Opinion for the Court filed by Circuit Judge WILLIAMS.
Opinion dissenting in part filed by Chief Judge MIKVA.
STEPHEN F. WILLIAMS, Circuit Judge:Choice means giving something up. In deciding whether to relax the previously established “corporate average fuel economy” (“CAFE”) standard for model year 1990, the National Highway Traffic Safety Administration (“NHTSA”) confronted a record suggesting that refusal to do so would exact some penalty in auto safety. Rather than affirmatively choosing extra energy savings over extra safety, however, NHTSA obscured the safety problem, and thus its need to choose. Because NHTSA failed to reason through to its decision, see Greater Boston Television Corp. v. FCC, 444 F.2d 841, 850-52 (D.C.Cir.1970), we remand the case for further consideration.
* * # * # #
The Energy Policy and Conservation Act, Pub.L. No. 94-163, 89 Stat. 871, codified at 15 U.S.C. § 2001 et seq. (1988), requires every major carmaker to keep the average fuel economy of its fleet, in each model year, at or above a prescribed level. The Act holds manufacturers to a standard of 27.5 miles per gallon for model year 1985 and each model year thereafter, but authorizes NHTSA to modify the standard, up or down. Where the agency chooses to modify, it must set the replacement standard at the “maximum feasible average fuel economy level”. 15 U.S.C. § 2002(a)(4). In determining “feasibility”, NHTSA has always taken passenger safety into account, see Competitive Enterprise Inst. v. NHTSA, 901 F.2d 107, 120 n. 11 (D.C.Cir.1990) (“CEI F’), and the agency maintains that safety concerns are relevant to whether the agency should adopt one CAFE standard over another.
*323In August 1988, at the behest of various parties, including several major carmakers and petitioner Competitive Enterprise Institute (“CEI”), NHTSA initiated a rulemak-ing proceeding on whether to reduce the CAFE standards for model years 1989 and 1990. See 53 Fed.Reg. 33080. The agency quickly lowered the standard for model year 1989 to 26.5 mpg, but it continued to hear public comment on whether to reduce the 1990 standard as well. Then, in May 1989, NHTSA terminated its proceedings on that issue and left the statutory standard in place.
While the agency rejected a variety of attacks on that standard, we are concerned with only one of the defeated arguments: the contention that the standard will force carmakers to produce smaller, less safe cars, thus making it more difficult and expensive for consumers to buy larger, safer cars. We find that the agency has not coherently addressed this concern.
* * * * * *
As a threshold matter, NHTSA claims that neither of the petitioners has standing to bring this action. In all respects but one, NHTSA’s arguments rehash those rejected in CEI I, and we need only refer it to that decision. 901 F.2d at 111-19. The only novel claim arises from our decision in General Motors Corp. v. NHTSA, 898 F.2d 165 (D.C.Cir.1990), issued a few weeks after CEI I, that NHTSA may properly refuse to entertain a petition, filed long after the beginning of a model year, to amend the standard for that year. From this NHTSA leaps to the view that this court cannot redress any wrong to Consumer Alert’s members. But nothing in General Motors undermines our power to order retroactive reconsideration of a standard where necessary to “grant appropriate relief” to aggrieved parties under 15 U.S.C. § 2004(a). See City of Los Angeles v. NHTSA, 912 F.2d 478, 485 (D.C.Cir.1990); CEI I, 901 F.2d at 117-18.
* * He * * *
On the merits, we must determine whether NHTSA has offered a reasoned explanation for terminating its inquiry into whether to relax the 27.5 standard. The statute does not specify criteria for when NHTSA must persevere in a rulemaking it has already initiated, but the agency does not dispute that there is “law to apply”. The parties do disagree on the appropriate degree of deference owed to an agency’s decision to terminate a rulemaking proceeding. Cf. Williams Natural Gas Co. v. FERC, 872 F.2d 438, 443-44 (D.C.Cir.1989) (suggesting that more deference than usual is owed in such contexts). But the exact degree is unimportant. Petitioners’ essential claim is an attack not on NHTSA’s judgment call, but on NHTSA’s attempt to paper over the need to make a call. We cannot defer to mere decisional evasion. Cf. SEC v. Chenery Corp., 332 U.S. 194, 196-97, 67 S.Ct. 1575, 1577-78, 91 L.Ed. 1995 (1947).
When the automaking firms petitioned for a reduction in the model year 1990 standard down to 26.5 mpg [see J.A. 60], and petitioners pressed the argument that failure to reduce the standard would cost lives, NHTSA had three basic choices. First, it might have concluded that the statute does not require it to consider safety effects when deciding whether to embark on a modification proceeding. It could then have dismissed petitioners’ claims without further ado. While a court might have reversed, the statutory framework is so loose and deference under Chevron U.S.A. v. NRDC, 467 U.S. 837, 104 S.Ct. 2778, 81 L.Ed.2d 694 (1984), so broad that the agency would have had a fair shot at being upheld.
