TABLE OF CONTENTS
Page
I. Background------------- 206
A. Early Regulation of Gasoline Lead---------------------------- 207
B. The Current Regulation_____________________________________ 207
C. Issues Presented___________________________________________ 209
II. EPA’s Statutory Authority.....„............._................ 209
A. Authority to Regulate Small Refiners_________________________ 210
B. The Relationship Between the Gasoline Lead Standard and the Ambient Air Quality Standard for Lead------------------------ 211
III. Judicial Review Under Clean Air Act § 307(d)__________________ _ _ 213
A. Notice of Rulemaking and Statement of Basis and Purpose-------- 213
B. The Record for Judicial Review______________________________ 214
C. Substantive Review________________________________________ 214
D. Procedural Review____________________________ 216
IV. The 1.10 gplg Final Standard -------------------------------218
A. EPA’s Explanation----------------------------------------- 218
1. Health Effects__________________________________ 218
2. Feasibility____________________________________________ 219
3. Equity and Efficiency-----------------------------------220
B. What Constitutes Adequate Agency Reasoning------------------220
C. Reasonableness of the Final Standard--------------------------221
1. Agency Reversal of Position------------------------------221
2. Health Effects__________ -222
a. Gasoline Lead Generally------------------------------222
i. The Correlation Between Gasoline Lead and Blood Lead Levels____________________________________222
ii. Incidence of Lead Poisoning--------- 224
iii. Public Comments________________________________225
iv. Conclusion on Health Effects----------------------226
b. The Health Effects of Small Refiner Lead Use..........-■ 226
i. Sales to Urban Areas-----------------------------226
ii. Health Effects in Rural Areas.......... 228
3. Feasibility____________I-------------------------------229
a. Timely Objection____________________________________229
b. EPA’s Use of Aggregate Analysis----------------------• 230
4. Equity and Efficiency •-----------------------------------231
5. Summary____________________________'-----------------231
D. Regulatory Flexibility Analysis-------------------------------• 232
1. Reviewability of the Analysis-----------------------------■ 232
2. Adequacy of the Analysis--------------------------------234
E. The July 1,1983 Effective Date------------------------------234
F. Late Docketing of Comments________________________________235
G. Conclusion_________________________________________________236
*206Page
V. The Interim 1.90 gplg Standard-------------------- ■ 237
A. Notice-------------------------------------- • 237
B. Feasibility----------------------------------- • 239
C. What Standard Does EPA Return To?------------ • 240
VI. The Definition of “Small Refinery”---------------- -• 240
A. The Standard for Adequate Notice -------------- ■ 241
B. Past Production Requirement-------------------•'242
C. Past Ownership Requirement--------------------243
1. Notice from EPA ------------------------- -243
2. Actual Notice-----------------------------.244
3. Notice by Others ------------------------- 244
4. The July 1,1981 Cutoff Date-----------------246
5. Conclusion--------------------------------246
D. Our Power to Issue a Stay---------------------246
VII. Conclusion.................................... -247
AppendixI---------------------------------------- - 247
Appendix II-------------------------------------- - 249
Before WILKEY, WALD and MIKVA, Circuit Judges.Opinion for the Court filed by Circuit Judge WALD.
WALD, Circuit Judge:Petitioners Small Refiner Lead Phase-Down Task Force (SRTF), Plateau, Inc., and Simmons Oil Co. seek review of an Environmental Protection Agency (EPA) regulation that sets lead-content limits for leaded gasoline produced by certain “small” refiners. 47 Fed.Reg. 49,322 (Oct. 29, 1982) (to be codified at 40 C.F.R. § 80.2, .4, .7, .20). In brief, the new rule: (1) narrows EPA’s previous definition of “small refinery”; (2) requires small refiners to meet an interim standard of no more than 1.90 grams of lead per gallon of leaded gasoline (grams per leaded gallon or gplg) as of November 1, 1982; and (3) requires small refiners to meet a final standard (equal to the large refiner standard) of no more than 1.10 gplg as of July 1, 1983.
We vacate the interim 1.90 gplg standard 1 because EPA promulgated it without adequate notice and the standard is not supported by the evidence in the record. We also vacate one clause in the definition of “small refinery” as promulgated without adequate notice and not supported by the evidence in the record. We uphold the remainder of the regulation, including the 1.10 gplg final standard, as within EPA’s statutory authority, not arbitrary, capricious, or an abuse of discretion, and not procedurally flawed.
I. Background
Adding lead to gasoline is an inexpensive way to produce the high-octane gasoline needed by today’s high-compression auto and truck engines. Other methods of producing high-octane gasoline require refiners to invest large sums in refining equipment and also involve higher operating costs. In particular, a refinery that uses less lead must use more crude oil to produce the same amount of gasoline. Use of lead in gasoline, however, has-grave social costs. Lead is highly poisonous to people and gasoline lead emissions are a major contributor to lead poisoning, with small children at greatest risk.1
Section 211(c)(1)(A) of the Clean Air Act, 42 U.S.C. § 7545(c)(1)(A), authorizes EPA to regulate fuel additives which “may reasonably be anticipated to endanger the public health or welfare.” At various times over the last decade, EPA has used this *207authority to regulate the lead content of gasoline. In this part, we first review EPA’s early efforts to regulate gasoline lead and then summarize the regulations that are challenged in this case.
A. Early Regulation of Gasoline Lead
EPA first issued lead content regulations in 1973. The regulations required refiners to limit the lead content of gasoline to 1.7 grams per gallon (gpg) beginning on January 1, 1975, steadily decreasing thereafter to 0.5 gpg on January 1,1979.2 In contrast to the grams per leaded gallon (gplg) standards at issue here, this earlier standard was a “pooled” average for all gasoline— leaded and unleaded — produced by a particular refiner. Thus, refiners that produced a high proportion of unleaded gasoline could use more lead per gallon of leaded gasoline than refiners that produced primarily leaded gasoline.3 EPA exempted small refiners from the lead content rules until January 1, 1977, “in recognition of the special lead-time problems faced by this group.”4 We upheld this regulation in its entirety. Ethyl Corp. v. EPA, 541 F.2d 1 (D.C.Cir.) (en banc), cert. denied, 426 U.S. 941, 96 S.Ct. 2662, 49 L.Ed.2d 394 (1976).
ÉPA later postponed the effective date of the 0.5 gpg standard until October 1, 1980, and large refiners have been obeying it since then.5 However, in 1977, Congress granted small refiners partial relief from EPA’s lead content rules until October 1, 1982.6 To qualify as “small,” a refinery had to: (1) have been in operation before October 1, 1976; (2) have crude oil capacity of 50,000 barrels per day (bpd) or less; and (3) be owned or controlled by a refiner with total crude oil capacity of 137,500 bpd or less. Clean Air Act § 211(g)(1)(B), 42 U.S.C. § 7545(g)(1)(B). Congress barred EPA from requiring a small refinery to reduce lead content below a “sliding-scale” standard that was linked to refinery size. Refineries producing gasoline at 5,000 bpd or less could use 2.65 gpg; 5,001 to 10,000 bpd refineries could use 2.15 gpg; and so on down to 0.80 gpg for 20,001 to 25,000 bpd refineries. Congress gave EPA broad discretion to regulate small refineries on and after October 1, 1982, “taking into account the experience under the [sliding-scale standard].” Id. § 211(g)(2), 42 U.S.C. § 7545(g)(2).
In 1979, EPA promulgated a new regulation requiring small refineries to meet the 0.5 gpg large refinery standard as of October 1, 1982.7 We upheld the regulation by unpublished order. Coastal States Gas Corp. v. EPA, No. 79-2174 (D.C.Cir. Sept. 17, 1980).
In addition to these direct controls on the lead content of gasoline, EPA also promulgated a national ambient air quality standard for lead of 1.5 micrograms per cubic meter (ug/m3).8 We upheld the regulation on appeal. Lead Industries Association v. EPA, 647 F.2d 1130 (D.C.Cir.), cert. denied, 449 U.S. 1042, 101 S.Ct. 621, 66 L.Ed.2d 503 (1980).
B. The Current Regulation
On February 22, 1982, roughly seven months before the 0.5 gpg small refinery standard was to take effect, EPA published a notice of proposed rulemaking in which it solicited comments on whether it should relax the 0.5 gpg standard either for all refiners or for small refiners only. EPA also solicited comments on how to define “small refinery.” The rulemaking notice was short and obviously tentative; EPA did not even issue a proposed rule. 47 Fed.Reg. 7812 (Feb. 22, 1982).9
*208In its notice of proposed rulemaking, EPA further proposed to suspend indefinitely the October 1,1982 deadline for small refiners to meet the 0.5 gpg standard. The agency explained that the suspension would permit small refiners to delay making the capital investments needed to meet the 0.5 gpg standard. It promised small refiners that the final rule “will take into account the lead time required for construction of any processing equipment needed for compliance.” Id. at 7814/2.
After receiving hundreds of comments from health specialists, industry, public interest groups, state and local governments, etc., EPA concluded that lead exposure was still a major health problem and that gasoline lead was “a significant contributor to this problem through its presence in the air, dirt and dust.” Id. at 38,078/1 (Aug. 27, 1982). For large refiners, EPA therefore decided that instead of relaxing the 0.5 gpg standard, it would tighten that standard further in the future, and issued a proposed rule to accomplish that goal.
The proposed rule changed the lead content limit from 0.5 grams per gallon (gpg) to 1.10 grams per leaded gallon (gplg). Id. at 38,078/3. Since slightly less than half of all gasoline produced today is leaded, this change would not affect current lead emissions. However, in the future, as use of leaded gasoline declines (cars built in 1975 and later years use unleaded gasoline only), the “pooled” gpg standard would have permitted refiners to use more lead per leaded gallon, while the gplg standard will not. EPA now estimates that under a uniform 0.50 gpg pooled standard, 1990 lead use would have been 38.8 billion grams, while under a uniform 1.10 gplg leaded-only standard, 1990 lead use will be only 16.2 billion grams, a 58% reduction.10
EPA proposed a looser 2.50 gplg standard for small refiners. Also, small refiners that used less than 2.50 gplg would be able to sell unused “lead credits” to other refiners under an inter-refiner averaging scheme. However, EPA proposed to define “small refinery” narrowly, to include only refineries that: (1) were in operation prior to October 1, 1976; (2) had average gasoline production of 10,000 bpd or less during the most recent calendar quarter; and (3) were not owned or controlled by a refiner with total average production during the most recent calendar quarter of more than 70,000 bpd. Id. at 38,088. EPA solicited comments on: whether a 2.50 gplg standard was too generous and “whether a standard such as 2.15 gp[l]g ... may be more appropriate”; whether the special small refiner standard “should be limited to some definite time period”; and whether there were any loopholes in the definition of small refinery that might permit small refiners to “expand their production of leaded gasoline.” Id. at 38,082/1, 38,085/3!
Finally, EPA suspended for one month the October 1, 1982 date for small refiners to meet the 0.5 gpg large refiner standard. EPA explained that it would take final action on the proposed 2.50 gplg small refiner standard before November 1, 1982, and stated again that the suspension would permit small refiners to defer the capital investments needed to comply with a strict standard. Id. at 38,091/1.
On October 29, 1982, only 2 days before the 0.5 gpg standard was to go into effect for small refiners, EPA promulgated a uniform 1.10 gplg standard for both large and small refiners, reasoning that permanent special treatment for small refiners would be “inequitable and inconsistent with the purposes of the Clean Air Act.” Id. at 49,324/1 (Oct. 29, 1982). The agency required large refiners to meet the new standard beginning November 1, 1982. It recognized, however, that its February and August proposals gave small refiners “some reason to ... delay[ ] projects designed to meet a more stringent standard,” and therefore gave small refiners until July 1, 1983 to meet the 1.10 gplg large refiner standard. Id. In the interim period from' November 1, 1982 to July 1, 1983, EPA required small refiners to meet only a 1.90 gplg standard, explaining that this standard “should generally be achievable by [small] refineries through the use of the [inter-refinery] averaging provisions in the regulations.” Id. at 49,324/2.
EPA also further tightened the definition of “small refinery” to include only refineries that: (1) were in operation before October 1, 1976; (2) had average gasoline production of 10,000 bpd or less during each *209calendar quarter since July 1, 1981; and (3) were not owned or controlled since July 1, 1981 by a refiner with total average production of more than 70,000 bpd. Id. at 49,-331-32.
C. Issues Presented
No large refiner challenges EPA’s change from a 0.5 gpg standard to a 1.10 gplg standard. SRTF, representing 13 small refiners, argues that EPA exceeded its statutory authority in promulgating a uniform I.10 gplg final standard for all refiners, that the 1.10 gplg final standard is substantively flawed, and that the 1.90 gplg interim standard is proeedurally and substantively flawed. Finally, SRTF argues that EPA did not give adequate notice that the final definition of “small refinery” might require a refiner to have produced no more than 10,000 bpd in any calendar quarter since July 1,1981 (“past production requirement”). Plateau, Inc.,- a small refiner, presents procedural and substantive challenges to the interim standard and to the effective date of the final standard.
Simmons Oil Co., a small refiner that was owned by a large refiner until January 4, 1982, argues that EPA did not give adequate notice that it might require a refinery, to qualify as “small,” not to have been owned or controlled by'a large refiner after July 1, 1981 (“past ownership requirement”). Amicus Giant Industries supports Simmons’ challenge to the past ownership requirement.
Several intervenors support the final rules: large refiners (Texaco, Gulf, Sun, and Exxon, on a joint brief); a mid-sized refiner (Texas City Refining); a refinery that was “small” under the old, but not under the new definition (United Refining); environmental organizations (Natural Resources Defense Council and Environmental Defense Fund); and a consumer organization (Consumers Union).
We denied SRTF’s motion for stay pend-mg appeal, but on January 12, 1983, granted Simmons’ motion for stay of the past ownership requirement. After oral argument on January 17, we issued an order on January 26 (reprinted as an appendix to this opinion), in which we upheld the final 1.10 gplg standard and the past production requirement, but vacated the 1.90 gplg interim standard primarily for lack of notice. Our order gave a brief statement of reasons and noted that a full opinion would follow. On February 9, we issued a second short order (also reprinted as an appendix) vacating the past ownership requirement for lack of notice.11 We chose this unusual course because our disposition made speed of the essence — our decisions vacating the interim standard and the past ownership requirement will have practical effect only until the final standard takes effect on July 1.
This opinion explains more fully the reasoning that underlay our initial orders. In part II, we uphold EPA’s statutory authority to set a small refiner standard stricter than the old sliding-scale standard. Part III discusses the standard for judicial review of EPA rulemaking under the Clean Air Act. Part IV concludes that the 1.10 gplg final standard is reasonable. Part V vacates the 1.90 gplg interim standard for lack of notice and lack of support in the record. Part VI upholds the past production requirement but vacates the past ownership requirement for lack of notice and lack of support in the record.
II. EPA’s Statutory Authority
The parties agree that EPA has authority under § 211(c)(1)(A), 42 U.S.C. § 7545(c)(1)(A), to regulate lead in gasoline, since lead is a fuel additive that “may reasonably be anticipated to endanger public health or welfare.” See Ethyl Corp., 541 F.2d at 11-32.12 SRTF argues, however, *210that § 211(g)(2), 42 U.S.C. § 7545(g)(2), prohibits EPA from requiring small refiners to reduce gasoline lead levels below the sliding-scale standard unless it makes specific findings concerning the need to do so. SRTF also argues that EPA cannot set a gasoline lead standard stricter than necessary to meet the ambient air quality standard for lead.13 We reject both arguments.
A. Authority to Regulate Small Refiners
Section 211(g)(2) grants EPA broad authority to regulate small refiners on and after October 1, 1982, subject only to the condition that EPA must “tak[e] into account” experience under the sliding-scale standard:
The Administrator may promulgate such regulations as he deems appropriate with respect to the reduction of the average lead content of gasoline refined by small refineries on and after October 1, 1982, taking into account the experience under the [sliding-scale] provisions of this paragraph.
(Emphasis added.)
On its face, this condition is a mild one. It merely requires EPA in good faith to consider past experience, and does not require EPA to make specific findings concerning the need for a strict small refiner standard. Congress did require specific findings elsewhere in § 211 and elsewhere in the Act. In particular, § 211(c)(2)(C), 42 U.S.C. § 7545(c)(2)(C), permits the Administrator to prohibit use of a fuel additive altogether only if “he finds, and publishes such finding” that the prohibition will not cause the use of another, equally dangerous additive.14 As we noted in a related context, “this is the only conclusion the Administrator is expressly required to ‘find’ before regulating a fuel or fuel additive for health reasons.” Ethyl Corp., 541 F.2d at 12 (emphasis in original). We must assume that Congress’ use of the weaker phrase “take into account” in § 211(g)(2) was intentional.15
The legislative history confirms this “plain language” reading of the statute. The House had granted permanent special treatment to small refiners. 123 Cong.Ree. 16,951-52 (1977), reprinted in 4 Leg.Hist., supra note 12, at 3320-24. The Senate, however, adopted a weaker version calling only for a five-year grace period, to last until October 1, 1982. Senator Muskie, the floor manager for the 1977 amendments, explained:
The amendment provides a temporary relaxation in the average lead requirement for small refineries . .. until October 1, 1982. This will allow time for them to install the necessary equipment to produce gasoline with lower lead content or to make alternate plans.
