies representing fifth-percentile adult females, six-year-old children, three-year-old children, and one-year-old infants. Id. at 30,685. In addition, the rule includes new and more stringent injury criteria. Id. at 30,691. Eight petitions for reconsideration were filed, including one by Public Citizen and several other consumer groups request- ing that the agency amend the unbelted rigid barrier test to require a thirty mile per hour maximum test speed for 9 passenger cars while retaining a twenty-five mile per hour top speed for light trucks, vans, and sport utility vehicles. NHTSA denied the reconsideration petition, reiterating the rationales offered in the final rule. Federal Motor Vehicle Safety Standards; Occupant Crash Protection, 66 Fed. Reg. 65,376, 65,379–82 (Dec. 18, 2001). Public Citizen and others filed petitions for review in the Ninth Circuit. After addressing various jurisdictional issues, the Ninth Circuit transferred the petition filed by Public Citizen and the Center for Auto Safety to this court. Pub. Citizen Inc. v. Mineta, 343 F.3d 1159, 1171 (9th Cir. 2003). (Throughout this opinion, we shall refer to petitioners Public Citizen and the Center for Auto Safety collectively as ‘‘Public Citizen.’’) In its petition, Public Citizen challenges only one element of the revised Standard No. 208: NHTSA’s decision to set the maximum unbelted crash test speed at twenty-five rather than thirty miles per hour. Public Citizen argues that the twenty-five mile per hour test speed (1) violates TEA 21’s requirement that Standard No. 208 ‘‘improve occupant pro- tection for occupants of different sizes, belted and unbelted,’’ and (2) amounts to arbitrary and capricious agency action. We consider each claim in turn. II. Because Public Citizen challenges NHTSA’s interpretation of a statute the agency is charged with implementing, we apply the two-part test of Chevron U.S.A. Inc. v. Natural Resources Defense Council, Inc., 467 U.S. 837 (1984). We thus begin by asking ‘‘whether Congress has directly spoken to the precise question at issue,’’ for if ‘‘the intent of Congress is clear, that is the end of the matterTTTT [T]he court, as well as the agency, must give effect to the unambiguously expressed intent of Congress.’’ Id. at 842–43. If the statute is ‘‘silent or ambiguous with respect to the specific issue,’’ we then determine whether NHTSA’s interpretation ‘‘is based on a permissible construction of the statute.’’ Id. at 843. Public Citizen does not claim that its statutory challenge can be resolved at Chevron step one, and for good reason. 10 TEA 21’s plain language—requiring NHTSA ‘‘to improve occupant protection for occupants of different sizes, belted and unbelted, under Federal Motor Vehicle Safety Standard No. 208, while minimizing the risk to infants, children, and other occupants from injuries and deaths caused by air bags, by means that include advanced air bags’’—unambiguously requires only three things: Standard No. 208 must (1) ‘‘im- prove occupant protection,’’ (2) ‘‘minimiz[e] the risk to infants, children, and other occupants’’ from air-bag induced injuries, and (3) accomplish these two goals at least partly through the use of ‘‘advanced air bags.’’ Nowhere does TEA 21 say anything about the particular vehicle-testing requirements NHTSA must adopt, much less the speed the agency must use in its unbelted rigid barrier crash test. To resolve Public Citizen’s claim that NHTSA’s twenty-five mile per hour unbelted test speed violates TEA 21’s require- ment to ‘‘improve occupant protection for occupants of differ- ent sizes, belted and unbelted’’—a Chevron step two issue— we must first answer an antecedent question: improve pro- tection compared to what? Specifically, does TEA 21 require improvement measured from the sled test, as NHTSA con- tends, or from the amount of protection provided by rede- signed air bags in use at the time of NHTSA’s rulemaking, as Public Citizen argues? If we can reasonably read TEA 21 to permit NHTSA to use the sled test as the baseline of comparison, then this becomes an easy case, for as Public Citizen concedes, the twenty-five mile per hour crash test improves protection over the sled test. In particular, the twenty-five mile per hour standard tests vehicles at a faster speed than the thirty mile per hour sled test, which has an effective crash test speed of twenty-two miles per hour. In addition, unlike the sled test, the twenty-five mile per hour rigid barrier test measures vehicle