Second, NHTSA might have seriously examined the record data. On its face this suggested (as we shall see) the overwhelming likelihood that a 27.5 mpg standard reduces the supply of safe cars available to American consumers. Conceivably, of course, a sophisticated analysis might have overcome the record’s apparent implications, but even if it did not, all NHTSA would have had to do was face the tradeoff. It could have said that while the 27.5 standard might cost, say, 200-to-500 American lives a year for ten years, it would also reduce American oil imports by, say, 50,-000-to-400,000 barrels a year, and that in *324its judgment the trade-off was worth it. And it could have expressed any such trade-off in less numerical terms.
Finally, NHTSA could have fudged the analysis, held the standard at 27.5, and, with the help of statistical legerdemain, made conclusory assertions that its decision had no safety cost at all. That is what it chose. The people petitioners represent, consumers who do not want to be priced out of the market for larger, safer cars, deserve better from their government.
We must remand this case to NHTSA if the agency has not adequately explained why one of the following is false: (1) adopting a 27.5 standard (as opposed to a lower standard) will have some constraining effect on carmakers; (2) carmakers will, as one consequence of the standard, decrease the average size of their cars below what it would have been absent the standard; (3) this decrease will make it more difficult for consumers to drive large cars; and (4) all other things being equal, a large car is safer than a small car. The agency actually admits the truth of the fourth proposition (see p. 326 below), and we can find no passage in the record where the agency has coherently explained the falsehood of any of the others.
Constraining automakers
As the agency conceded at oral argument, the 27.5 mpg standard obviously affects carmakers’ behavior — if not in model year 1990, at least in subsequent years. Under the statute, if a carmaker exceeds the applicable CAFE standard in one year, it earns credits that it may use to offset CAFE deficiencies over the next three years. See 15 U.S.C. § 2002(i). At the very least, keeping the 1990 standard at 27.5 mpg reduces the number of carryover credits that GM can use to blunt the effect of the CAFE standards for model years 1991-93. Accordingly, NHTSA expressed its quite reasonable belief that “the potential actual [sic] impacts on energy conservation [from retention of the 27.5 mpg standard for model year 1990] are largely related to multi-year considerations.” NHTSA, Passenger Automobile Average Fuel Economy Standards for Model Year 1990, Termination of Rulemaking, 54 Fed.Reg. 21985, 21994/3 (1989). In fact, NHTSA recently declared that it would be unlawful for it to set “CAFE standards deliberately low enough to be ‘noncon-straining' ”. NHTSA, Passenger Automobile Average Fuel Economy Standards for Model Years 1989 and 1990, Notice of Proposed Rulemaking, 53 Fed.Reg. 33080, 33094/3 (1988). It seems obvious, then, that the 27.5 mpg standard is constraining in one way or another.
Automakers’ likely choice to downsize
Second, the agency insisted at oral argument that even if the 27.5 standard constrains the behavior of carmakers, it will not lead to smaller cars. Yet nowhere has the agency actually justified this claim or even purported to make such a finding. It came closest in the following passage:
[T]here are still a number of fuel-efficiency enhancing methods that [GM and Ford have] not fully utilized throughout their fleets.... NHTSA believes that the domestic manufacturers should be able to improve their fuel economy in the future by these and/or other technological means, without outsourcing their larger cars, without further downsizing or mix shifts toward smaller cars, and without sacrificing acceleration or performance.
54 Fed.Reg. at 21996/2-3 (emphasis added). Why the agency expressed itself in the normative (“should be”) is anybody’s guess. At any rate, it has never claimed that domestic manufacturers will in fact meet the standard without downsizing their fleets, or even that there is a substantial probability that they will do so, or even that there is a substantial likelihood that they will use methods other than downsizing for the lion’s share of the work. Presumably NHTSA does not assert such facts because it could not ground them in the record.
Moreover, to the extent that carmakers choose technological innovation over downsizing (and further assuming that such innovation would not itself compromise aspects of auto safety), that choice would involve significant costs in implementation, *325even if we assume that research and development are complete. That cost would translate into higher prices for large cars (as well as small), thereby pressuring consumers to retain their old cars and make the associated sacrifice in safety (see pp. 325-26 below). The result would be effectively the same harm that concerns petitioners and that the agency fails to negate or justify.