Id. at 18,474, 3 Leg.Hist. 1059 (emphasis added).
In conference, the House “concur[red] in the Senate provision” with several small changes. H.R.Rep. No. 564, 95th Cong., 1st Sess. 162 (1977) (“Conf.Rep.”), 3 Leg.Hist. 381, 542, 1977 U.S.Code Cong. & Ad.News 1502, 1543. One of these changes was to add the “taking into account” requirement. The Conference Committee explained, however, that EPA would nevertheless have *211broad discretion to regulate small refiners after October 1, 1982:
[0]n the basis of experience under this provision, the Administrator shall promulgate such standards as he sees fit for the period beyond October 1, 1982 ....
Id. (emphasis added).
In short, the legislative history of § 211 (g)(2) confirms EPA’s broad discretion to regulate small refiners, and give no hint that the agency must make specific findings before doing so. Finally, EPA’s interpretation is “entitled to deference.” Federal Election Commission v. Democratic Senatorial Campaign Committee, 454 U.S. 27, 32, 102 S.Ct. 38, 42, 70 L.Ed.2d 23 (1981).16
We conclude, then, that § 211(g)(2) requires EPA in good faith to take experience under the sliding-scale standard into account, but does not require specific findings concerning that experience. Accord Ethyl Corp., 541 F.2d at 32 n. 66 (the requirement in § 211(c)(2)(A) that EPA, before regulating a fuel additive must “consider[]” regulating auto emissions under § 202 instead, “means, of course, no more than it says: actual good faith consideration of the specified ... options”).17
The only remaining question is whether EPA adequately considered experience under the sliding-scale standard in selecting a lead-content limit for small refiners. As we held in construing similar language in the Clean Water Act, EPA must reach an “express and considered conclusion” about the bearing of experience under the sliding-scale standard, but need not give “any specific weight” to this factor. Weyerhaeuser Co. v. Costle, 590 F.2d 1011, 1045 (D.C.Cir.1978); accord BASF Wyandotte Corp. v. Costle, 598 F.2d 637, 662 (1st Cir.1979), cert. denied, 444 U.S. 1096, 100 S.Ct. 1063, 62 L.Ed.2d 784 (1980). Here, EPA devoted an entire rulemaking to the issue of the appropriate lead standard for small refiners. It concluded that:
experience under the two-tiered approach has been that it fosters inequities, as small refineries are granted a competitive advantage over large refineries.... [T]his two-tiered approach can no longer be justified, in light of the health effects of lead exposure and the scope and profitability of small refinery operations.
47 Fed.Reg. at 49,324/3. Thus, EPA has complied with the statutory command to take past experience into account.18
B. The Relationship Between the Gasoline Lead Standard and the Ambient Air Quality Standard for Lead
SRTF also argues that EPA lacks authority to control gasoline lead except to the extent necessary to meet the ambient air quality standard for lead, and that even under the sliding-scale standard, no areas that are served by small refineries would violate the ambient air standard. To prevail, SRTF must overcome the deference that we give to EPA’s interpretation of its governing statute. It falls far short; indeed, we find no support for SRTF’s construction in the statutory language, the legislative history, or sound policy.
Nothing in § 211 expressly prevents EPA from regulating fuel additives unless necessary to meet the ambient air quality standards established under sections 108 and *212109. Indeed, § 211 does not even cross-refer to sections 108 and 109. In contrast, when Congress wanted EPA to consider other sections of the Act before regulating fuel additives, it said so. In particular, § 211(c)(2)(A), 42 U.S.C. § 7545(c)(2)(A), requires EPA, before regulating a fuel additive, to “consider[]” regulating auto emissions under § 202 instead. Thus, § 211 on its face permits EPA to regulate gasoline lead whether or not the ambient air standard for lead is being met. Cf. Ethyl Corp., 541 F.2d at 54 n. 124 (finding “no basis in the statute” for the claim that EPA must establish an ambient air standard for lead under § 108 before regulating gasoline lead under § 211).
SRTF points out that both § 108 (governing ambient air quality standards) and § 211 (governing fuel additives) allow EPA to regulate a pollutant only if it “may reasonably be anticipated to endanger public health or welfare.” But the fact that the Clean Air Act establishes the same threshold standard for EPA’s power to establish ambient air standards and its power to regulate fuel additives does not mean that the substantive standard that EPA sets for a fuel additive must be tied to the substantive standard it sets for ambient air quality. On the contrary, § 109(b)(1), 42 U.S.C. § 7409(b)(1), prescribes a specific substantive standard — “requisite to protect the public health” allowing “an adequate margin of safety.”19 Section 211(c), in contrast, broadly authorizes EPA to “control or prohibit” the sale of fuel additives.
Similarly, other sections of the Act have the same threshold standards as sections 108 and 211, yet each has its own substantive standard. See, e.g., §§ 111 (new source performance standards), 202 (motor vehicle emission standards).20 Thus, Congress’ use of the same threshold standard in these disparate sections cannot imply that they share the same substantive standard.
The legislative history confirms the conclusion that EPA can set different substantive standards under sections 109 (ambient air quality) and 211 (fuel additives). Until 1977, different sections of the Act had different threshold standards. In particular, § 108(a)(1)(A) had a different standard (“has an adverse effect on public health or welfare”) than § 211(c)(1)(A) (“will endanger the public health or welfare”).21 In 1977, Congress established a uniform formula (“may reasonably be anticipated to endanger public health or welfare”) for the threshold decision whether “the Administrator may regulate a pollutant.” H.R.Rep., supra note 12, at 50, 4 Leg.Hist. 2517, 1977 U.S.Code Cong. & Ad.News at 1128 (emphasis added).22 Nothing suggests that Congress also meant to establish a uniform substantive standard tied to the ambient air quality standards. The legislative history of the 1977 amendments mentions § 108 only as one of a list of sections to which the new threshold standard will apply.23 Moreover, Congress left the substantive standards in §§ 111 and 202 unchanged, and thus could not possibly have intended to set up a single overriding substantive standard.
*213SRTF’s construction would also be foolish as a policy matter. First, sections 109 and 211 are designed to accomplish different goals. Section 109 establishes only ambient air standards. To meet the ambient standards, the states must develop emission limits for individual pollution sources. Necessarily, these individual source limits will be stricter in some geographic areas than in others. Section 211, in contrast, permits EPA to establish nationally uniform emission limits on fuel additives.24 Those limits may be more or less stringent than needed to meet the ambient air standards in any particular area. Thus, it would not make sense to force regulation under § 211 to move in lockstep with the ambient air standards promulgated under § 109.
Second, the ambient air standard for lead does not fully protect the public health against the adverse effects of gasoline lead. Lead is a heavy element that quickly settles out of the air onto the ground, where it can be ingested or inhaled along with dust and dirt. The ambient standard does not take full account of dirt and dust lead. See Environmental Protection Agency, Fuels and Fuel Additives: Denial of Petition to Repeal Lead Phasedown Regulations, 45 Fed.Reg. 54,090, 54,093/1 (1980) (ambient standard for lead was “not designed to protect against exposure from non-air sources, including those resulting from automobile emissions”) (footnote omitted). Also, some of the lead used by rural refineries will make its way into the food chain to be eaten by urban children (who are at greatest risk of lead poisoning). We decline to read § 211 as precluding EPA from regulating these indirect sources of lead exposure.25
For the foregoing reasons, we hold that EPA may issue lead-content regulations under § 211(c) that are stricter than necessary to meet the ambient air quality standard for lead issued under § 109(b).26 Having disposed of SRTF’s statutory objections to EPA’s authority, we turn to the substantive and procedural validity of the lead-content regulations. We begin by reviewing the special statutory provisions for judicial review of EPA rulemaking under the Clean Air Act.
III. Judicial Review Under Clean Air Act § 307(d)
In 1977, Congress — concerned that the Administrative Procedure Act (APA), 5 U.S.C. § 553, did not provide procedures adequate for the complex scientific issues involved in EPA rulemaking — created new procedures for most rulemaking under the Clean Air Act, including rulemaking under § 211. Clean Air Act § 307(d), 42 U.S.C. § 7607(d). This part contains an overview of the § 307(d) procedures.
A. Notice of Rulemaking and Statement of Basis and Purpose
Under the APA, 5 U.S.C. § 553(b)(3), an agency must publish in the Federal Register a notice of proposed rulemaking which “shall include ... either the terms or substance of the proposed rule or a description of the subjects and issues involved.” Clean Air Act § 307(d)(3) requires a much more detailed notice of rulemaking. EPA must publish a notice of proposed rulemaking “as provided under [APA] section 553(b).” In addition, this notice “shall be accompanied by a statement of its basis and purpose,” which:
*214shall include a summary of—
(A) the factual data on which the proposed rule is based;
(B) the methodology used in obtaining the data and in analyzing the data; and
(C) the major legal interpretations and policy considerations underlying the proposed rule.
Thus, there are basically two major differences between APA § 553(b) and Clean Air Act § 307(d)(3). First, the Clean Air Act, unlike the APA, requires EPA to issue a “proposed rule.” Although § 307(d)(3) does not explicitly require EPA to issue a proposed rule, it refers to the agency’s duty to explain “the factual data on which the proposed rule is based” and “the major legal interpretations and policy considerations underlying the proposed rule.” Also, at the final rule stage, § 307(d)(6) requires EPA to explain “any major changes in the promulgated rule from the proposed rule.” EPA cannot comply with these statutory requirements unless it issues a “proposed rule.”27 Second, the APA requires an agency to explain its reasoning only when it issues a final rule. In contrast, Clean Air Act § 307(d)(3) requires EPA to give a detailed explanation of its reasoning at the “proposed rule” stage as well. Turning to the notice of final rulemaking, APA § 553(c) requires an agency to provide only a “concise general statement of basis and purpose” (emphasis added). In contrast, § 307(d)(6) requires a more detailed explanation — a “statement of basis and purpose” (without the qualifiers “concise” and “general”). In particular, the statement of basis and purpose must include the specific items required at the proposed rule
stage and must also include “an explanation of the reasons for any major changes in the promulgated rule from the proposed rule” and “a response to each of the significant comments, criticisms, and new data submitted ... during the comment period.”
B. The Record for Judicial Review
The APA does not prescribe any procedures for assembling a “record” for judicial review of informal rulemaking. The Clean Air Act, in contrast, requires EPA to establish a “rulemaking docket” that is open to the public and must include: (1) the notice of proposed rulemaking and accompanying statement of basis and purpose; (2) “[a]ll data, information, and documents” on which EPA relies in its statement of basis and purpose; (3) other documents which become available after the proposed rule is published and which “the Administrator determines are of central relevance to the rulemaking”; (4) all written comments and documents submitted to EPA during the comment period; (5) a transcript of any public hearings; and (6) the notice of final rulemaking and accompanying EPA explanation.28 The final rule must be based entirely on material that has “been placed in the docket as of the date of ... promulgation,” and the material in the docket is the exclusive record for judicial review.29
C. Substantive Review
The standard for substantive judicial review of EPA action under the Clean Air Act is taken directly from the APA: The court may reverse only if EPA’s action was “arbitrary, capricious, an abuse of discre*215tion, or otherwise not in accordance with law.” Clean Air Act § 307(d)(9)(A), 42 U.S.C. § 7607(d)(9)(A); see 5 U.S.C. § 706(2)(A) (parallel APA provision).
Congress considered but ultimately rejected a proposal to establish a “substantial evidence” standard of review for the Clean Air Act. It did so to avoid any suggestion that EPA should be subject to a stricter standard of review than other agencies. See 123 Cong.Rec. 26,851 (1977), 3 Leg.Hist. 366 (statement of Sen. Muskie) (“the Environmental Protection Agency should [not] be singled out in this way”).30 Congress was aware, however, that there may be “little practical difference” between substantial evidence review and arbitrary and capricious review. Conf.Rep. at 178, 3 Leg. Hist. 558, 1977 U.S.Code Cong. & Ad.News at 1559;31 see Sierra Club v. Costle, 657 F.2d 298, 323 n. 67 (D.C.Cir.1981); Lead Industries Association v. EPA, 647 F.2d 1130, 1146 n. 30 (D.C.Cir.1980), cert. denied, 449 U.S. 1042, 101 S.Ct. 621, 66 L.Ed.2d 503 (1980) (both discussing substantive review under § 307(d)).32
The scope of review under the “arbitrary and capricious” formula is well established. We must engage in a “searching and careful” review of the record. Citizens to Preserve Overton Park v. Volpe, 401 U.S. 402, 416, 91 S.Ct. 814, 823,28 L.Ed.2d 136 (1971); see Ethyl Corp., 541 F.2d at 35. For its decision to be sustained, “the agency must consider all of the relevant factors and demonstrate a reasonable connection between the facts on the record and the resulting policy choice.” Sierra Club v. Costle, 657 F.2d at 323 (footnote omitted). When the facts are uncertain, the Administrator “should so state and go on to identify the considerations he found persuasive.” Industrial Union Department, AFL-CIO v. Hodgson, 499 F.2d 467, 476 (D.C.Cir.1974); see Lead Industries Association, 647 F.2d at 1146-47. In short, we must, in Judge Leventhal’s phrase, take a “hard look” at both the facts and the agency’s reasoning.33
In taking this hard look, however, we have only limited power to second-guess the agency’s reasoning. We can and do insist that the agency’s reasons and policy choices do “ ‘not deviate from or ignore the ascertainable legislative intent.’ ” Ethyl Corp., 541 F.2d at 36 (quoting Greater Boston Television Corp. v. FCC, 444 F.2d 841, 850 (D.C.Cir.1970), cert. denied, 403 U.S. 923, 91 S.Ct. 2233, 29 L.Ed.2d 701 (1971)). Beyond that, however, we can only require the *216agency’s reasons and policy choices to conform to “certain minimal standards of rationality.” Id. If so, the rule is reasonable and must be upheld.
D. Procedural Review
Under the APA, the reviewing court must reverse agency action taken “without observance of procedure required by law,” 5 U.S.C. § 706(2)(D), after taking “due account ... of the rule of prejudicial error,” id. § 706 (last clause). Clean Air Act § 307(d), at least on its face, more narrowly circumscribes judicial power to reverse EPA action on procedural grounds. As under the APA, the court must reverse agency action taken “without observance of procedure required by law.” Clean Air Act § 307(d)(9)(D), 42 U.S.C. § 7607(d)(9)(D). However, the court must also find that EPA’s “failure to observe such procedure is arbitrary or capricious,” id. § 307(d)(9)(D)(i), 42 U.S.C. § 7607(d)(9)(D)(i), and that the procedural errors were “so serious and related to matters of such central relevance to the rule that there is a substantial likelihood that the rule would have been significantly changed if such errors had not been made,” id. § 307(d)(8), 42 U.S.C. § 7607(d)(8).34
Unfortunately, it is not clear what these restrictions on procedural reversal mean. “Arbitrary and capricious” review generally refers to the requirement that an agency “must consider all the relevant factors” and reach a “reasonable” conclusion. Sierra Club v. Costle, 657 F.2d at 323. Yet Congress could hardly have intended to leave EPA free to ignore the procedural requirements of § 307(d), so long as the agency gives a decent reason for doing so.
The requirement that the errors be “so serious and related to matters of such central relevance” that “there is a substantial likelihood that the rule would have been significantly changed if such errors had not been made” is also ambiguous. The first clause (“so serious and related to matters of such central relevance”) has no independent force; it will automatically be satisfied by any error that creates “a substantial likelihood that the rule would have been significantly changed.” This leaves the problem of interpreting the two key terms — “substantial likelihood” and “significant change” — neither of which has a precise meaning. These terms appear on their face to be stricter than the APA’s “prejudicial error” rule, which requires only a possibility that the error would have resulted in some change in the final rule. See, e.g., Weyerhaeuser Co. v. Costle, 590 F.2d 1011, 1031 n. 27 (D.C.Cir.1978) (“we cannot be sure that under the correct procedures the Agency would have reached the same conclusion”). But it is hard to say how much stricter the Clean Air Act standard is.
The legislative history provides a partial answer to these puzzles, and leads us to conclude that there is less to § 307(d)’s requirements for procedural reversal than meets the eye. Section 307(d) derives from the House version of the 1977 Clean Air Act Amendments; the Senate bill had no comparable provision. The House bill provided for oral hearings and for limited cross-examination on “disputed issue[s] of material fact.” Cross-examination could be conducted only “to such extent and in such manner as the Administrator considers necessary and appropriate in view of the nature of the issues involved, the number of participants and the nature of their interests, and any need for expedition.” H.R. 6161, 95th Cong., 1st Sess. sec. 305(a), § 307(d)(5)(B) (1977), 4 Leg.Hist. 2220, 2431. Moreover, the House report explains that the right to cross-examine on issues of “fact” refers only to “ ‘adjudicative facts’ ” and further cautions that “[i]n many cases, such facts may not be present, or ... policy considerations [may be] so dominant as to reduce to insignificance the question whether one *217particular ‘adjudicative’ fact is true or not.” H.R.Rep. at 821, 4 Leg.Hist. 2788,1977 U.S. Code Cong. & Ad.News at 1400 (footnote omitted).