performance in actual collisions. See 65 Fed. Reg. at 30,698 (stating that because the sled test does not actually crash the vehicle, ‘‘it cannot measure the performance provided by the vehicle structure in combination with the air bags or even the full air bag system by itself’’). Given these more rigorous testing features, NHTSA estimates that air bags designed to pass the twenty- 11 five mile per hour crash test would save at least sixty-four to 144 more lives than air bags designed to pass the generic sled test. Id. at 30,735. According to Public Citizen, however, the sled test cannot serve as the standard from which to measure the improve- ment TEA 21 requires because NHTSA, in its 1997 rulemak- ing, made the sled test temporary and because Congress, in TEA 21, preserved the sled test only until the agency promul- gates a new crash test standard. Public Citizen argues that in order to satisfy Congress’s directive to improve occupant protection, NHTSA must issue a standard that forces manu- facturers to design air bags to provide better protection than that provided by the redesigned air bags in use at the time of NHTSA’s rulemaking—a showing the agency cannot make, Public Citizen contends, because those air bags already satis- fy the more stringent thirty mile per hour crash test. More- over, Public Citizen claimed at oral argument that NHTSA itself had interpreted TEA 21 as requiring the agency to measure improvement by reference to the level of ‘‘real world protection,’’ not the sled test’s regulatory standard. Counsel stated: [O]n the very first page of the final rule[,] [the agency says] that the occupant protection criterion is satisfied ‘‘as long as the manufacturers improve the already substantial overall level of real-world protec- tion provided by current redesigned air bags.’’ That’s where they on the very first page of their final rule defined the baseline, and that’s the base- line that they don’t satisfy. Tr. of Oral Argument at 55 (emphasis added). Public Citizen’s arguments are unconvincing. Not only does TEA 21 section 7103(a)(1) require NHTSA to improve protection ‘‘under Federal Motor Vehicle Safety Standard No. 208,’’ TEA 21 § 7103(a)(1) (emphasis added), which in- cluded the sled test at the time the statute was enacted, but section 7103(a)(4) expressly authorizes NHTSA to retain the sled test: the ‘‘requirements of S13 of Standard No. 208 12 [prescribing the vehicle sled test] shall remain in effect unless and until changed by the rule required by this subsection,’’ id. § 7103(a)(4) (emphasis added). As the congressional confer- ees explained, this provision ensures that ‘‘[t]he availability of the current sled test certification option available under [Fed- eral Motor Vehicle Safety Standard] 208 (S13) remains in effect unless and until phased out according to the schedule in the final rule.’’ H.R. CONF. REP. NO. 105–550, at 521 (1998), reprinted in 1998 U.S.C.C.A.N. 70, 196. Because Congress directed NHTSA to keep the sled test ‘‘unless’’ otherwise phased out, NHTSA’s use of that test as the baseline of comparison can hardly be unreasonable. See Chevron, 467 U.S. at 843. We also disagree with Public Citizen that NHTSA acted inconsistently with its own interpretation of TEA 21. NHTSA did not, as counsel claimed at oral argument, inter- pret TEA 21 as requiring improvement to be measured by reference to redesigned air bags actually installed in vehicles. Under the heading, ‘‘Rationales for Protection Improvement Requirements TTT Selection of TTT 25 mph[ ] as [the] Top Speed for [the] Unbelted Rigid Barrier Test,’’ 65 Fed. Reg. at 30,686, NHTSA maintained that its ‘‘decision to replace the TTT 30 mph[ ] generic sled test with the TTT 25 mph[ ] unbelted rigid barrier test requires a significantly higher level of safety [because] the sled test is roughly equivalent to a TTT 22 mph[ ] rigid barrier TTT crash,’’ id. at 30,689. Moreover, responding to comments that a twenty-five mile per hour standard would violate TEA 21’s protection-improvement mandate, NHTSA explained: We also note that the suggestion that TEA 21 somehow requires an unbelted barrier test with a test speed not lower than TTT 30 mph[ ] is inconsis- tent with the language of that statute. In fact, TEA 21 expressly left open the possibility of our retaining the sled test. That test has a severity level signifi- cantly below that of a TTT 30 mph[ ] barrier test and a TTT 25 mph[ ] barrier test. 13 Id. at 30,705 (emphasis added). And despite Public Citizen’s claim to the contrary, NHTSA never said that ‘‘the occupant protection criterion is satisfied ‘as long as the manufacturers improve the already substantial overall level of real-world protection provided by current redesigned air bags.’’’ Tr. of Oral Argument at 55. NHTSA actually said this: ‘‘As long as the manufacturers improve the already substantial overall level of real world protection provided by current redesigned air bags, the uncertainty associated with the challenge of simultaneously achieving the twin goals of TEA 21 is best resolved at this point in favor of minimizing risk.’’ 65 Fed. Reg. at 30,680. Read in its entirety and in the context of NHTSA’s clear explanation of its view of TEA 21, this sentence cannot plausibly be read to mean that the agency viewed the ‘‘level of real world protection’’ as the statutorily required baseline of comparison. III. We turn next to Public Citizen’s arbitrary-and-capricious challenge. Under the Administrative Procedure Act, we will ‘‘hold unlawful and set aside agency action, findings, and conclusions found to be TTT arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law.’’ 5 U.S.C. § 706(2)(A) (2000). As the Supreme Court explained in its seminal decision in Motor Vehicle Manufacturers Ass’n of the United States, Inc. v. State Farm Mutual Automobile Insurance Co., 463 U.S. 29 (1983), a case that also involved air bags and Standard No. 208: The scope of review under the ‘‘arbitrary and capri- cious’’ standard is narrow and a court is not to substitute its judgment for that of the agency. Nev- ertheless, the agency must examine the relevant data and articulate a satisfactory explanation for its action including a rational connection between the facts found and the choice made. In reviewing that explanation, we must consider whether the decision was based on a consideration of the relevant factors and whether there has been a clear error of judg- 14 ment. Normally, an agency rule would be arbitrary and capricious 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 exper- tise. Id. at 43 (citations and internal quotation marks omitted). With this highly deferential standard in mind, we consider each of Public Citizen’s specific claims. Public Citizen contends first that nothing in the record supports NHTSA’s conclusion that the twenty-five mile per hour unbelted test speed is ‘‘in the best overall interest of safety.’’ 65 Fed. Reg. at 30,687. According to Public Citizen, unless NHTSA found that ‘‘retaining the 30 mph test speed could result in a new risk of substantial fatalities or serious injuries that would outweigh the hundreds of fatalities among teenagers and adults TTT that might flow from reducing the 30 mph test to [a] 25 mph standard,’’ NHTSA could not rationally conclude that a twenty-five mile per hour standard best serves the overall interests of safety. Pet’rs’ Br. at 35– 36. Because NHTSA failed to make such a finding, Public Citizen concludes, the agency violated the APA standard of reasoned decisionmaking. We disagree. To begin with, the factual premise underlying Public Citi- zen’s claim—that establishing a twenty-five mile per hour unbelted test speed will result in hundreds of new fatalities among unbelted teens and adults—is unsupported by the record, for it rests on the assumption, rejected by NHTSA, that manufacturers will depower air bags to the minimum level permitted by the twenty-five mile per hour standard. NHTSA explained that auto makers (1) have no economic incentive to depower air bags because doing so would achieve no significant cost savings, (2) did not depower their bags to the bare minimum permitted by the sled test, see supra p. 5, and (3) have indicated that, due to other regulatory require- 15 ments, they will not do so now. 65 Fed. Reg. at 30,689; see also id. at 30,704 (stating that NHTSA’s Final Economic Analysis shows that air bags certified under a twenty-five mile per hour standard cost essentially the same as those certified to a thirty mile per hour standard). Given these findings, NHTSA’s prediction that manufacturers will not likely reduce the protectiveness of current air bags is reason- able. ‘‘Predictions regarding the actions of regulated entities are precisely the type of policy judgments that courts routine- ly and quite correctly leave to administrative agencies.’’ Pub. Utils. Comm’n v. FERC, 24 F.3d 275, 281 (D.C. Cir. 1994). Moreover, NHTSA did not, as Public Citizen claims, rely on manufacturers’ ‘‘voluntary action’’ to satisfy TEA 21’s protec- tion-improvement requirement. Pet’rs’ Br. at 41. As we have just explained, NHTSA’s revised Standard No. 208 improves occupant protection by requiring auto makers to satisfy a performance test (the twenty-five mile per hour crash test) that is more stringent than the agency’s preexist- ing test (the sled test). Public Citizen cites Public Citizen v. Nuclear Regulatory Commission, 901 F.2d 147 (D.C. Cir. 1990), but there, interpreting a very different statute, we held that an agency ordered by Congress to promulgate binding regulatory requirements may not issue a non-binding policy statement that encourages but does not compel action. Id. at 157. Here, by contrast, NHTSA’s revised Standard No. 208 imposes mandatory testing standards. In other words, man- ufacturers must meet NHTSA’s twenty-five mile per hour unbelted crash test, and NHTSA’s conclusion that manufac- turers are unlikely to depower to the minimum permitted by the twenty-five mile per hour standard does not transform that mandatory standard into an optional one. Public Citizen’s challenge to NHTSA’s conclusion that the revised Standard No. 208 serves the best overall interest of safety fails for a second reason: NHTSA explained, reason- ably in our view, why a twenty-five mile per hour unbelted test speed, considered in the context of the entire rule, serves the agency’s overall safety goals. Given the complex array of new requirements imposed by the rule—including additional crash test dummies, revised injury criteria, and new risk- 16 minimization tests—NHTSA was ‘‘concerned about the diffi- culties of trying to meet the unbelted rigid barrier test at TTT 30 mph[ ] with both adult dummies while simultaneously trying to reduce the risks of air bag-induced injuries and deaths.’’ 65 Fed. Reg. at 30,687. And because air bags, unlike other safety devices, can themselves injure or even kill vehicle occupants, NHTSA thought it should ‘‘be cautious in how far and how fast vehicle manufacturers are required to advance the state of advanced air bag technologies in their vehicles.’’ Id. NHTSA continued: Since a significant percentage of current vehicles can already satisfy the new unbelted barrier crash test at TTT 25 mph[ ] with both the 5th percentile adult female dummy and the 50th percentile adult male dummy, we conclude that setting the maximum speed at that level will help vehicle manufacturers to focus their resources and compliance efforts TTT on meeting the risk reduction requirements. Id. Giving manufacturers the ability to so focus their efforts, NHTSA maintained, would help ensure that ‘‘the installation of advanced air bag technologies by the vehicle manufacturers across the full spectrum of their fleets [is] done correctly— the first timeTTTT Compared with a TTT 30 mph[ ] unbelted rigid barrier test, a TTT 25 mph[ ] unbelted rigid barrier test presents less chance of inadvertently increasing risks to out- of-position occupants.’’ Id. at 30,688. In view of NHTSA’s twin statutory obligations—to improve occupant protection while minimizing the risks of air-bag induced injuries—and our highly deferential standard of review, we have no basis for second-guessing the agency’s safety assessment. Public Citizen next argues that ‘‘NHTSA TTT acted unrea- sonably by reducing the maximum test speed based on ‘uncer- tainties’ about how air bag improvements will be implemented even though it has no evidence that these uncertainties warrant such action.’’ Pet’rs’ Br. at 25. Again, we disagree. NHTSA did not, as Public Citizen suggests, ‘‘merely recite the terms ‘substantial uncertainty’ as a justification for its actions.’’ State Farm, 463 U.S. at 52. Instead, just as the 17 Supreme Court required in State Farm, ‘‘[t]he agency TTT explain[ed] the evidence which is available, and TTT offer[ed] a rational connection between the facts found and the choice made.’’ Id. (quoting Burlington Truck Lines, Inc. v. United States, 371 U.S. 156, 168 (1962)) (internal quotation marks omitted). Specifically, NHTSA stated that although it be- lieved that vehicles could eventually pass a thirty mile per hour unbelted rigid barrier crash test without compromising safety for children and small women, ‘‘[o]ur laboratory tests and knowledge of advanced technologies do not tell us how or when developments might reach that point. They also do not provide us with a full picture of the real world consequences of adopting that test speed.’’ 