The historical fact is, however, that car-makers respond to CAFE standards by reducing the size of their fleets. NHTSA itself has explicitly acknowledged as much in the past, see, e.g., NHTSA, Passenger Automobile Average Fuel Economy Standards for Model Year 1989: Final Rule, 53 Fed.Reg. 39278/1 (Oct. 6, 1988); MY 1987-88 Environmental Assessment 19-20, and we ourselves have insisted that “the evidence shows that manufacturers are likely to respond to lower CAFE standards by continuing or expanding production of larger, heavier vehicles,” CEI I, 901 F.2d at 117. Even in the decision below the agency acknowledged this link, explaining that “Chrysler’s CAFE has been higher than that of GM or Ford in recent years primarily because it does not compete, or compete as heavily, in all the market segments in which GM and Ford sell cars, particularly the large car market.” 54 Fed.Reg. at 21991/2-3.
The agency now tries to obscure this reality by pointing out that “the average fuel economy of the new car fleet has improved steadily from 26.6 mpg in model year 1982 to 28.2 mpg in model year 1987, while the average weight of a new car increased two pounds during the same period.” NHTSA Br. at 13, citing 54 Fed.Reg. at 21993/1. This argument misses the point. The appropriate comparison, which NHTSA must but did not address, is between the world with more stringent CAFE standards and the world with less stringent standards. The fact that weight has remained constant over time despite mileage improvements shows the effect of technological improvements, to be sure, but in no way undermines the natural inference that weight is lower than it would be absent CAFE regulation. Here we can be quite sure that it is lower, since, as NHTSA observed in this decision, economic recovery and declining gasoline prices sharply raised consumer demand for large cars over the relevant period. 54 Fed.Reg. at 21987/1; see also CEI I, 901 F.2d at 117 (“consumer demand has shifted back toward larger vehicles”). If consumers demanded substantially bigger cars, carmak-ers — absent regulation — would have produced substantially bigger cars, not cars that remained, on average, within two pounds of the cars made when consumers favored smaller cars. Moreover, NHTSA has given us no reason to think that whatever technological innovations permitted automakers to meet CAFE requirements while keeping weight constant did not also cost consumers more, again pricing some consumers out of the market for new large cars.1
Effect on consumer access to large cars
NHTSA also argues that even if the 27.5 mpg standard will deplete the supply of large GM or Ford cars, a consumer looking for a big car “will buy a large car from another manufacturer, or will buy a minivan, or will keep his or her older, large car.... [A]ny one of those alternative consumer outcomes is far more likely than the possibility that the consumer will buy a smaller car than he or she wanted to buy.” 54 Fed.Reg at 21993/2. Nothing in the record suggests that any of these will give consumers large-car safety at the prices that would have prevailed if NHTSA had made a less stringent choice.
The reference to buying large cars from “another manufacturer” is somewhat in the spirit of Marie Antoinette’s suggestion to “let them eat cake.” By NHTSA’s own hypothesis, the “other manufacturers” are Chrysler, which has essentially 'removed itself from the large car market, id. at *32621991/2-3, and foreign manufacturers, which are subject to CAFE standards on their U.S. sales, see 15 U.S.C. § 2002(a)(1). To the limited extent that foreign firms produce truly large cars at all, they are expensive ones. See generally Highway Loss Data Institute, Injury and Collision Loss Experience, Cars by Make and Model (September 1988), reprinted at J.A. 239.
In suggesting minivans (which are exempt from the 27.5 standard), the agency disingenuously obscures their dangers by citing safety figures only for vans in general. 54 Fed.Reg at 21993/2-3. As NHTSA itself has amply documented, however, minivans are considerably less safe than vans generally, with a fatality rate per registered vehicle about 25-33% higher than that of large cars. NHTSA, Safety Programs for Light Trucks and Multipurpose Passenger Vehicles (April 1988), at 4, reprinted at J.A. 94. Finally, NHTSA’s notion that the consumer should “keep his or her older, large car” ignores both its own finding that new cars “appear to experience fewer accidents per mile traveled,” NHTSA, Final Rule: Federal Motor Vehicle Safety Standards, 42 Fed.Reg. 34289, 34292/3 (1977), and the plight of consumers seeking to buy a large car for the first time.
Impact on safety
By making it harder for consumers to buy large cars, the 27.5 mpg standard will increase traffic fatalities if, as a general matter, small cars are less safe than big ones. They are, as NHTSA itself acknowledges. See 54 Fed.Reg at 21993/3-21994/1. The agency explains:
Occupants of the smaller ears generally are at greater risk because: (a) the occupant’s survival space is generally less in small cars (survival space, in simple terms, means enough room for the occupant to be held by the vehicle’s occupant restraint system without being smashed into injurious surfaces, and enough room to prevent being crushed or hit by a collapsing surface); (b) smaller and lighter vehicles generally have less physical structure available to absorb and manage crash energy and forces; and (c) in most collisions between vehicles of different weight, the forces imposed on occupants of lighter ears are proportionately greater than the forces felt by occupants of heavier vehicles.