The House committee recognized that the various limits on the right to cross-examination were imprecise, so that questions could arise over, e.g., “whether a given question involves ‘facts’ or ‘policy’ or whether a given fact is ‘legislative’ or ‘adjudicative.’ ” Id. at 322, 4 Leg.Hist. 2789, 1977 U.S.Code Cong. & Ad.News at 1401. It restricted procedural reversal to avoid disputes over the manner in which EPA structured cross-examination:
To prevent rulemaking from bogging down in arguments about such matters, and to underline that the agency is authorized to adapt rulemaking procedures to the individual case, the committee has limited the extent to which the Administrator’s decisions on such procedural matters may be reversed during judicial review.
... [T]he court is directed to consider two factors. The first is whether the Administrator’s determination on the procedural point is “arbitrary or capricious.” The second is whether the procedural errors “were so serious and related to matters of such central relevance to the rule that there is a substantial likelihood that the rule would have been significantly changed if such errors had not been made.”
Id. (citations to the bill omitted).
In the House bill, consistent with the committee explanation, the “arbitrary and capricious” requirement was contained not in § 307(d)(9)(D) (governing procedural review generally), but in § 307(d)(5), the subsection that directed EPA to hold an oral hearing and allow cross-examination. H.R. 6161, supra, sec. 305(a), § 307(d)(5)(C), 4 Leg.Hist. 2432. Thus, the “arbitrary and capricious” test applied only to EPA decisions concerning the oral hearing and cross-examination and did not apply to procedural issues arising out of other subsections. Moreover, the provision did not authorize EPA to violate the procedural dictates of the Act; rather, it merely insured that EPA could take reasonable steps to implement the cross-examination procedures.
Less clear is the scope of the requirement that a procedural error create a “substantial likelihood that the rule would have been significantly changed.” The House also directed this requirement primarily at procedures for cross-examination, but it applies on its face to all procedural determinations. The legislative history, unfortunately, provides no help on what “substantial likelihood” and “significant change” mean. We can say this much, though. The House did not intend to cut back on the procedural review provisions of the APA. Rather, the House was concerned that the APA procedures were “inadequate,” H.R.Rep. at 318, 4 Leg.Hist. 2785, 1977 U.S.Code Cong. & Ad. News at 1397, and wanted to add new procedural protections, but also wanted to minimize disputes over EPA’s compliance with the new procedures.
The Conference Committee deleted the cross-examination provisions in the House bill but retained the limits on procedural reversal (moving the “arbitrary and capricious” requirement from the truncated § 307(d)(5) to § 307(d)(9)(D)). These limits are not mentioned in either the Conference Report or in the House and Senate debates on the Conference Committee bill. So far as appears, Congress never considered their residual meaning once the right to cross-examination was gone.35 It appears, however, that the Senate accepted the House understanding that § 307(d) did not cut back on existing APA procedural safeguards. Thus, Senator Muskie explained that he had opposed the House proposal for “substantial *218evidence” review because he believed that EPA ought to be subject to the same scrutiny as other agencies, no more, no less:
I am not suggesting that the courts should not discharge their duties with respect to EPA as they do with other agencies; at the same time, I want to emphasize ... that nothing in the bill or its legislative history was meant to single out EPA for special scrutiny either.
123 Cong.Rec. 26,851 (1977), 3 Leg.Hist. 366; see id. at 27,075, 3 Leg.Hist. 333 (statement of Rep. Broyhill) (“these new procedural requirements ... will assure the opportunity for more extensive public participation in the rulemaking process”).
In light of this history, we must interpret the final version of § 307(d) to carry forward the House understanding of what the procedural restrictions meant, so far as that understanding makes sense in the context of the final bill. We therefore conclude that § 307(d)(9)(D)(i)’s requirement that EPA’s procedural error be “arbitrary and capricious” means that we must affirm reasonable EPA decisions about how to implement the § 307(d) procedures. It does not mean that EPA can ignore those procedures so long as it has a reasoned basis for doing so.
The scope of § 307(d)(8)’s requirement that a procedural error create a “substantial likelihood that the rule would have been significantly changed” remains uncertain, for we do not know what “substantial likelihood” and “significant change” mean. At a minimum, failure to observe the basic APA procedures, if reversible error under the APA, is reversible error under the Clean Air Act as well. On the other hand, § 307(d)(8) sets a restrictive tone for our review of procedural errors that would not violate the APA. See Sierra Club v. Costle, 657 F.2d at 391 (“The essential message of so rigorous a standard [for procedural reversal] is that Congress was concerned that EPA’s rulemaking not be casually overturned for procedural reasons.”).36 We defer any effort to give further content to the terms “substantial likelihood” and “significant change” until our discussion below of concrete factual situations.
IV. The 1.10 gplg Final Standard
To restate briefly the aspects of the regulatory scheme that are relevant to this part, EPA’s prior regulations limited large refiner lead use to 0.5 gpg. Small refiners were subject to a sliding-scale standard that was to expire on October 1, 1982, after which they would have to meet the 0.5 gpg large refiner standard. In August 1982, EPA proposed to replace the 0.5 gpg large refiner standard with a stricter 1.10 gplg standard. For small refiners, EPA proposed a permanent 2.50 gplg standard, but also solicited comments on whether the special standard for small refiners should be permanent or temporary.
On October 29, 1982, the agency promulgated a uniform 1.10 gplg standard, effective November 1, 1982 for large refiners and July 1, 1983 for small refiners. Until July 1, 1983, small refiners had to meet a 1.90 gplg interim standard. This part reviews the basis for the 1.10 gplg final standard; part V reviews the 1.90 gplg interim standard.
A. EPA’s Explanation
In deciding to issue a uniform 1.10 gplg standard, EPA considered basically three factors. It relied principally on the adverse health effects of gasoline lead, but also considered at length the cost of reducing gasoline lead (including the cost to small refiners of meeting a uniform standard), and the equity and efficiency aspects of requiring small refiners to meet the same standard as large refiners.
1. Health Effects
After reviewing the available evidence, EPA reached the following general conclusions. First, “lead exposure is a major *219health problem” that results from the cumulative impact of “a combination of sources, including food, water, air, leaded paint and dust.” Therefore, efforts should be made to “reduce [all] preventable sources of lead exposure.” 47 Fed.Reg. at 38,072.37 Second, gasoline lead is a major and preventable source of air and dust lead. Id. Therefore, “further action by EPA to reduce lead in gasoline is ... prudent and reasonable.” Id. at 49,330/1. Third, the 1.10 gplg standard would significantly reduce future lead use as compared to a 0.50 gpg standard. Id. at 49,329 (estimating a 58% reduction in 1990, from 38.8 billion grams to 16.2 billion grams).
In its notice of proposed rulemaking, EPA had justified a looser standard for small refiners on the grounds that excess lead exposure was more of an urban than a rural problem and small refineries “are characterized by a more rural production and consumption pattern.” Id. at 38,081/1. Moreover, the possibility of high lead emission levels in specific urban areas served by small refiners did “not appear to be a significant problem.” Id. at 38,085/1.
The notice of final rulemaking discusses in general terms the deleterious effects of gasoline lead. It does not, however, directly rebut the inference from the notice of proposed rulemaking that gram for gram, small refinery emissions are less harmful because most small refineries serve rural customers. However, in a separate document entitled “Supplemental Responses to Comments,” issued with and incorporated by reference in its notice of final rulemaking, EPA explained that small rural refiners could sell gasoline to pipelines serving urban consumers and that it could not effectively regulate such sales.38 EPA also explained that small refiner lead use could not be ignored as insignificant. To the contrary, a permanent dual standard would increase lead use in 1990 from 16.2 billion grams to 20.0 billion grams (a 23% increase); small refiners would produce 34% of all gasoline lead.39
2. Feasibility
EPA analyzed the cost to small refiners of complying with the 1.10 gplg standard and implicitly concluded that most small refineries could meet it.40 Moreover, while large refineries enjoyed some economies of scale, the proposed 2.50 gplg small refiner standard “was far in excess of that amount which was necessary to put [small refiner*220ies] on an equal footing with large refineries.” Id. at 49,324/2. Finally, EPA noted that for small refiners, who produce mostly leaded gasoline, the 1.10 gplg leaded-only standard would be significantly less stringent than a 0.50 gpg pooled standard. Id. at 49,330.
3. Equity and Efficiency
EPA found that a uniform standard was preferable on efficiency grounds because a dual standard in effect subsidized small refiners that would otherwise be unable to produce gasoline at competitive prices. A uniform standard was also preferable on equity grounds because: (1) a dual standard gave a competitive advantage to small refiners; (2) the greatest benefits from a dual standard would go to those few small refiners who had not already made the capital investments needed to comply with the 0.5 gpg standard; and (3) small refiners had been on notice for over five years that they would eventually have to meet the same standard as large refiners. Id. at 49,324/1.
B. What Constitutes Adequate Agency Reasoning
EPA argues that to justify the uniform 1.10 gplg standard, it need only show that gasoline lead threatens human health and that the new regulation will reduce gasoline lead. By EPA’s logic, adverse health effects would permit it to justify any lead standard at all, without explaining why it chose the level it did. We cannot accept such incomplete reasoning. EPA must also explain why it chose 1.10 gplg rather than some other standard.
We do not demand certainty where there is none. There may be no strong reason for choosing 1.10 gplg rather than a somewhat higher or lower number. If so, we will uphold the agency’s choice of a numerical standard if it is within a “zone of reasonableness.” National Association of Broadcasters v. Copyright Royalty Tribunal, 675 F.2d 367, 374 (D.C.Cir.1982); Hercules, Inc. v. EPA, 598 F.2d 91,107 (D.C.Cir.1978) (citing cases). But our deference to its ultimate choice does not relieve EPA of the duty to explain why 1.10 gplg is an appropriate standard. A simpleminded argument that “gasoline lead is bad and our rule reduces gasoline lead” does not satisfy that duty.
Second, EPA must explain why small refiners contribute to the health problems created by lead emissions. We note that this requirement implicitly presupposes that EPA must consider small refiners as a separate category of the gasoline refining industry. In general, an agency can sub-classify an industry for purposes of regulatory analysis in any reasonable manner. See Weyerhaeuser Co. v. Costle, 590 F.2d at 1055 (petitioners have a “heavy burden” in challenging EPA’s subcategorization of the paper mill industry). Here, however, EPA was obliged to treat small refiners as a discrete class for three reasons.
Section 211(g)(2), 42 U.S.C. § 7545(g)(2), directs EPA to “tak[e] into account” experience under the old sliding-scale standard before regulating small refiners. Also, EPA proposed a separate standard for small refiners and must, under § 307(d)(6)(A), 42 U.S.C. § 7607(d)(6)(A), explain “the reasons for any major changes in the promulgated rule from the proposed rule.” Moreover, the Regulatory Flexibility Act, 5 U.S.C. §§ 603-604, requires EPA to consider “significant alternatives” to the chosen rule that accomplish the same end but minimize economic impact on “small entities.”41 Unless EPA analyzes small refiners as a discrete class, it cannot review experience under the sliding-scale standard, explain why it abandoned a dual standard, or consider significant alternatives under the Regulatory Flexibility Act.
Third, we cannot restrict our review to health considerations. Whether or not § 211(c)(2)(A) requires EPA to consider the economic effects of fuel additive regula*221tions, a question we do not address, all parties agree that it permits EPA to do so,42 and EPA in fact considered economic effects at length. Since economic considerations formed a major part of EPA’s rationale, we must review its economic reasoning as part of our review of whether its overall decision was reasonable.
C. Reasonableness of the Final Standard
SRTF challenges the basis for EPA’s conclusions on the health impacts of lead use and the feasibility of the 1.10 gplg standard.43 We consider its principal arguments in turn and conclude that EPA’s decision to impose a uniform 1.10 gplg standard is supported by substantial evidence in the record viewed as a whole.
1. Agency Reversal of Position
As an initial matter, SRTF argues that we must review EPA’s action with extra rigor because, in its final rule, EPA reversed its position in the August proposal that small refiners deserved special treatment. We do not think that heightened scrutiny is appropriate in this case.
We have held that “sudden and profound alterations in an agency’s policy constitute ‘danger signals’ that the will of Congress is being ignored,”44 and that “an agency changing its course must supply a reasoned analysis indicating that prior policies and standards are being deliberately changed, not casually ignored.”45 Here, however, there has been no change of policy. Since 1973, EPA has consistently concluded that while small refiners need some extra lead *222time, they should eventually meet the same standard as large refiners. In February 1982, EPA announced that it was considering a change in policy, and in August 1982, it proposed a change. But the agency ultimately adhered to its longstanding position in favor of a uniform standard.
We do not see why EPA’s action in considering but not adopting a change of course sends a “danger signal” that the final rule ignores the will of Congress. On the contrary, that EPA rethought but ultimately adhered to its original view is positive evidence of careful decisionmaking. Cf. National Wildlife Federation v. Gorsuch, 693 F.2d 156, 169 (D.C.Cir.1982) (EPA’s reconsideration of its interpretation of the Clean Water Act “is sufficient evidence of thoroughness to meet the standard for deference” to the agency’s interpretation).
2. Health Effects
SRTF presents two arguments concerning the health basis for the uniform 1.10 gplg standard. First, it notes that a 1.10 gplg/2.50 gplg dual standard would already produce substantial reductions in lead use, and argues that nothing in the record shows that the additional reductions from limiting small refiners to 1.10 gplg will have significant health benefits. Second, SRTF argues that small refineries are generally located in rural areas and therefore do not adversely affect the health of urban children. In subsection a, we find that EPA had ample basis in the record for generally seeking further reductions in gasoline lead. In subsection b, we find sufficient basis in the record for EPA’s belief that small refiners contribute to the lead emissions problem.
a. Gasoline Lead Generally
When EPA first issued lead-content regulations in 1973, there was no firm evidence that gasoline lead was harmful, only a “significant risk” that it was. Ethyl Corp., 541 F.2d at 12. Today, in contrast, there is compelling evidence that gasoline lead is a major cause of lead poisoning in young children. We discuss here only a few of the major pieces of evidence.
i. The Correlation Between Gasoline Lead and Blood Lead Levels
Gasoline lead correlates strongly with blood lead levels. A comprehensive study (“NHANES II”) by the Centers for Disease Control (CDC) shows that between 1976 and 1980, as gasoline lead use (under the spur of EPA regulation) declined from 105,000 tons per six-month period to 50,000 tons, mean blood levels declined in virtual lockstep from 16 ug/dl to less than 10 ug/dl. See Figure l.46 The percentage of people with excessive blood lead levels (30 ug/dl or above) showed a similar decline. As Dr. Vernon Houk, Acting Director of CDC’s Center for Environmental Health, explained:
This reduction was real. It was not due to chance, laboratory error, nor sampling of age, sex, race, urban vs. rural areas, income levels, or geographic regions. The most significant environmental change during this time was the reduced amount of lead used in the production of gasoline... . [This data] clearly demon*223strates that as we have removed lead from gasoline, we have also removed lead from ourselves and our children.47
Other studies show that when gasoline lead use peaks sharply each summer, blood lead levels peak sharply as well.48
[[Image here]]
Leaded gasoline is the only lead source that can explain this data. Both of the other two major sources of lead (lead paint and food) are roughly constant year-round; thus, neither would produce a summer peak in blood lead levels. Also, neither showed a sharp decline in recent years.49
Other evidence also suggests that gasoline lead is a major cause of lead poisoning, *224through its presence in the air, in dirt, and in dust. Gasoline lead is the source of over 90% of air lead, and there is a well-established relationship between inhalation of air lead and increases in blood lead.50 Also soil analysis shows that “lead levels ... in the front yards of urban homes are two to three times greater than ... in the back yards.”51 Gasoline lead can explain this pattern; the other major sources of dirt and dust lead (lead paint and stationary lead sources) cannot. Moreover, the concentrated lead in dirt and dust can easily produce lead poisoning. An urban child who licks her finger after running it across a table left undusted for several days can consume several times the “safe” daily dose of lead.52
ii. Incidence of Lead Poisoning
In the NHANES II study, the percentage of young children (aged 6 months to 5 years) with blood lead levels of 30 ug/dl or more dropped from 4.65% in 1976-1978 to 2.07% in 1978-1980.53 The record does not indicate how many children are currently suffering from lead poisoning, but the percentage probably dropped again in 1981 in response to the EPA-mandated decrease in large refiner lead use from 0.8 gpg to 0.5 gpg effective October 1, 1980.54 Nevertheless, lead exposure remains a serious national problem. Even if the percentage of children with blood lead levels of 30 ug/dl or more has dropped as low as 1%, that still means 150,000 children with lead poisoning.