65 Fed. Reg. at 30,688. In addition, with respect to certain vehicles such as light trucks and vans, NHTSA stated that at the time it issued the final rule, its experience testing such vehicles with fifth-percentile female dummies was limited. Id. at 30,689. The record also shows that NHTSA had cause for concern over how manufac- turers could simultaneously achieve TEA 21’s protection- improvement and risk-minimization goals: in the agency’s thirty mile per hour unbelted crash tests, over half the vehicles failed to meet the new injury criteria for small female occupants, one of the groups most at risk of air-bag induced injuries. See National Highway Traffic Safety Administra- tion, Final Economic Assessment, FMVSS No. 208 Advanced Air Bags at E–3, IV–32 to IV–33 (May 2000). And despite Public Citizen’s claim to the contrary, NHTSA did consider alternatives—including the thirty mile per hour crash test standard that Public Citizen favors—and coherently explained why the twenty-five mile per hour standard better accommo- dated its uncertainty about future technological develop- ments. See 65 Fed. Reg. at 30,686–89. Public Citizen calls our attention to Public Citizen v. Steed, 733 F.2d 93 (D.C. Cir. 1984), in which we granted a petition for review challenging NHTSA’s decision to suspend a tire- testing regulation. In that case, however, NHTSA was ‘‘un- certain[ ]’’ whether the evidence on which it had relied to suspend the regulation truly supported its ultimate decision. Id. at 101 (‘‘As for the evidence indicating flaws in the test 18 procedures themselves, NHTSA’s statement explaining the suspension [of the rule] reflects considerable uncertainty con- cerning the extent to which those flaws lead to significant inaccuracies in the test results.’’). NHTSA had no such uncertainty here. It merely explained, based on entirely rational reasons, its inability to forecast the pace or nature of technological change; NHTSA had no uncertainty about that uncertainty. For its final APA argument, Public Citizen contends that NHTSA acted arbitrarily and capriciously by failing to ex- plain cogently why it declined to increase the unbelted test speed to thirty miles per hour after the year 2006. This claim fails because NHTSA offered rational reasons for adopting an ‘‘interim final rule’’ establishing the unbelted crash test speed through August 2006 only. The agency explained that because it could not ‘‘assess whether the uncertainty about the manufacturers’ ability to improve pro- tection further and minimize risk simultaneously will persist’’ into the future, it would leave open the question of the post- 2006 unbelted crash test speed, using that time instead to undertake ‘‘a multi-year effort to obtain additional data.’’ 65 Fed. Reg. at 30,685. ‘‘Based on the results of those informa- tion gathering and analysis efforts’’ and public input, NHTSA said that it would then make a final decision regarding the maximum test speed for unbelted dummy testing in the long run. Id. We see no defect in this explanation, for nothing in the APA precludes an agency from collecting data and moni- toring real-world experience with regulatory standards before adopting new standards governing periods of time far into the future—especially in cases, as here, that involve unpredict- able technological change. Indeed, gathering evidence before making a long-term decision is eminently sensible. See Nat’l Ass’n of Broadcasters v. FCC, 740 F.2d 1190, 1211 (D.C. Cir. 1984) (holding that the FCC acted reasonably in postponing a decision on certain details of a spectrum allocation rule because ‘‘the Commission acted against an evolving back- ground,’’ certain relevant factors could not ‘‘be known at the 19 present,’’ and ‘‘[a]s a result, too fine a calibration of the relocation [of spectrum space] would have been premature’’). At bottom, Public Citizen’s arbitrary-and-capricious chal- lenge boils down to a policy disagreement with NHTSA. Public Citizen believes that NHTSA should have set the unbelted test speed at thirty miles per hour. Perhaps the record could have supported such a standard. But because NHTSA’s selection of twenty-five miles per hour is both supported by the record and rationally explained, we have no basis for substituting Public Citizen’s views for the agency’s, particularly given NHTSA’s judgment that doing so would increase the risk of harm to children and small women. See State Farm, 463 U.S. at 43. IV. The petition for review is denied. So ordered.