NHTSA, Small Car Safety in the 1980’s at 64 (1980), reprinted at J.A. 85.2
The agency tries to skirt the obvious conclusion with two specious arguments. First, it essentially argues that the 27.5 mpg standard will have no effect on the availability of large cars (i.e., will accomplish nothing at all). This, we have seen, is simply untrue. Second, the agency observes that new cars now come with a variety of mandatory and optional safety features (airbags, anti-lock brakes, etc.) that will presumably compensate for a decline in size. Id. at 21994/1-2.
There are two things wrong with this latter argument. First, so far as we can tell, the agency nowhere claims that these safety innovations fully or even mostly compensate for the safety dangers associated with downsizing. More critically, as in the relation between fuel economy and downsizing, the relevant inquiry is whether stringent CAFE standards reduce auto safety below what it would be absent such *327standards. That new safety devices may be coming on the market is all well and good, but it is immaterial to our inquiry unless the implementation of those devices somehow depends on or is caused by more stringent CAFE standards; no one even hints at such a link. Whatever extra safety devices may contribute to either type, small cars remain more dangerous than large ones, all other things being equal.3
******
Nothing in the record or in NHTSA’s analysis appears to undermine the inference that the 27.5 mpg standard kills people, although, as we observed before, we cannot rule out the possibility that NHTSA might support a contrary finding. Assuming it cannot, the number of people sacrificed is uncertain. Forced to confront the issue, the agency might arrive at an estimate lower than that of two independent analysts who came up with an annual death rate running into the thousands (for the cars produced in any one model year). See Robert W. Crandall & John D. Graham, The Effect of Fuel Economy Standards on Automobile Safety, 32 J.L. & Econ. 97 (April 1989). Yet the actual number is irrelevant for our purposes. Even if the 27.5 mpg standard for model year 1990 kills “only” several dozen people a year, NHTSA must exercise its discretion; that means conducting a serious analysis of the data and deciding whether the associated fuel savings are worth the lives lost.
When the government regulates in a way that prices many of its citizens out of access to large-car safety, it owes them reasonable candor. If it provides that, the affected citizens at least know that the government has faced up to the meaning of its choice. The requirement of reasoned decisionmaking ensures this result and prevents officials from cowering behind bureaucratic mumbo-jumbo. Accordingly, we order NHTSA to reconsider the matter and provide a genuine explanation for whatever choice it ultimately makes.
So ordered.
. It is significant that even NHTSA makes no more than the lame claim that ‘‘[t]his example illustrates the point that not all CAFE gains come by reducing weight.” 54 Fed.Reg. at 21993/1. The issue is whether a material portion of the “CAFE gains" are likely to entail downsizing. NHTSA never even purports to deny this.
. One might argue that the third factor indicates that if all cars were small, there would be fewer traffic fatalities. Any such inference appears quite doubtful. Cars can hit a variety of objects, including trucks, trees, and other cars; fatalities in car-to-car crashes do not account for even a majority of passenger-car occupant fatalities. See Failure Analysis Associates, The Impact of "Downsizing" the American Automobile Fleet on Overall Motor Vehicle Safety (August 1986), at 38, reprinted at J.A. 180. Unless NHTSA outlaws trucks and trees, smaller cars will probably always mean higher fatality rates, as NHTSA recognizes. See Small Car Safety in the I980’s at 59, reprinted in J.A. at 80 ("in single vehicle crashes, there is increased risk of serious injury or death”); see also The Impact of “Downsizing" at 6, reprinted at J.A. 176; Insurance Institute for Highway Safety, Status Report, Dec. 30, 1982, at 5, reprinted at J.A. 223. Moreover, while the record is not clear on the matter, it appears that the chance of fatality in crashes involving two big cars is substantially lower than the chance of fatality in crashes involving two small ones. See NHTSA, Traffic Safety Trends and Forecasts (October 1981), at 58, reprinted at J.A. 79.
. The point is widely recognized. See, e.g., NHTSA, Small Car Safety in the 1980's at 77 ("In the event of a crash, the likelihood of injury is increased as the car’s size decreases.”); cf. Center for Auto Safety, Small on Safety 87 (Clarence Ditlow ed. 1972) ("Small size and light weight impose inherent limitations on the degree of safety that can be built into a vehicle. All known studies relating car size to crash injury conclude that occupants of smaller cars run a higher risk of serious or fatal injury than occupants of larger cars.”).