While urban children are at greatest risk, lead also poses a significant risk to the health of rural children. In the NHANES II survey, 2.1% of rural children, including 10% of rural black children, had blood lead levels of 30 ug/dl or higher (1976-1980 average), as compared to 4.0% of all children.55
Moreover, recent studies suggest that the recognized danger point of 30 ug/dl is too high and that lead reduces intelligence at blood lead levels as low as 10-15 ug/dl. For example, one study found a systematic relationship between intellectual performance and dentine lead level;56 another found “an average difference of seven IQ points between children with blood lead levels of 12 [ug/dl] and below, and those of 13 [ug/dl] and above.”57 Other studies have correlated blood lead levels of 10-15 ug/dl *225with altered brain activity. For example, a recent study showed a linear decrease in the amplitude of “slow” brain waves for all lead levels measured, from 6 ug/dl to 59 ug/dl.58
EPA concluded that these studies of low-level health effects, while not fully conclusive, were “suggestive of a potentially lower maximum safe level of blood lead [than 30 ug/dl].” 47 Fed.Reg. at 38,077/1. Even a moderate cut in the “safe” blood lead level from 30 ug/dl to 25 ug/dl would almost double the number of children with excessive blood lead levels, and a major cut to, say, 10 ug/dl would put half of the children in the U.S. at risk of lead poisoning.59 EPA deferred any attempt to set a new “safe” level pending further research, but stated that the studies supported “the need to take all reasonable steps to control lead emissions.” Id. at 38,077/2. We can find no fault with that conclusion,
iii. Public Comments
EPA’s decision to reduce gasoline lead further was supported by the overwhelming majority of comments from health experts. All 60 comments from independent health professionals opposed EPA’s February proposal to relax the 0.5 gpg standard; many supported a tighter standard.60 For exam-pie, Dr. Glenn Austin, President of the American Academy of Pediatrics, stated:
[W]e now recognize that much lower levels of lead absorption cause impairment of intellectual function in growing children. ... We urge you to retain and further tighten the standards for reduction of lead in the atmosphere from gasoline and all other sources.61
Similarly, all 32 comments from state and local governments opposed loosening the 0.5 gpg standard; the 25 commenters from this group who discussed a special standard for small refiners unanimously opposed such a standard.62
These comments are consistent with a 1980 report by the National Academy of Sciences, Lead in the Human Environment. The National Academy of Sciences found “convincing” evidence of “a significant hazard of detrimental biological effects in children,” 63 and recommended a “serious effort ... to reduce the baseline level of exposure to lead for the general population of the United States.”64 These conclusions deserve special weight because Congress specifically instructed EPA to consider the recommendations of the National Academy of Sciences and, if a regulation “differs in any important respect from any of these recom*226mendations, [to explain] the reasons for such differences.” Clean Air Act § 307(d)(3), 42 U.S.C. § 7607(d)(3).
iv. Conclusion on Health Effects
In sum, the demonstrated connection between gasoline lead and blood lead, the demonstrated health effects of blood lead levels of 30 ug/dl or above, and the significant risk of adverse health effects from blood lead levels as low as 10-15 ug/dl, would justify EPA in banning lead from gasoline entirely. Necessarily, then, there are health benefits from any intermediate reduction.
EPA did not, however, adequately explain the health basis for choosing a 1.10 gplg standard rather than some other standard. An EPA consultant used NHANES II data to estimate that the proposed 1.10 gplg/2.50 gplg standard would produce a mean blood level in 1990 of 7.7 ug/dl for whites and 10.1 ug/dl for blacks.65 However, EPA did not estimate the mean blood level it expected from any other standard, or state what mean blood lead level it wanted to reach.
EPA has been more careful than this in the past (e.g., in promulgating the ambient air quality standard for lead). It had ample data to work from and has little excuse for being so sloppy. Fortunately, we are able in subsection 5 to find a rational basis other than health effects for EPA’s choice of 1.10 gplg as the stopping place.
b. The Health Effects of Small Refiner Lead Use
EPA, in its August proposal, noted that lead emissions are a more serious problem in urban than in rural areas and that small refineries are mostly located in rural areas. It concluded that small refiner lead use had little effect on the health problems of urban children. In October, however, EPA retracted this conclusion, reasoning that small refineries could ship gasoline to urban areas via pipeline.66
SRTF argues that the record does not support EPA’s concern about possible sales to pipelines. In particular, EPA failed to adduce data to rebut SRTF’s own survey of small refiners, which showed that only four small refiners sold leaded gasoline to major Northeast and Midwest pipelines, for a total amount of about 20,000 bpd.67 EPA responds that a mere possibility of sales to pipelines justifies its decision to regulate small refiners.68
As an initial matter, we cannot agree with EPA’s view of how much factual support it needs to justify a decision to regulate. EPA has undoubted power to regulate even where “the evidence [is] difficult to come by, uncertain, or conflicting.” Ethyl Corp., 541 F.2d at 28. But the agency must make a reasonable effort to develop the facts. Where it has not made that effort, EPA cannot regulate on the basis of a guess about what the facts might be. See id. (agency has no power “to act on hunches or wild guesses”).
On the merits of SRTF’s claim, we find, despite some flaws in EPA’s reasoning, sufficient basis in the record for its conclusion that small refiner lead use is a health hazard. Even if small refiners use only a small percentage of total gasoline lead, EPA can reasonably require them to do their part in reducing total lead emissions, so long as the health impact of small refiner lead use is roughly commensurate with their market share. We think EPA has shown that,
i. Sales to Urban Areas
SRTF correctly notes that neither EPA nor the commenters claiming that small refiners had access to pipelines produced any tangible evidence to that effect beyond citing isolated examples.69 Nevertheless, *227SRTF is at least partially hoist on its own data. Its figure of 20,000 bpd is a bare minimum estimate. SRTF only counted sales to major Northeast and Midwest pipelines, and excluded other pipelines.70 Also, SRTF polled only the 64 small “processing” refineries, and excluded 14 small “blenders” that EPA treats as small refineries.71 Despite these distortions, SRTF’s own number shows that 16% of the 124,000 bpd of leaded gasoline produced by small processing refineries is sold to pipelines.72 This is a significant percentage, especially since it is a minimum estimate.
Also, some small refiners are located in or near urban areas and directly sell gasoline to those areas, without using pipelines. EPA did not discuss this point in its notice of final rulemaking, but was certainly aware of it. The issue of direct small refiner sales to urban areas was a major one, and EPA both commissioned an independent contractor to study direct sales and discussed such sales in its notice of proposed rulemaking.73
The contractor, Sobotka & Co., Inc. (“Sobotka”), estimated that small refiners directly sold 33% of their output in metropolitan areas of at least 250,000 population.74 Sobotka did not explain its methodology, and called its estimate “extremely rough and . .. based on little data.”75 Moreover, SRTF criticized the Sobotka study and arrived at a lower estimate of direct sales (23%), and EPA did not explain the discrep*228ancy.76 But even SRTF’s estimate that 23% of small refiner output is sold in urban areas is high enough to refute any claim that small refiner lead use does not threaten the health of urban children.
Moreover, total small refiner sales to urban areas are the sum of direct small refiner sales to nearby urban areas (SRTF estimate = 23%) and small refiner sales to more distant areas by pipelines (SRTF estimate = 16%). Thus, we can estimate that at least 39% of small refiner output, and probably more, is sold in urban areas. Furthermore, this percentage would increase over time under a dual standard. EPA estimated that under a 2.50 gplg/1.10 gplg dual standard, small refiners would have a competitive advantage and would gain market share in the leaded gasoline market at the expense of large refiners. By 1990, they would produce 34% of all gasoline lead compared to 9% in 1982.77 This gain in market share would almost surely portend increased penetration of urban markets now served primarily by large refiners,
ii. Health Effects in Rural Areas
EPA was not concerned solely with small refiner sales to urban areas. Rather, it was also concerned, both in the proposed and final rules, that lead emissions threaten health in rural areas as well. For example, EPA’s notice of proposed rulemaking referred to the “widespread” exposure to lead “among the general population,” 47 Fed. Reg. at 38,072/2, and solicited comments “on the impact of the proposed small refinery definition on lead emissions in specific urban and rural areas,” id. at 38,085/2 (emphasis added).
In fact, rural blood lead levels are only slightly lower than urban blood lead levels. As noted earlier, a significant percentage of rural children have blood lead levels of 30 ug/dl or more.78 Moreover, in the NHANES II survey, rural blood lead levels show the same dramatic decline as urban lead levels in response to reductions in gasoline lead.79 To be sure, EPA did not explicitly focus on this data. Had EPA done so, however, it necessarily would have had to conclude, using the same logic that it applied on a nationwide basis, that gasoline lead use threatens health in rural areas to much the same extent as in urban areas.
In sum, the record supports EPA’s belief that some small refineries sell gasoline to urban areas and also supports the agency’s concern with lead emissions in rural as well as urban areas. Thus, EPA’s ultimate conclusion that small refiner lead use threatens health is reasonable.
In so concluding, we recognize that EPA was less explicit than it could have been in explaining the health dangers of small refiner lead use. We recognize, too, the settled doctrine that neither the courts nor counsel for the agency may “substitute new reasons for those offered by the agency.” Portland Cement Association v. Ruckelshaus, 486 F.2d 375, 395 (D.C.Cir.1973) (citing, e.g., Burlington Truck Lines v. United States, 371 U.S. 156, 168-69, 83 S.Ct. 239, 245, 9 L.Ed.2d 207 (1962)), cert. denied, 417 U.S. 921, 94 S.Ct. 2628, 41 L.Ed.2d 226 (1974); see Weyerhaeuser Co. v. Costle, 590 F.2d at 1030. But this is not a case where EPA failed to give any reasons or gave unsupported reasons for its belief that small refiner lead use threatens health. Rather, *229EPA merely failed to articulate its reasons in any detail, forcing us to dig into the record to understand them fully. In the circumstances of this case, we think that EPA has not crossed “the line from the tolerably terse to the intolerably mute.” Greater Boston Television Corp. v. FCC, 444 F.2d at 852 (footnote omitted).
3. Feasibility
As noted earlier, EPA concluded that most small refiners could meet the 1.10 gplg final standard but the agency was willing to let those who could not meet the standard fall by the wayside.80 SRTF argues that EPA improperly used an aggregate model to predict that most small refiners could meet the standard; in particular, the model relies on an untested scheme for interrefinery trading of lead credits.
EPA did indeed base most of its analysis of the cost and feasibility of a 1.10 gplg standard on an aggregate model developed by the Department of Energy and used by EPA’s contractor Sobotka. That model treated all small refineries as if they were a single small refinery with costs, output, lead use, octane-enhancing capacity, etc., equal to a weighted average for all small refineries. Sobotka modeled large refineries similarly. As it recognized, the model assumes that small refineries can engage in “unlimited” trading of blending components at zero cost. Since interrefinery trading in fact involves some transaction costs, the model “overstate[s]” small refinery ability to produce high-octane gasoline without using lead and understates the cost of meeting a uniform standard.81
EPA concluded that the assumption of interrefinery trading was basically sound. It stated in explaining the final rule that “refineries that are temporarily short of octane producing capability will be able to attain the 1.10 gplg standard ... through purchases of blending stocks or by use of the [lead credit trading] provisions of the regulation.” 47 Fed.Reg. at 49,324/2.
SRTF does not dispute Sobotka’s conclusion that small refiners, in aggregate, have enough octane-enhancing capacity to meet the 1.10 gplg standard. It argues instead that the record does not support EPA’s belief that most individual small refineries can meet the 1.10 gplg standard through interrefinery trading.
a. Timely Objection
Before turning to the merits of EPA’s use of an aggregate model, we address a procedural matter. EPA included Sobotka’s methodology in the record at an early date and referred to that methodology in its notice of proposed rulemaking. See id. at 38,080/3 (Sobotka has calculated costs “using the Department of Energy’s Refinery model”). SRTF, however, failed directly to attack that methodology. EPA argues that SRTF is barred from doing so now by § 307(d)(7)(B), 42 U.S.C. § 7607(d)(7)(B), which provides:
Only an objection .. . which was raised with reasonable specificity during the period for public comment (including any public hearing) may be raised during judicial review.
We agree that SRTF cannot raise at this late date technical objections to the model. On the other hand, EPA retains a duty to examine key assumptions as part of its affirmative “burden of promulgating and explaining a non-arbitrary, non-capricious rule.” National Lime Association v. EPA, 627 F.2d 416, 433 (D.C.Cir.1980); see Inter-*230national Harvester Co. v. Ruckelshaus, 478 F.2d 615, 643 (D.C.Cir.1973). Here, aggregate analysis is a vital assumption underlying the Sobotka model. Thus, EPA must justify that assumption even if no one objects to it during the comment period, and SRTF may properly ask this court to review whether EPA has done so.
b. EPA’s Use of Aggregate Analysis
Any model is an abstraction from and simplification of the real world. Nevertheless, administrative agencies have undoubted power to use predictive models. See Sierra Club v. Costle, 657 F.2d at 332-35. The agency must “explain the assumptions and methodology. used in preparing the model” and, if the methodology is challenged, must “provide a ‘complete analytic defense.’ ” Id. at 333 (quoting American Public Gas Ass’n v. FPC, 567 F.2d 1016, 1039 (D.C.Cir.1977), cert. denied, 435 U.S. 907, 98 S.Ct. 1457, 55 L.Ed.2d 499 (1978)). We will also look for evidence that the agency is conscious of the limits of the model. See id. at 334. Ultimately, however, 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. We can reverse only if the model is so oversimplified that the agency’s conclusions from it are unreasonable.
The key abstraction at issue here is treating all small refineries as a single refinery with characteristics equal to a weighted average of small refineries. This abstraction is reasonable only if, as EPA believed, small refineries that lack octane-enhancing equipment can purchase high-octane blending components or lead credits from better equipped refineries at reasonable cost.82
EPA did not attempt to determine how many small refiners would be able to purchase blending components or at what cost. However, refiners have long been free to buy and sell blending components in the normal course of business, and many do so. “Blenders” produce gasoline entirely by buying, blending components and mixing those components together in appropriate proportions. Moreover, a number of processing refineries produce primarily low-octane components, which they mix with purchased high-octane components to make gasoline. Indeed, as Sobotka noted, the 17 small refineries that lack catalytic reformers already “probably cannot make merchantable gasoline [even] at ... 3 or 4 grams of lead per gallon ... [unless] they use blending components made in other and better equipped refineries.”83
Also, the final rule facilitates a second form of interrefinery trading: Refineries that wish to use more than 1.10 gplg can buy “lead credits” from refineries that use less than 1.10 gplg. Although lead-credit trading was a new idea, EPA had sufficient reason to believe that a market for lead credits would develop. First, the gasoline refining industry is highly competitive. Thus, refiners with excess octane-enhancing capacity will likely be willing to sell lead credits if it is profitable to do so, just as they now sell high-octane blending compo*231nents.84 EPA predicted that the market price for lead credits would settle at about the marginal value of lead to large refiners of one cent per gram.85 Second, EPA’s experience with similar programs — the “offset” and “bubble” policies under the Clean Air Act, and the oil import quota and crude oil “entitlements” programs — suggests that such “paper” trading is likely to develop.86 Third, many commenters supported .interrefinery averaging,87 and most commenters who opposed averaging took for granted that averaging would take place.88
In short, evidence in the record supports Sobotka’s assumption that there will develop a smoothly functioning market in which small refiners will be able to buy blending components, lead credits, or both. Accordingly, we must uphold EPA’s conclusions about the cost and feasibility of the 1.10 gplg standard.
4. Equity and Efficiency
There is little for us to review in EPA’s conclusion that equity and efficiency favor a uniform standard. SRTF does not dispute EPA’s conclusion that a uniform standard is more efficient.89 EPA’s decision that a uniform standard is more equitable than a dual standard is a policy choice that is “not susceptible to the same type of verification or refutation by reference to the record as are some factual questions.” Industrial Union Department, AFL-CIO v. Hodgson, 499 F.2d 467, 475 (D.C.Cir.1974); see Lead Industries Association, 647 F.2d at 1147. Our cases require EPA adequately to explain the reasons for its policy choice, but EPA has done so. It reasoned that a dual standard would give a competitive advantage to small refiners, especially those small refiners who had not already invested in octane-enhancing equipment, and that small refiners had long been on notice that they would have to meet a uniform standard. We must accept those reasons so long as they are rational, and we think they are.
5. Summary
We have separately considered and upheld each of the major conclusions — on health effects, feasibility, and equity and efficiency — underlying EPA’s decision to impose a uniform standard. Necessarily, then, EPA’s justification for the 1.10 gplg standard, taken as a whole, is reasonable, with one caveat adverted to earlier: We are given scant aid by EPA in ascertaining why it chose 1.10 gplg rather than some other uniform standard.
EPA explained that it had balanced health benefits against cost to the industry as a whole. The 1.10 gplg standard would produce lead emissions “equivalent to [emissions] under the existing standards” and, in *232the future, would “result in lower lead levels than the current standards.” 47 Fed. Reg. at 38,070/1. The degree of future reduction was an “accommodation” that “will achieve EPA’s goals of accelerated reductions in lead usage ... without giving rise to unreasonable costs.” Supplemental Response to Comments, supra note 38, at II — 2, J.A. at 1327. Although we continue to be troubled by EPA’s failure explicitly to consider the health benefits of alternative standards, we cannot say that EPA’s balancing of costs against benefits was so seriously flawed that the final rule is unreasonable.
D. Regulatory Flexibility Analysis
Under the Regulatory Flexibility Act, Pub.L. No. 96-354,94 Stat. 1164 (1980) (codified at 5 U.S.C. §§ 601-612), all agencies, as part of the rulemaking process, must conduct a “regulatory flexibility analysis” for any rule that has “a significant economic impact on a substantial number of small entities.” 5 U.S.C. § 605(b). The flexibility analysis must, among other things, discuss how a rule will affect small entities, describe “significant alternatives” that would “minimize any significant economic impact of the rule on small entities,” and explain “why each one of such alternatives was rejected.” Id. § 604(a)(3) (describing the “final regulatory flexibility analysis” to be issued with the final rule); see id. § 603 (describing the “initial regulatory flexibility analysis” to be issued with the proposed rule).
EPA prepared and included in the record both an initial and a final regulatory flexibility analysis. SRTF complains that the final analysis failed to discuss significant alternatives to a uniform standard or explain why those alternatives were rejected. EPA responds that Congress has foreclosed judicial review of agency compliance with the Regulatory Flexibility Act.
The scope of judicial review of a regulatory flexibility analysis is a question of first impression. We conclude after reviewing the statute and the legislative history that we may consider the adequacy of EPA’s regulatory flexibility analysis as part of our overall review of whether the rule is reasonable. On the merits of SRTF’s claims, however, we conclude that EPA has considered a reasonable range of alternatives and given sufficient reasons for rejecting them.
1. Reviewability of the Analysis
Section 611(b) of the Regulatory Flexibility Act expressly limits judicial review of agency compliance with the Act:
Any regulatory flexibility analysis prepared under sections 603 and 604 of this [Act] and the compliance or noncompliance of the agency with the provisions of this [Act] shall not be subject to judicial review.
However, § 611(b) continues:
When an action for judicial review of a rule is instituted, any regulatory flexibility analysis for such rule shall constitute part of the whole record of agency action in connection with the review.
On its face, § 611(b), at a minimum, forbids the courts to review agency compliance with the Regulatory Flexibility Act except as part of their review of the underlying rule. It may mean no more than this, so that a defect in the flexibility analysis would be a factor in the reviewing court’s overall decision whether a rule is reasonable. It may also mean, as EPA claims, that a reviewing court can only consider in its overall decision whatever flexibility analysis the agency sees fit to prepare, and cannot overturn a rule because the regulatory flexibility analysis is grossly inadequate or even totally absent. Neither result is compelled by the statutory language. For further guidance, we therefore turn to the legislative history.
Congress adopted the Regulatory Flexibility Act in an unusual fashion which limits the value of some of the traditional guides to legislative intent. Its immediate precursors were H.R. 4660, 96th Cong., 1st Sess. (1979) and S. 299, 96th Cong., 2d Sess., 126 Cong.Rec. S10,930-31 (daily ed. Aug. 6, 1980), neither of which restricted judicial *233review of agency compliance with the Act.90 However, the Senate rejected S. 299 as reported by the Judiciary Committee and adopted instead a substitute bill offered by Senator Culver on the Senate floor. See 126 Cong.Rec. S10,931-33 (daily ed. Aug. 6, 1980).91
Senator Culver offered along with the bill an extensive section-by-section analysis of the bill akin to a committee report. Id. at S10,934-43. We think it appropriate to give substantial weight to this description of the bill. Indeed, since the Senate did not debate the judicial review provisions of the Act, we have little else to go on in ascertaining its intent.
Senator Culver’s substitute report explained that § 611(b) was meant to ensure that agencies obey the Act by permitting courts to review the flexibility analysis as part of their overall review of the reasonableness of a rule. On the other hand, it avoided disruption of the rulemaking process by precluding courts from separately reviewing the flexibility analysis:
In any judicial review, the issue before the court will remain whether the rule itself is valid.... In making that determination, the court will consider the contents of the regulatory flexibility analysis along with other relevant material.
. .. [T]he bill strikes a balance between two central aims .... The first is to ensure that an agency’s compliance with the objectives of this bill be subject to meaningful ... judicial oversight.. ..
On the other hand, the bill avoids the substantial disruption of agency rulemaking inherent in allowing separate judicial review of the regulatory flexibility analysis itself in the manner of an environmental impact statement.
Id. at S10,939. Senator Culver also made it clear that a major error in the regulatory flexibility analysis may be, but does not have to be, grounds for overturning a rule:
[I]f an agency ... completely and consciously ignores the ... requirement to perform regulatory flexibility analyses, an injured party would have grounds to argue that this fact is evidence of the unreasonableness of the rule.... In addition, if an agency completely fails to respond to a clearly available significant alternative ... raised in public comments, then this failure would be grounds for finding the rule unreasonable under established case law.92
The Senate bill was brought directly to the House floor and adopted without change. The House apparently accepted the Senate view that a defect in the analysis may lead a court to conclude that the rule is unreasonable. For example, Representative Butler, a primary sponsor of the now superseded House bill, explained:
Judicial review of the lack of or adequacy of the regulatory flexibility analysis is permitted to the extent it is relevant to a review of the validity of the final rule.... [T]he failure to perform an analysis of a reasonable alternative consistent with the desired regulatory goal may be adequate grounds for a determination that the final rule is unreasonable.
Id. at H8472-73 (daily ed. Sept. 8, 1980); see id. at H8459 (statement of Rep. Kastenmeier):
The bill’s treatment of judicial review is intended to strike a balance between two necessary goals. First, to [avoid interlocutory review]; and Second, the desire to insure that agencies take seriously their obligation under the law by providing for *234review of regulatory flexibility analyses as part of the entire record.93
We hold, based on this history, that a reviewing court should consider the regulatory flexibility analysis as part of its overall judgment whether a rule is reasonable and may, in an appropriate case, strike down a rule because of a defect in the flexibility analysis. We turn, therefore, to the merits of SRTF’s objections.
2. Adequacy of the Analysis
In its Final Regulatory Flexibility Analysis, EPA analyzed five alternatives: a uniform 0.50 gpg pooled standard; a uniform 0.40 gpg pooled standard; a uniform 1.10 gplg leaded-only standard; a 1.10 gplg/1.90 gplg dual standard; and a uniform 1.10 gplg standard with a “phase-in” period for small refiners. Final Regulatory Flexibility Analysis, supra note 84, at 3, J.A. at 1421. SRTF complains that EPA failed to consider less restrictive standards such as a 1.10 gplg/2.50 gplg dual standard.
This complaint is frivolous. Taking the rulemaking as a whole, EPA considered an appropriate range of alternatives, including the 1.10 gplg/2.50 gplg dual standard. EPA discussed that standard at length in its statement of basis and purpose and in its Supplemental Response to Comments. We concluded in subsection C that EPA properly rejected it. EPA should have analyzed that option in its regulatory flexibility analysis as well, but its failure to do so is a purely technical flaw that does not affect the reasonableness of the final rule.94
E. The July 1, 198S Effective Date
Plateau argues that the July 1, 1983 effective date for the 1.10 gplg standard does not give it enough time to install the new equipment it needs to meet the standard. It explains that it had been building the needed equipment but stopped when EPA issued its February notice of proposed rule-making and cannot now finish the aborted construction by July 1. We conclude that EPA’s decision to grant an eight-month period between date of promulgation and effective date was reasonable.
EPA recognized that its February notice gave small refiners “some reason to believe that they might not be subject to the same standard as large refineries” and that small refiners “may therefore have delayed projects designed to meet a more stringent standard.” 47 Fed.Reg. at 49,324/1. It granted an eight-month delay in the effective date to compensate for the eight months of uncertainty between the February notice and the October final rule. Id. at 49,324/2.
Plateau claims that it will have to start all over with a new “isomerization project” that will take two years to complete because its previous plan to install a “reformer” will not suffice to meet the “severe” final standard. We fail to understand the factual basis for this claim. Because small refiners produce mostly leaded gasoline, the 1.10 gplg final standard permits them to use substantially more lead than the 0.5 gpg standard it replaces. Thus, nothing EPA has done has forced Plateau to make new plans.
*235To be sure, start-up delays may in some cases make it difficult to complete a mothballed project within the originally scheduled time. However, even then, small refiners that eventually plan to install new equipment can probably comply in the interim by purchasing blending components or lead credits. EPA recognized and relied on this fallback possibility:
Even if the eight-month period is too short to permit compliance through construction programs in some cases, refineries that are temporarily short of octane producing capability will be able to attain the 1.10 gpg standard ... through purchases of blending stocks or by use of the averaging provisions of the regulation.
Id. at 49,324. Especially in light of this fallback option, EPA’s choice of an eight-month delay in the effective date to compensate for the eight-month duration of the rulemaking proceeding strikes us as sensible.95
F. Late Docketing of Comments
SRTF also objects to the final rule on a procedural ground — EPA relied on evidence that the agency had added to the record either near the end of or after the close of the comment period and too late for effective rebuttal. In particular, EPA docketed a number of Sobotka studies on October 7 and 8 (the comment period closed on October 8), and docketed several other Sobotka studies on October 27, only two days before the final rule was issued.
This late docketing was highly improper. We have held that EPA may, in an appropriate case, add documents submitted by others to the docket after the close of the comment period. Sierra Club v. Costle, 657 F.2d at 396-400. Here, however, EPA itself added evidence to the record both near the end of and after the close of the comment period, and presumably intended to rely on that evidence. Interested parties had little or no opportunity to reply. Sierra Club prohibits such conduct in no uncertain terms:
If ... documents . .. upon which EPA intended to rely had been entered on the docket too late for any meaningful public comment ..., then both the structure and spirit of section 307 would have been violated.
Id. at 398; see Kennecott Corp. v. EPA, 684 F.2d 1007, 1019 (D.C.Cir.1982) (EPA improperly placed economic forecast data in the record only one week before issuing its final regulations).
We do not go so far as to hold that it is per se improper for EPA to add evidence to the record at the end of or after the close of the comment period. EPA may sometimes be able to show that the late entry did not foreclose an opportunity for “meaningful public comment.” For example, it might be proper for EPA to develop new evidence in order to respond to a particular comment, so long as it gives the commenter an opportunity to reply to the new evidence. Cf. id. at 399 (Environmental Defense Fund was aware that industry group had submitted new data after the close of the comment period, had “many weeks” to respond, and in fact did so). Here, however, EPA has not shown that it afforded any rebuttal opportunity.
It is also incumbent upon a petitioner objecting to the agency’s late submis*236sion of documents to indicate with “reasonable specificity” what portions of the documents it objects to and how it might have responded if given the opportunity. Cf. Clean Air Act § 307(d)(7)(B), 42 U.S.C. § 7607(d)(7)(B) (objections to a rule must be raised with “reasonable specificity during the period for public comment ... [unless] it was impracticable to raise such objection within such time”); American Petroleum Institute v. Costle, 665 F.2d 1176, 1191 (D.C.Cir.1981) (requiring parties to raise objections which arise after the end of the comment period with “reasonable specificity” in a petition to the agency for reconsideration), cert. denied, 455 U.S. 1034, 102 S.Ct. 1737, 72 L.Ed.2d 152 (1982). Here, however, SRTF raises a number of specific objections to the Sobotka studies.96 We therefore turn to the standard for reviewing those objections.
In general, there are two possible sources of reversible error from EPA’s late docketing of evidence. First, since EPA did not provide an opportunity to reply to the late evidence, we should not, we think, consider any evidence to which a petitioner objects in determining whether there is enough evidence in the record to support the final rule. In reaching our conclusion that the 1.10 gplg standard is reasonable, we have not relied on the portions of the Sobotka studies to which SRTF objects. Necessarily, then, any controversial portions of the late-submitted studies are not “vital to EPA’s support for the rule.” Sierra Club, 657 F.2d at 399.
Second, it is possible that although we can find support for the rule without relying on the late evidence, the agency may in fact have relied on that evidence to reach a different result than it would otherwise have reached. To convince us to reverse on this ground, SRTF must show that there is a “substantial likelihood that the rule would have been significantly changed” had it been given a rebuttal opportunity. Clean Air Act § 307(d)(8), 42 U.S.C. § 7607(d)(8).
In this case, EPA did refer briefly to some of the late studies in its final statement of basis and purpose. See, e.g., 47 Fed.Reg. at 49,324/2 (small refiners can produce gasoline at 1.40 gplg at about the same cost as large refiners at 1.10 gplg). Thus, the studies played some role in EPA’s decision — how much, it is hard to say. However, the same information was available (though in somewhat less detail) from the timely-submitted comments of the Department of Energy, on which EPA could properly have relied.97 Moreover, SRTF’s specific complaints are grounded primarily on a single underlying objection to Sobotka’s use of aggregate analysis. SRTF had ample opportunity to object to that modeling assumption, and we have already considered and rejected its objections. Thus, there is not a “substantial likelihood” that SRTF, if given the chance to comment on the additional Sobotka studies, could have convinced EPA to choose a different standard.
We conclude that EPA has committed error, but not reversible error.
G. Conclusion
In places the record is sparse and EPA’s analysis weak. We have had to search for scraps of hard data or indirect evidence that the agency even thought about major issues. But the ultimate standard is not whether the agency’s analysis is impeccable, but whether it is reasonable; not whether most of the evidence supports the agency’s position but whether enough of it does; not whether the agency’s policy choices are wise but whether they are rational. Against that high standard, SRTF has the burden of convincing us that EPA improperly declined to subsidize small refiners by allowing them to spew more lead into the air than their *237competitors. In that task, it has failed. The 1.10 gplg final standard is reasonable.
V. The Interim 1.90 gplg Standard
To review the relevant aspects of the final rule, prior EPA regulations made small refiners subject to a sliding-scale lead content standard until October 1, 1982 and to a 0.5 gpg standard thereafter. In February 1982, EPA proposed to relax the 0.5 gpg standard and also proposed to suspend the October 1, 1982 deadline pending completion of the rulemaking. The suspension, EPA explained, would permit small refiners to defer investing in equipment needed to meet the 0.5 gpg standard:
Because attainment of the 0.5 gpg standard will likely require substantial capital investment on the part of small refineries in the near future ..., the effective date of that standard for small refineries should be postponed pending the outcome of the analysis on the entire lead phase-down issue.
47 Fed.Reg. at 7814/2. EPA unequivocally promised small refiners that the final rule would give them enough time to construct any new equipment they might need:
Any final decision regarding standards for small refineries will take into account the lead time required for construction of any processing equipment needed for compliance.
Id. (emphasis added).
In August 1982, EPA proposed a permanent 2.50 gplg standard for small refiners. It suspended the October 1 compliance date for the 0.5 gpg standard for only one month, but explained that it would issue a final rule before November 1. EPA expected and wanted small refiners to continue to defer capital investment needed to meet the 0.5 gpg standard:
[T]he basis for this suspension is the need to defer the necessity for small refineries to decide whether to make the substantial expenditures required to meet a more stringent lead standard during the time that the Agency is still considering amendments to the lead phasedown program.
Id. at 38,091/1.
On October 29, 1982, EPA promulgated a 1.10 gplg final standard for small refiners effective July 1,1983 and a 1.90 gplg interim standard effective November 1,1982. It explained that small refiners currently averaged 2.17 gplg and that the 1.90 gplg standard would require “only a slight decrease in average lead use” and “should generally be achievable by [small] refineries through use of the averaging provisions in the regulations.” Id. at 49,324/2.
SRTF and Plateau argue that EPA did not give adequate notice that it might impose with no lead time a standard significantly stricter than the sliding-scale standard and that the record does not support EPA’s conclusion that most small refiners could meet the interim standard.
A. Notice
As an initial matter, the 1.90 gplg standard is significantly stricter than the proposed 2.50 gplg standard, which was itself stricter than the sliding-scale standard.98 Thus, it does not satisfy EPA’s promise to the small refiners that they could safely delay capital investments pending EPA’s issuance of a final rule. EPA implicitly recognized as much when it explained that small refiners “generally” could meet the standard “through averaging.” The question, then, is whether anything in the proposed rulemaking gave adequate notice that EPA might issue such a strict interim standard. We think not.
*238We discuss in part VI infra what constitutes adequate notice under Clean Air Act § 307(d). For present purposes, we need merely note that, as under the APA, the final rule must be a “logical outgrowth” of the agency’s proposal. United Steelworkers v. Marshall, 647 F.2d 1189, 1221 (D.C.Cir.1980), cert. denied, 453 U.S. 913, 101 S.Ct. 3148, 69 L.Ed.2d 997 (1981); Weyerhaeuser Co. v. Costle, 590 F.2d 1011, 1031 (D.C.Cir.1978).
EPA argues that the 1.90 gplg interim standard was a logical outgrowth of the August proposal because it was “within the range of alternative standards being considered by EPA.”99 The August notice of proposed rulemaking expressed concern that the proposed 2.50 gplg standard might be too high and invited comments on “whether a standard such as 2.15 gp[l]g . .. may be more appropriate.” 47 Fed.Reg. at 38,082/1. The 1.90 gplg standard, EPA reasons, is encompassed by the phrase “such as 2.15 gplg.”
This argument misses the point. EPA has argued that the August proposal gave fair notice of a 1.90 gplg final standard. The issue, however, is whether the August proposal gave notice of a 1.90 gplg interim standard, effective immediately. On that issue, EPA did not give fair notice. On the contrary, EPA stated unequivocally, both in February and August, that its final rule “will” give small refiners who wish to meet the new standard by constructing new equipment enough lead time to do so. It never hinted that it might immediately require small refiners to meet a strict interim standard.100 Unsurprisingly, therefore, no one commented on what interim standard would be feasible.
Intervenors argue that the small refiners had no right to rely on EPA’s statements that it would give adequate lead time, and that the small refiners should have prepared for the possibility that EPA would allow the 0.5 gpg standard to take effect on November 1,1982. Therefore, the small refiners cannot complain about the much less stringent 1.9 gplg interim standard.101
However, even if EPA could have allowed the 0.5 gpg standard to take effect, it did not do so; instead, it promulgated a new 1.90 gplg interim standard. EPA could not validly issue that new standard without first providing notice and an opportunity to comment. Cf. Weyerhaeuser Co. v. Costle, 590 F.2d at 1031 n. 27 (EPA’s failure to give notice deprived the industry of an opportunity to comment on key facts; it is beside the point that EPA might have been able to justify the rule even using the industry’s figures).
Second, we doubt the validity of intervenors’ premise — that EPA could have allowed the 0.5 gpg standard to take effect on November 1. Unless EPA had a strong reason for breaking its promise of adequate lead time, a court might well have concluded that its action was arbitrary.
The question remains whether EPA’s failure to give notice meets the standard for reversal on procedural grounds laid out in Clean Air Act § 307(d)(8), 42 U.S.C. § 7607(d)(8) — “a substantial likelihood that the rule would have been significantly changed if such errors had not been made.” As we concluded in part III.D supra, where the procedural error would have been reversible error under the APA, § 307(d)(8) *239does not restrict our power, indeed our duty, to reverse EPA’s action on procedural grounds. Under that test, EPA’s failure to give notice on a major portion of a rule is reversible error.102
B. Feasibility
We also do not find substantial evidence to support EPA’s belief that most small refiners could meet the 1.90 gplg interim standard at reasonable cost. As noted above, EPA reasoned that small refiners currently averaged 2.17 gplg and that 1.90 gplg was only a “slight decrease” that “should generally be achievable through use of the averaging provisions in the regulations.” We have two problems with this analysis.
First, the 2.17 average is not representative of what most individual small refiners can achieve. Most small refineries are processing refineries, and EPA’s own data show that small processing refineries averaged 2.36 gplg.103 Only by including blenders could EPA arrive at 2.17. Moreover, SRTF cogently objects (and EPA fails to respond) that EPA’s data include three refineries that do not qualify as small, and that corrected data would show that small processing refineries averaged 2.50 gplg. This would be consistent with SRTF’s own evidence — that of the 64 small processing refineries, 40 used 2.50 gplg or more, and 14 of these used 3.00 gplg or more.104 Thus, to meet the 1.90 gplg interim standard, a typical small refiner would have to reduce lead use by much more than EPA believed.
Second, EPA recognized that small refiners could generally meet the interim standard only through averaging. We concluded in part IV.C.3.b supra that the record supports EPA’s belief that a market for lead credits would develop. But EPA imposed the interim standard with only two days lead time. Nothing in the record supports EPA’s implausible belief that a lead credit market would spring up overnight.
We recognize that the time lag before a lead credit market will develop is not susceptible of proof by hard evidence. Had EPA considered the issue, we would be reluctant to second-guess its judgment. But, so far as we can tell, EPA never did so. Instead, it blithely assumed that lead credits would be readily available.105 We conclude that EPA improperly relied on lead-credit trading as a reason to believe that small refiners could immediately meet the interim standard.106
Both of these flaws are serious. Taken together, they undermine the principal bases for EPA’s conclusion that the 1.90 gplg standard is feasible. Moreover, their import is magnified by the lack of an opportunity for public comment, which “undermines our usual assumption that notice and comment rulemaking, by virtue of its acces*240sibility to public scrutiny, will achieve rational results.” Weyerhaeuser Co. v. Costle, 590 F.2d at 1031. We conclude that the interim standard is not supported by the evidence in the record.107
C. What Standard Does EPA Return To?
In light of the foregoing, we vacate the portion of 40 C.F.R. § 80.20(b)(l)(i) that requires small refineries to limit gasoline lead content to 1.90 gplg for gasoline production up to the refinery’s past production level. We leave in force the requirement in § 80.20(b)(l)(i) that leaded gasoline produced in excess of a refinery’s past production level may not exceed 1.10 gplg.
When an agency replaces an existing rule with a new rule, and we vacate all or part of the new rule, we must decide whether the prior rule continues in effect or whether our action leaves no rule in effect. In at least some cases, we have no power to reinstate the prior rule and must remand to the agency to determine the appropriate replacement. See Burlington Northern, Inc. v. United States, — U.S. —, 103 S.Ct. 514, 74 L.Ed.2d 311 (1982) (Court of Appeals lacks power to reinstate old ICC rate after invalidating revised rate). On the other hand, we have sometimes assumed that a decision vacating a new rule will reinstate the old rule. E.g., Natural Resources Defense Council, Inc. v. Gorsuch, 685 F.2d 718, 728 (D.C.Cir.1982) (implicitly assuming that EPA will return to its previous regulation defining “source” under the Clean Air Act).
We believe the better course is generally to vacate the new rule without reinstating the old rule. This avoids any problem of the court overstepping its authority, and leaves it to the agency to craft the best replacement for its own rule. Failure to reinstate the old rule creates a temporary regulatory vacuum, but the court’s power to delay the mandate in a case for a reasonable period, plus the agency’s limited power to issue a replacement on an emergency basis without notice and comment, should usually suffice to avoid untoward consequences.
We think it appropriate to follow this general rule here.108 Indeed, it would be irrational to order the agency to return to the old 0.5 gpg standard for small refiners (which would have become effective on November 1, 1982) because that standard is stricter than both the invalidated 1.90 gplg interim standard and the 1.10 gplg final standard. We therefore vacate the interim standard and leave it to the agency to choose a suitable replacement.
Our decision leaves EPA without a regulatory limit on small refinery lead use. The unexpected nature of this regulatory vacuum, plus the public health danger posed by unrestricted lead use and the absence of any unfairness to small refiners in such a course of action, would justify EPA in immediately promulgating a temporary lead content regulation which does not require small refiners to reduce lead use to a level *241significantly below previous lead-use levels. See 5 U.S.C. §§ 553(b)(B) (agency may for “good cause” issue rules without notice or public procedure) (made applicable to Clean Air Act by 42 U.S.C. §§ 7607(d)(1) (last sentence)), 553(d)(3) (“good causé” exception to 30-day period between publication of a rule and the rule’s effective date); American Federation of Government Employees v. Block, 655 F.2d 1153, 1156 (D.C.Cir.1981) (these exceptions to the § 553 procedures permit agency to issue temporary rules in “emergency situations”).109
The last paragraph is taken almost verbatim from our order of January 26, 1983. See Appendix I. In that order, we delayed issuing the mandate for one week to give EPA an opportunity to issue such an emergency rule. In response, EPA established a new interim standard equal to the old sliding-scale standard, and made assorted conforming changes in the fuel additive regulations. 48 Fed.Reg. 5724 (1983). This seems to us to be an entirely appropriate resolution, and confirms our decision not to reinstate the old rule.
VI. The Definition of “Small Refinery”
We turn next to the basis for EPA’s definition of “small refinery.” To review the relevant regulatory background, in 1977, Congress had defined “small refinery” in terms of capacity limits. A refinery, to qualify as “small,” needed to have crude oil capacity of 50,000 bpd or less and could not be owned or controlled by a refiner with total crude oil capacity of 137,500 bpd or more. Clean Air Act § 211(g)(1)(B), 42 U.S.C. § 7545(g)(1)(B).
In August 1982, EPA proposed to redefine “small refinery” in terms of production limits. Under the proposal, a “small” refinery could not have produced more than 10.000 bpd during the most recent calendar quarter and could not be owned or controlled by a refiner with total production during the most recent calendar quarter of 70.000 bpd or more [hereinafter a “large refiner”]. 47 Fed.Reg. at 38,083-85. The agency also requested comments on means to ensure that the new definition would not permit small refineries to increase current production levels:
The standard is not intended to allow small refineries to expand their production of leaded gasoline above their current levels, and thereby increase total emissions of lead into the environment. The Agency invites comments on methods to assure that such a proliferation of leaded gasoline does not occur.
Id. at 38,085/3.
In October, EPA revised the definition of small refinery to include only refineries that had produced no more than 10,000 bpd during each calendar quarter since July 1, 1981 (“past production requirement”) and were not owned or controlled since July 1, 1981 by a large refiner (“past ownership requirement”). The past production requirement, it explained, would prevent refineries that could produce more than 10,-000 bpd from “backsliding” — reducing production to less than 10,000 bpd in order to qualify as “small” refineries. Id. at 49,325/2. The past ownership requirement, EPA explained, would prevent a large refiner that owns a small refinery from selling the small refinery so that the newly independent small refiner could use more lead. Id. at 49,325-26.
SRTF claims that EPA did not give adequate notice of the past production requirement and Simmons claims that EPA did not give adequate notice of the past ownership requirement.
A. The Standard for Adequate Notice
EPA undoubtedly has authority to promulgate a final rule that differs in some particulars from its proposed rule. As we noted in International Harvester Co. v. Ruckelshaus, 478 F.2d 615, 632 n. 51 (D.C.Cir.1973), “[a] contrary rule would lead to the absurdity that ... the agency can learn *242from the comments on its proposals only at the peril of starting a new procedural round of commentary.” However, if the final rule deviates too sharply from the proposal, affected parties will be deprived of notice and an opportunity to respond to the proposal.
Courts have devised various verbal formulas for the extent to which an agency can make changes in the final rule that were not clearly presaged by the notice of proposed rulemaking. This court has held, both under the APA and under Clean Air Act § 307(d), that the final rule must be a “logical outgrowth” of the proposed rule. Sierra Club v. Costle, 657 F.2d at 352 (Clean Air Act); United Steelworkers v. Marshall, 647 F.2d at 1221 (APA); accord BASF Wyandotte Corp. v. Costle, 598 F.2d 637, 642 (1st Cir.1979) (APA), cert. denied, 444 U.S. 1096, 100 S.Ct. 1063, 62 L.Ed.2d 784 (1980); Taylor Diving & Salvage Co. v. United States Department of Labor, 599 F.2d 622, 626 (5th Cir.1979) (APA).110 The Third Circuit has phrased the test as whether the agency’s notice would “fairly apprise interested persons of the ‘subjects and issues’ [of the rulemaking].” American Iron & Steel Institute v. EPA, 568 F.2d 284, 293 (3d Cir.1977); accord Consolidation Coal Co. v. Costle, 604 F.2d 239, 248 (4th Cir.1979), rev’d on other grounds sub nom. EPA v. National Crushed Stone Association, 449 U.S. 64, 101 S.Ct. 295, 66 L.Ed.2d 268 (1980).
These general formulas, however, more rephrase than answer the underlying question of how much notice is enough. In any particular case, the answer to that question must turn on how well the notice that the agency gave serves the policies underlying the notice requirement. Notice, as we see it, serves three distinct purposes.
First, notice improves the quality of agency rulemaking by ensuring that agency regulations will' be “tested by exposure to diverse public comment.” BASF Wyandotte Corp., 598 F.2d at 641. Second, notice and the opportunity to be heard are an essential component of “fairness to affected parties.” National Association of Home Health Agencies v. Schweiker, 690 F.2d 932, 949 (D.C.Cir.1982). Third, by giving affected parties an opportunity to develop evidence in the record to support their objections to a rule, notice enhances the quality of judicial review. See Marathon Oil Co. v. EPA, 564 F.2d 1253, 1271 n. 54 (9th Cir.1977) (“Such comment is often an invaluable source of information to a reviewing court attempting to evaluate complex statistical and technological decisions.”).
Against these values, we must balance the public interest in expedition and finality. The notice requirement should not force an agency endlessly to repropose a rule because of minor changes, nor should a court vacate and remand an otherwise reasonable rule because of a minor procedural flaw. With these competing policies in mind, we turn to the particular notice problems posed by the past production and past ownership requirements.
B. Past Production Requirement
The basic concern that EPA expressed in its August notice was that the new definition — framed in terms of production limits rather than capacity limits— might contain loopholes that would allow small refiners to increase lead use over current levels. EPA’s request for comments on how to prevent increases in lead use, on its face, was directed primarily at production increases by already qualifying small refineries rather than increases in the number of refineries that qualify as “small.” But we think that, in context, EPA’s request can fairly' be read also to encompass loopholes that would permit refineries to alter production in order to qualify as small. In response, numerous commenters noted a loophole of the latter sort — refiners that produced slightly more than 10,000 bpd would find it profitable to cut production to just under 10,000 bpd to take advantage of *243a dual standard.111 EPA responded by adding the past production requirement to the definition of “small refinery.”
We think that SRTF received adequate notice of the past production requirement. It is true that the proposal did not list specific “loopholes” that EPA might try to close. Without any specific indication of what changes EPA might try to make, commented were hampered in effectively opposing those changes. See Home Box Office, Inc. v. FCC, 567 F.2d 9, 36 (D.C.Cir.) (agency must “make its views known to the public in a concrete and focused form so as to make criticism or formulation of alternatives possible”), cert. denied, 434 U.S. 829, 98 S.Ct. 111, 54 L.Ed.2d 89 (1977). This shortcoming is especially important in light of Congress’ intent, expressed in § 307(d), that EPA provide a detailed proposal for interested parties to focus their comments on.112
Nevertheless, we think SRTF was on adequate notice of the past production requirement, for two independent reasons. First, SRTF was aware generally that EPA was seeking to identify and close as-yet-undefined loopholes that might affect SRTF’s members. We think SRTF was therefore obliged to take reasonable steps — a letter of inquiry to EPA ought to have sufficed — to keep informed of EPA’s thinking on this matter. Had SRTF done so, EPA presumably would have informed it that other commented had proposed a past production requirement.113
Second, SRTF was almost surely in fact aware that United Refining Co. and others had proposed a past production requirement. SRTF’s attorney attended the public hearing and thus presumably heard United propose the requirement at the hearing. Also, SRTF was closely involved in the entire rulemaking, was monitoring the docket, and responded to other information entered in the docket to which it objected.114
C. Past Ownership Requirement
We are unable, however, to find adequate notice of the past ownership requirement. The requirement derives from a single comment, by Asamera Oil, noting that large refiners might sell small refining facilities to permit the small facilities to qualify for a higher lead standard, and that two large refiners had recently sold small facilities.115 EPA responded by adding the past ownership requirement. Id. at 49,325-26.
1. Notice from EPA
In contrast to the past production requirement, the past ownership requirement does not close a loophole created by EPA’s switch from capacity limits to production limits. Large refiners would have the same opportunity to sell small facilities either way. The rule, seen with the benefit of hindsight, makes some sense because it closes a loophole of sorts. But the test, imper*244fectly captured in the phrase “logical outgrowth,” is whether Simmons, ex ante, should have anticipated that such a requirement might be imposed. We think not. The connection between EPA’s request for comments and the past ownership requirement is simply too tenuous.
EPA also argues that it gave general notice that it might make unspecified changes in the definition of small refinery. This purported notice, however, is too general to be adequate. Agency notice must describe the range of alternatives being considered with reasonable specificity. Otherwise, interested parties will not know what to comment on, and notice will riot lead to better-informed agency decision-making. See Home Box Office, 567 F.2d at 36 (quoted in subsection B supra); Rodway v. United States Department of Agriculture, 514 F.2d 809, 814 (D.C.Cir.1975) (general reference to changes in food stamp program is inadequate notice of particular change); S.Rep. No. 752, 79th Cong., 1st Sess. 14 (1945), reprinted in Senate Judiciary Committee, Administrative Procedure Act: Legislative History 187, 200 (Comm. Print 1946) (“APA Legislative History”) (notice must “fairly apprise interested persons of the issues involved”).116
This is doubly true under Clean Air Act § 307(d)(3), which requires EPA to issue a specific “proposed rule” as a focus for comments. We note that the requirement of “reasonable specificity” merely holds EPA to the same standard that private parties must meet in objecting to a rule. See § 307(d)(7)(B), 42 U.S.C. § 7607(d)(7)(B) (parties must raise objections to a rule “with reasonable specificity during the period for public comment”).
2. Actual Notice
EPA also points out that Platt’s Oil Marketing Bulletin discussed Asamera’s comment and argues that this source put Simmons on actual notice of the past ownership requirement.117 Our cases recognize that even if the agency has not given notice in the statutorily prescribed fashion, actual notice will render the error harmless. See, e.g., Sierra Club v. Costle, 657 F.2d at 355, 360, 398-99; cf. 5 U.S.C. § 552(a) (“Except to the extent that a person has actual and timely notice ... thereof, a person may not in any manner be required to resort to, or be adversely affected by, a matter required to be published in the Federal Register and not so published.”). Here, however, EPA proffers no proof that Simmons had actual notice, only speculation. Simmons denies receiving actual notice,118 and we have no basis for doubting its denial.
3. Notice by Others
EPA also argues that Simmons was under a duty to monitor the docket and that if it had done so, it would have seen Asamera’s comment and realized that it ought to respond. We disagree. As a general rule, EPA must itself provide notice of a regulatory proposal. Having failed to do so, it cannot bootstrap notice from a comment.
The APA does not require comments to be entered on a public docket. Thus, notice necessarily must come—if at all—from the agency. See S.Rep. No. 752, supra, at 14, APA Legislative History at 200 (“Agency notice must be sufficient to fairly apprise interested parties of the issues involved ....”) (emphasis added); cf. Wagner Electric Corp. v. Volpe, 466 F.2d 1013, 1019 (3d Cir.1972) (that some “knowledgeable manufacturers” responded to an inadequate no*245tice with comments relating to the final rule “is not relevant. Others [were] possibly not so knowledgeable .... ”).
We see no warrant for applying a different rule under Clean Air Act § 307(d) merely because § 307(d) provides for a public record. Section 307(d), like the APA, is phrased in terms of the notice that the agency must give. Congress established a' clearly defined record to facilitate judicial review and to enhance agency decisionmaking. See Sierra Club v. Costle, 657 F.2d at 392-95; Pedersen, Formal Records and Informal Rulemaking, 85 Yale L.J. 38, 70-74 (1975). Nothing in the text or legislative history of the statute hints that Congress intended to require private parties to scrutinize the record for notice of issues that the agency did not raise itself.
On the contrary, the additional notice requirements in § 307(d)(3) suggest that Congress intended agency notice under the Clean Air Act to be more, not less, extensive than under the APA. The legislative history supports this view. See, e.g., H.R. Rep. at 319,4 Leg.Hist. 2786,1977 U.S.Code Cong. & Ad.News at 1398 (the new procedures will “insure an effective opportunity for public participation in the rulemaking process”); 123 Cong.Rec. 27,075 (1977), 3 Leg.Hist. 333 (statement of Rep. Broyhill) (the new procedures “will assure the opportunity for more extensive public participation in the rulemaking process”).
Finally, EPA’s construction would ill-serve the purposes behind the notice requirement. It would turn notice into an elaborate treasure hunt, in which interested parties, assisted by high-priced guides (called “lawyers”), must search the record for the buried treasure of a possibly relevant comment. Inevitably, many parties will not attempt this costly search and many others will fail in their search. The agency will not get the informed feedback it needs, the parties will feel unfairly treated, and there will be a meager record for us to review.
This case well illustrates these problems. Simmons, a small Montana refiner, would have had to hire Washington counsel to read over 1100 written comments in search of two relevant pages tucked away in the middle of a single comment, on an issue that EPA gave no notice of. Unsurprisingly, Simmons failed to notice or to rebut Asamera’s comment.
Had Simmons been on notice that EPA was considering a past ownership requirement, it could have advanced several arguments against the requirement. First, Simmons could have argued that large refiners would not sell small facilities merely to take advantage of the short period remaining until July 1,1983, after which small refiners must meet the same lead standard as large refiners. Simmons might also have argued that February 22,1982, the date when EPA initially proposed relaxing the lead-content •limits for small refiners, was a better cutoff date than July 1, 1981, because there was no incentive for large refiners to sell small facilities prior to the February proposal.
On a factual level, Simmons could have noted that it bought its refinery on January 4, 1982, at a time when it expected to have to meet a uniform 0.5 gpg standard beginning on October 1,1982. It could also have pointed out that it was the only small refinery to have been sold by a large refiner (EPA incorrectly thought there were two119).
We do not know, of course, how EPA would have responded to these comments. But whether or not notice would have led to a different rule, it would have permitted EPA to get its facts straight. It also would have forced EPA to focus on one key issue — the choice of cutoff date — on which the notice of final rulemaking is conspicuously silent.
*2464. The July 1, 1981 Cutoff Date
EPA’s silence on why it chose July 1, 1981 as a cutoff date provides an independent reason for vacating the past ownership requirement. A rule without a stated reason is necessarily arbitrary and capricious. See, e.g., Action on Smoking & Health v. CAB, 699 F.2d 1209, 1219 (D.C.Cir.1983) (remanding for further proceedings “the three proposals ... that the Board disposed of without reasons”); State Farm Mutual Automobile Insurance Co. v. Department of Transportation, 680 F.2d 206, 231 (D.C.Cir.) (“There must be some reason to support any reasoned decision.”) (emphasis in original), cert. granted, — U.S. —, 103 S.Ct. 340, 74 L.Ed.2d 382 (1982). Agency counsel’s attempted explanation, of course, comes too late.
5. Conclusion
For the foregoing reasons, we vacate the requirement in 40 C.F.R. § 80.2(p)(3) that a refinery, to qualify as “small,” must not have been owned or controlled during any period since July 1, 1981 by a large refiner. We leave in force the remainder of § 80.-2(p)(3), which requires that a small refinery not be currently owned or controlled by a large refiner.
D. Our Power to Issue a Stay
We stayed EPA’s enforcement of the past ownership requirement over EPA’s objection that we had no power to do so. Therefore, a brief explanation of why we think we had this power is appropriate.
EPA relies on Clean Air Act § 211(f)(5), 42 U.S.C. § 7545(f)(5), which states (emphasis added): “No action of the Administrator under this section may be stayed by any court pending judicial review of such action.” Our review of the legislative history, however, convinces us that Congress meant to preclude judicial stays only under subsection 211(f) (governing “new fuels and fuel additives”), and did not prohibit a judicial stay of regulations issued under other parts of § 211.
Subsection 211(f), including the no-stay provision, was not part of the 1970 version of the Clean Air Act, but was added by § 222 of the 1977 amendments.120 The subsection derives entirely from the Senate bill; the House had no comparable provision. In brief, the Senate wanted to prohibit all new fuel additives, but also permitted EPA to waive the prohibition if certain conditions were met. Significantly, the Senate bill did not prohibit judicial stays.121
The no-stay provision in subsection (f)(5) was added in conference. The Conference Report gives no hint that this provision applies to anything except the particular amendment — concerning new fuel additives — that Congress was considering. The Committee explained:
The House concurs in the Senate amendment [§ 211(f)] with the following amendments: .... (5) no action of the Administrator shall be stayed by any court pending judicial review of such action.
Conf.Rep. at 161, 3 Leg.Hist. 541,1977 U.S. Code Cong. & Ad.News at 1541 — 42. Significantly, this explanation omits the key words “under this section” contained in the actual amendment.
The House, in post-Conference debate, made no reference to § 211(f) at all. In the Senate, Senator Muskie explained § 211(f), but did not refer to the no-stay provision in § 211(f)(5). Rather, his remarks were narrowly focused on new fuel additives:
The conference adopted an amendment to deal with the problem of the fuel additive MMT which will effectively deal with this situation .... [EPA] has been given broad authority to prohibit the use of an additive expeditiously between the period 6 months after enactment until September 15, 1978.
123 Cong.Rec. 26,849 (1977), 3 Leg.Hist. 362.122
*247In short, the legislative history gives no hint that Congress meant to prohibit judicial stays under all of § 211. Had Congress so intended, we think some explanation would have been forthcoming. We also do not know why Congress prohibited stays under § 211(f), but a plausible reason, suggested by Senator Muskie’s comments, would be to ensure that the Administrator could move quickly to ban new fuel additives without the risk that manufacturers of such additives could obtain a judicial stay of the ban.
In light of this history, we hold that § 211(f)(5) prohibits only judicial stays of agency action under § 211(f). The broader phrasing of § 211(f)(5) — “under this section” — we take to be a product of rushed drafting in conference.123
VII. Conclusion
EPA’s final 1.10 gplg gasoline lead standard is affirmed, as is the past production requirement in the definition of “small refinery.” EPA’s interim 1.90 gplg standard is vacated for lack of notice and lack of record evidence that a market for lead credits will develop virtually overnight. We also vacate the past ownership requirement in the definition of “small refinery” for lack of notice and lack of an explanation for the July 1, 1981 cutoff date.
APPENDIX I
ORDER
Nos. 82-2282, 82-2283, 82-2308, 82-2395, 82-2521
Order issued January 26, 1983.
Before WILKEY, WALD and MIKVA, Circuit Judges.Order PER CURIAM.
PER CURIAM:Petitioners Small Refiner Lead Phase-Down Task Force (SRTF), Plateau, Inc., and Simmons Oil Co. challenge various aspects of an EPA regulation that sets lead-content limits for leaded gasoline produced by certain “small” refiners. 47 Fed.Reg. 49,322 (Oct. 29, 1982) (to be codified at 40 C.F.R. §§ 80.2, .4, .7, .20). In brief, the new rule narrows EPA’s previous definition of “small refinery” and requires small refiners to meet an interim standard of no more than 1.90 grams of lead per gallon of leaded gasoline (gplg) as of November 1,1982, and a final standard of no more than 1.10 gplg as of July 1, 1983. (The rule also requires large refiners to meet the more stringent 1.10 gplg standard beginning November 1, 1982, but no large refiners have challenged this standard.) Because our disposition of the case makes speed of the essence, we are issuing an order today, with a full opinion explaining our rationale to follow shortly.
With two exceptions, we uphold the regulation in its entirety as within EPA’s statutory authority, not arbitrary, capricious, or an abuse of discretion, and not procedurally flawed. First, we vacate the interim 1.90 gplg standard because EPA did not give adequate notice that it might immediately require small refiners to significantly reduce lead use. Clean Air Act § 307(d)(3), 42 U.S.C. § 7607(d)(3); see United Steel*248workers v. Marshall, 647 F.2d 1189, 1221 (D.C.Cir.1980) (final rule must be a “ ‘logical outgrowth’ of the rulemaking proceeding”) (quoting South Terminal Corp. v. EPA, 504 F.2d 646, 659 (1st Cir.1974)), cert. denied, 453 U.S. 913, 101 S.Ct. 3148, 69 L.Ed.2d 997 (1981); Weyerhaeuser Co. v. Costle, 590 F.2d 1011, 1028-31 (D.C.Cir.1978). To the contrary, throughout the eight-month rulemaking period, EPA had assured small refiners that its final rule “will take into account the lead time required for construction of any processing equipment needed for compliance.” Notice of Proposed Rulemaking, 47 Fed.Reg. 7814 (Feb. 22, 1982); see also Final Notice of Rulemaking, 47 Fed.Reg. 38,090, 38,091 (Aug. 27, 1982) (suspending effective date of small refiner standard for 30 days “to defer the necessity for small refineries to decide whether to make the substantial capital expenditures required to meet a more stringent lead standard during the time that the Agency is still considering amendments to the lead phasedown program”).
In addition, EPA’s conclusion that the interim standard is feasible relies heavily on a new and untested scheme for inter-refinery trading of lead credits. See 47 Fed. Reg. at 49,324, col. 2. EPA assumed both that a market for lead credits would develop and that it would develop swiftly enough to permit small refiners to meet the interim standard by purchasing lead credits. EPA’s belief that a market for lead credits will eventually develop is reasonable and is supported by the record. However, given the paucity of evidence in the record to support EPA’s belief that a lead credit market will develop fast enough and extensively enough to materially assist the small refiners in meeting the 1.90 gplg interim standard, EPA was not warranted in factoring such a scheme into its decision to impose that standard.
Second, we defer decision on Simmons’ challenge to the historic ownership requirement in 40 C.F.R. § 80.2(p)(3) pending receipt of briefs on the issue by Simmons and EPA. Our stay of the historic ownership requirement contained in that section remains in effect. We reject EPA’s argument that § 211(f)(5) of the Clean Air Act, 42 U.S.C. § 7545(f)(5), prohibits a stay. Our review of the legislative history of that section convinces us that it applies only to actions of the EPA Administrator under § 211(f), and does not prohibit a judicial stay of regulations promulgated under other parts of § 211.
In light of the foregoing, we vacate so much of 40 C.F.R. § 80.20(b)(l)(i) as requires small refineries to limit the lead content of leaded gasoline to 1.90 grams per gallon for gasoline production not exceeding the refinery’s historic production level. We leave in force the requirement in § 80.20(b)(l)(i) that production of leaded gasoline “in excess of the historic production level may not exceed 1.10 grams per gallon.”
When an agency replaces an existing regulation with a new regulation, and we vacate all or part of the new regulation, we must decide whether the agency’s prior regulation continues in effect or whether our action leaves no regulation in effect. Compare Natural Resources Defense Council, Inc. v. Gorsuch, 685 F.2d 718, 728 (D.C.Cir.1982) (implicitly assuming that EPA will return to its previous regulation defining “source” under the Clean Air Act) with Burlington Northern, Inc. v. United States, — U.S. —, 103 S.Ct. 514, 74 L.Ed.2d 311 (1982) (Court of Appeals lacks power to reinstate old ICC rate after invalidating revised rate). In this case, returning to the old standard would be irrational, because the 0.5 grams per gallon (gpg) standard for small refiners (which would have become effective on November 1, 1982) is stricter than both the invalidated 1.90 gplg interim standard and the 1.10 gplg final standard. We therefore vacate only the interim standard and not EPA’s conclusion that the old standard should be replaced. See Burlington Northern, 103 S.Ct. at 522 n. 9 (Court of Appeals should have vacated “only the Commission’s new rate calculation and not *249the Commission’s conclusion that the [previous] rate was too low”).
Our decision leaves EPA without a regulatory limit on small refinery lead use for production of leaded gasoline not exceeding the refinery’s historic production level. The unexpected nature of this regulatory vacuum, plus the public health danger posed by unrestricted lead use and the absence of any unfairness to small refiners in such a course of action, would justify EPA in immediately promulgating a temporary lead-content regulation which does not require small refiners to reduce lead use to a level significantly below previous lead-use levels. See 5 U.S.C. §§ 553(b)(B) (agency may for “good cause” issue rules without notice or public procedure) (made applicable to Clean Air Act by 42 U.S.C. § 7607(d)(1) (last sentence)), 553(d)(3) (“good cause” exception to 30-day period between publication of a rule and the rule’s effective date); American Federation of Government Employees v. Block, 655 F.2d 1153, 1156 (D.C.Cir.1981) (these exceptions to the § 553 procedures permit an agency to issue temporary rules in “emergency situations”). It is for EPA, not for the courts, to decide if an interim standard is needed and what that standard should be. Cf. Burlington Northern, supra (Court of Appeals lacks power to set railroad rates). We will delay issuing the mandate in this case until Wednesday, February 2, 1983 to give EPA an opportunity to promulgate such an emergency rule.1
APPENDIX II
ORDER
No. 82-2521
Order issued February 9, 1983.
Before WILKEY, WALD and MIKVA, Circuit Judges.Order PER CURIAM.
PER CURIAM:Petitioner Simmons Oil Co. challenges one aspect of an EPA regulation that sets lead-content limits for gasoline produced by certain “small” refiners. 47 Fed.Reg. 49,-322 (Oct. 29, 1982) (to be codified at 40 C.F.R. §§ 80.2, .4, .7, .20). In relevant part, the new rule defines “small refinery” to include only refiners that are not and were not “during any period of ownership or control since July 1, 1981” owned or controlled by a refiner with total gasoline production greater than 70,000 barrels per day (bpd). Id. at 49,332 (to be codified at 40 C.F.R. § 80.2(p)(3)). Simmons, a small refiner that was owned by a large refiner until January 4, 1982, challenges this past ownership requirement as having been promulgated without adequate notice and as arbitrary and capricious.
On January 12, 1983, we stayed the past ownership requirement pending decision on the merits. Our disposition of the case makes speed of the essence; the practical effect of our decision expires on July 1, 1983, after which small refiners must meet the same lead-content standard as large refiners. We are therefore issuing an order today, with a full opinion explaining our rationale to follow shortly. We now lift the stay but vacate the past ownership requirement because EPA did not give adequate notice that it might impose such a requirement.
Previous regulations had used capacity limits to define “small refinery” and had a current ownership requirement — a refinery could qualify as small only if it (1) had crude oil capacity of 50,000 bpd or less and (2) was not owned or controlled by a refiner with crude oil capacity greater than 137,500 bpd [hereinafter called a “large refiner”]. 40 C.F.R. § 80.2(p) (1982). In August, 1982, EPA issued a proposed rule in which it switched from capacity limits to production limits. See 47 Fed.Reg. 38,078, 38,084 (Aug. 27, 1982) (explaining the proposed change). *250Under the proposed rule, a refinery, to qualify as “small,” had to (1) have average gasoline production of 10,000 bpd or less during the most recent calendar quarter and (2) not be owned or controlled by a refiner with average gasoline production greater than 70,000 bpd during the most recent calendar quarter. Id. at 38,088 (proposed to be codified at 40 C.F.R. § 80.-2(p)(2) — (3)). The proposal did not alter the current ownership requirement except to the extent that it redefined “large refiner.”
EPA received some 400 written comments on various aspects of the proposed lead-content rule. One commenter, Asamera Oil, suggested that under the current ownership requirement, there was an incentive for large refiners to sell small facilities to take advantage of the less stringent lead-content limits applicable to small refiners. Asamera Oil (U.S.), Inc. Comments, Sept. 22, 1982, at 11-13, J.A. at 381, 392-94.1 EPA was persuaded by Asamera’s comment and therefore added the past ownership requirement to which Simmons now objects. See 47 Fed.Reg. at 49,325-26. The change affects Simmons and perhaps one other refiner.
EPA presents two arguments why Simmons was on notice of this change. First, EPA points out that its Notice of Proposed Rulemaking solicited comments on methods to assure that the change from capacity limits to production limits would not permit small refiners “to expand their production of leaded gasoline.” Id. at 38,085. This request for comments, EPA suggests, should have put Simmons on notice that the ownership criterion might be -revised. In our view, however, the change from a current ownership requirement to a past ownership requirement was not a “logical outgrowth” of EPA’s invitation for comments on loopholes created by the change from capacity limits to production limits. See United Steelworkers v. Marshall, 647 F.2d 1189, 1221 (D.C.Cir.1980) (final rule must be a “ ‘logical outgrowth’ of the rulemaking proceeding”) (quoting South Terminal Corp. v. EPA, 504 F.2d 646, 659 (1st Cir.1974)), cert. denied, 453 U.S. 913, 101 S.Ct. 3148, 69 L.Ed.2d 997 (1981); Weyerhaeuser Co. v. Costle, 590 F.2d 1011, 1028-31 (D.C.Cir.1978).
EPA also argues that Asamera’s comment was sufficient to put Simmons on notice that it. ought to submit comments opposing the past ownership requirement. We disagree. Under Clean Air Act § 307(d), no less than under the Administrative Procedure Act, the agency itself must provide fair notice of what it plans to do. Having failed to do this, EPA cannot bootstrap notice from a comment. See S.Rep. No. 752, 79th Cong., 1st Sess. 14 (1946) (“Agency notice must be sufficient to fairly apprise interested parties of the issues involved, so that they may present responsive data or arguments relating thereto.”) (emphasis added); cf. Wagner Electric Corp. v. Volpe, 466 F.2d 1013, 1019 (3d Cir.1972) (that some “knowledgeable manufacturers” responded to an inadequate notice with comments relating to the final rule “is not relevant. Others [were] possibly not so knowledgeable .... ”).
Had it received comments from Simmons on this issue, EPA might well have concluded that the change from a current ownership requirement to a past ownership requirement was not needed in light of the temporary nature of the small refiner exemption. Alternatively, Simmons might have persuaded EPA that the July 1, 1981 cutoff date (which EPA did not explain) should have been changed — perhaps to February 22, 1982, the date when EPA initially proposed relaxing lead-content limits for small refiners. Thus, there is a “substantial likelihood that the rule would have been significantly changed” had EPA given proper notice. Clean Air Act § 307(d)(8), 42 *251U.S.C. § 7607(d)(8) (standard for invalidating a rule on procedural grounds).
In light of the foregoing, we vacate the requirement in 40 C.F.R. § 80.2(p)(3) that a refinery, to qualify as “small,” must not have been owned or controlled during any period since July 1, 1981 by a refiner with total gasoline production greater than 70,-000 bpd. We leave in force that portion of § 80.2(p)(3) which requires that a small refinery not be currently owned or controlled by a large refiner. EPA remains free, of course, to institute expedited rulemaking with a view toward possibly repromulgating a new regulation to replace the vacated portion of § 80.2(p)(3).
The mandate will issue immediately.
. Notes 46-65, 78-79 infra and accompanying texts discuss the health hazards posed by gasoline lead.
. 38 Fed.Reg. 33,734, 33,741 (1973) (codified at 40 C.F.R. § 80.20(a)(1) (1976)).
. For example, if a refiner produces 50% unleaded gasoline, it can use as much as 1.0 grams per leaded gallon (gplg) and still comply with a 0.5 gpg standard.
. 38 Fed.Reg. at 33,740/2; see 40 C.F.R. § 80.-20(b) (1976). (In citing to the Federal Register, we use the notation “33,740/2” as shorthand for page 33,740, col. 2.)
. 44 Fed.Reg. 53,144 (1979) (codified at 40 C.F.R. § 80.20(a)(6) (1980)).
. Clean Air Act Amendments of 1977, Pub.L. No. 95-95, § 223, 91 Stat. 685, 764 (codified at Clean Air Act § 211(g), 42 U.S.C. § 7545(g)).
. 44 Fed.Reg. 46,275 (1979) (codified at 40 C.F.R. § 80.20(a)(5), (b) (1982)).
. 43 Fed.Reg. 46,246 (1978) (codified at 40 C.F.R. § 50.12 (1982)). Clean Air Act § 109, 42 U.S.C. § 7409, directs EPA to issue “primary” and “secondary” ambient air quality standards for “criteria” pollutants identified under § 108(a), 42 U.S.C. § 7408(a). For lead, EPA established 1.5 ug/m 3 as both the primary and the secondary ambient air quality standard.
. Six months earlier, Vice President Bush had listed the lead content regulation as one of 30 regulations that would be subject to “in-depth agency reconsideration” and possible regulatory relief. Remarks of Vice President George Bush at the Presidential Task Force on Regulatory Relief Briefing 1 (Aug. 12, 1981), reprinted in Joint Appendix (“J.A.”) at 1555, 1555.
. See Notice of Final Rulemaking, 47 Fed.Reg. 49,322, 49,329 (Oct. 29, 1982).
. Our earlier order had deferred decision on this issue pending receipt of briefs from the parties. Simmons had filed its petition for review much later than SRTF and Plateau, and was therefore on a later briefing schedule.
. When Ethyl Corp. was decided, § 211 required EPA to find that a fuel additive “will endanger” the public health or welfare. We held that this language did not require EPA to prove actual harm to health or welfare — a “significant risk” of harm was enough. 541 F.2d at 12. Congress subsequently reworded this standard — to “may reasonably be anticipated to endanger” — with the specific purpose of endorsing the Ethyl decision. See H.R.Rep. No. 294, 95th Cong., 1st Sess. 43-51 (1977) [hereinafter cited as H.R.Rep.], reprinted in 4 Environmental Policy Division, Congressional Research Service, A Legislative History of the Clean Air Act Amendments of 1977, at 2468, 2510-18 (Comm.Print 1978) [hereinafter cited as Leg. Hist. ], and in 1977 U.S.Code Cong. & Ad.News 1077, 1121-29.
. SRTF Brief at 53-59; SRTF Reply Brief at 41 — 18.
. See also 42 U.S.C. § 7521(f)(3) (before issuing high-altitude motor vehicle regulations, “the Administrator shall consider and make a finding with respect to [economic impact, technological feasibility, and expected air quality improvement]”) (emphasis added); cf. id. § 7408(a)(2) (“Air quality criteria for an air pollutant shall accurately reflect the latest scientific knowledge....”) (emphasis added).
. Both in the original passage of the Clean Air Act in 1970 and the 1977 amendments, Congress rejected a proposal to require EPA to make “findings” before regulating fuel additives under § 211 generally. See Ethyl Corp., 541 F.2d at 21-24 (reviewing legislative history of 1970 Act); H.R.Rep., supra note 12, at 51, 4 Leg.Hist. 2518, 1977 U.S.Code Cong. & Ad. News at 1129 (in considering the 1977-amend-ments, “the committee expressly rejected an amendment which would have ... required a finding by the Administrator”). Thus, it is most unlikely that Congress’ failure to require findings under § 211(g)(2) (dealing specifically with small refiners) was inadvertent.
. See also, e.g., Federal Election Commission, 454 U.S. at 39, 102 S.Ct. at 46 (agency interpretation must be accepted if “sufficiently reasonable” even if it is not “the only reasonable one or even the reading the court would have reached” on its own); EPA v. National Crushed Stone Ass’n, 449 U.S. 64, 83, 101 S.Ct. 295, 306, 66 L.Ed.2d 268 (1980); National Wildlife Fed’n v. Gorsuch, 693 F.2d 156, 166-71 (D.C.Cir.1982).
. Cf. Weyerhaeuser Co. v. Costle, 590 F.2d 1011, 1046 (D.C.Cir.1978) (Clean Water Act § 304(b)(1)(B), 33 U.S.C. § 1314(b)(1)(B), directs EPA to “consider” certain factors; this requires the agency only to “ ‘take into account’ ” those factors “without prescribing any structure for EPA’s deliberations”); BASF Wyandotte Corp. v. Costle, 598 F.2d 637, 656 & n. 36 (1st Cir.1979) (construing same section of Clean Water Act in similar fashion), cert. denied, 444 U.S. 1096, 100 S.Ct. 1063, 62 L.Ed.2d 784 (1980).
.We address in part IV infra the separate question whether EPA’s decision to adopt a uniform 1.10 gplg standard was reasonable.
. Section 109(b), 42 U.S.C. § 7409(b), directs EPA to issue both “primary” and “secondary” ambient air quality standards. The primary standard, contained in § 109(b)(1), is quoted in text. The secondary standard, prescribed in § 109(b)(2), is the level “requisite to protect the public welfare from any known or anticipated adverse effects associated with the ... pollutant.”
. The substantive standards are: § 111(a), 42 U.S.C. § 7411(a) (“degree of emission limitation ... achievable through application of the best technological system of continuous emission reduction which ... has been adequately demonstrated”); § 202(a)(3)(A)(i), 42 U.S.C. § 7521(a)(3)(A)(i) (“greatest degree of emission reduction achievable through the application of [available] technology”).
. See 42 U.S.C. §§ 1857c-3(a)(l)(A), 1857f-6c(c)(l)(A) (1976) (the previous codification of §§ 108(a)(1)(A) and 211(c)(1)(A)).
. See also H.R.Rep. No. 564, 95th Cong., 1st Sess. 183 (1977), 3 Leg.Hist. 381, 563, 1977 U.S.Code Cong. & Ad.News 1502, 1564 [hereinafter cited as Conf.Rep.] (1977 amendments establish a “uniform standard of proof’ for when EPA “could” regulate air pollutants).
. See id.; H.R.Rep., supra note 12, at 49-50, 4 Leg.Hist. 2516-17, 1977 U.S.Code Cong. & Ad. News at 1127-28.
. In this case, EPA issued nationally uniform lead-content standards, and petitioners do not challenge the agency’s decision to do so. Thus, we have no occasion to decide whether or not § 211 requires EPA to issue uniform standards.
. In adopting the “may reasonably be anticipated to endanger” standard, Congress specifically intended to “assure consideration of the cumulative impact of all sources of a pollutant,” including food and water sources. H.R. Rep., supra note 12, at 49, 4 Leg.Hist. 2516, 1977 U.S.Code Cong. & Ad.News at 1127.
.In light of this conclusion, we need not address the other premise on which SRTF’s argument rests — that small refineries do not contribute to violations of the ambient air standard for lead. We note, however, that this premise is shaky as well, since at least some portion of small refiner output is consumed in urban areas. See notes 66-77 infra and accompanying text.
. We express no view on whether the “proposed rule” must take any particular form, so long as it is specific enough to comply with § 307(d).
. Clean Air Act § 307(d)(2), (3), (4)(B)(i), (6), 42 U.S.C. § 7607(d)(2), (3), (4)(B)(i), (6).
. Id. § 307(d)(6)(C), (7)(A), 42 U.S.C. § 7607(d)(6)(C), (7)(A). For a more thorough review of the rulemaking record requirements of § 307(d), including a discussion of the legislative history, see Sierra Club v. Costie, 657 F.2d 298, 392-400 (D.C.Cir.1981).
There is one exception to the rule that the material in the docket becomes the record for judicial review. The docket must include drafts of proposed rules submitted by EPA to the Office of Management and Budget (OMB) for interagency review, any comments on the drafts by OMB or other agencies, and any written EPA responses. Id. § 307(d)(4)(B)(ii), 42 U.S.C. § 7607(d)(4)(B)(ii). This material is not, however, part of the record for judicial review. See id. § 307(d)(7)(A), 42 U.S.C. § 7607(d)(7)(A).
. See also Conf.Rep., supra note 22, at 178, 3 Leg.Hist. 558, 1977 U.S.Code Cong. & Ad.News at 1559 (“to avoid misinterpretations or confusion regarding the intent underlying a change in the existing scope of review, it was the decision of the conferees to retain the ‘arbitrary and capricious’ scope of review”).
. See also 123 Cong.Rec. 27,075 (1977), 3 Leg. Hist. 333 (statement of Rep. Broyhill explaining Conference Committee version of § 307(d)) (similar to Conference Report).
. One possible source of a practical distinction between the two tests under the APA is absent under § 307(d). A “basic requirement” for substantial evidence review under the APA is “a record that is to be the basis of agency action.” Citizens to Preserve Overton Park v. Volpe, 401 U.S. 402, 415, 91 S.Ct. 814, 822, 28 L.Ed.2d 136 (1971); see Ethyl Corp., 541 F.2d at 37 n. 79 (the “primary difference” is that “ ‘substantial evidence’ review is limited to evidence developed in formal hearings”). Section 307(d), in contrast, combines “arbitrary and capricious” review with a well-defined record on which EPA must base its final rule.
.See Leventhal, Environmental Decisionmaking and the Roie of the Courts, 122 U.Pa.L.Rev. 509, 514 (1974) (“The court does not make the ultimate decision, but it insists that the agency take a ‘hard look’ at ail relevant factors.”). For a review of the development of the phrase “hard look review,” see National Lime Ass’n v. EPA, 627 F.2d 416, 451 n. 126 (D.C.Cir.1980). See id. at 452-53 & nn. 128-38 for a survey of the strictures that such review imposes on agencies.
In enacting § 307(d), Congress specifically approved this “hard look” review. See Conf. Rep., supra note 22, at 178, 3 Leg.Hist. 558, 1977 U.S.Code Cong. & Ad.News at 1559 (“[T]he conferees intend that the courts continue their thorough, comprehensive review which has characterized judicial proceedings under the Clean Air Act thus far.”); H.R.Rep., supra note 12, at 323,4 Leg.Hist. 2790, 1977 U.S.Code Cong. & Ad.News at 1402 (endorsing the judicial practice of “engaging in searching review without substituting their judgment for that of the Administrator”). '
. In addition, all objections to an EPA rule, whether substantive or procedural, must be “raised with reasonable specificity during the period for public comment ... [unless] it was impracticable to raise such objection within such time.” Clean Air Act § 307(d)(7)(B), 42 U.S.C. § 7607(d)(7)(B); see Lead Industries Association, 647 F.2d at 1173 (§ 307(d)(7)(B) “applies to all objections”) (emphasis in original).
. The Conference Committee’s consideration of § 307(d) was apparently rushed. Its report includes two paragraphs explaining a change in § 307(d)(3)(C) (from “policy considerations” to “policy options”) that was never made. See Conf.Rep., supra note 22, at 178, 3 Leg.Hist. 558, 1977 U.S.Code Cong. & Ad.News at 1558-59. The 84 technical amendments to the Clean Air Act contained in § 14(a) of the Safe Drinking Water Amendments of 1977, Pub.L. No. 95-190, 91 Stat. 1393, 1399-1405, are further evidence of the Committee’s haste.
. See also American Petroleum Inst. v. Costle, 665 F.2d 1176, 1187 (D.C.Cir.1981) (quoting this statement from Sierra Club), cert. denied, 455 U.S. 1034, 102 S.Ct. 1737, 72 L.Ed.2d 152 (1982).
. EPA discussed the health effects of lead exposure at length in its August notice of proposed rulemaking. See 47 Fed.Reg. at 38,071-77. The notice of final rulemaking reaffirms the tentative findings in the notice of proposed rulemaking but does not repeat the analysis of health effects, id at 49,330/1. Our discussion therefore relies heavily on the notice of proposed rulemaking.
. Environmental Protection Agency, Supplemental Responses to Comments on the February 22, 1982, and August 27, 1982, Proposals to Amend the Gasoline Lead Content Regulations III — 1 to -2 (Oct. 29, 1982), J.A. at 1319, 1329-30 [hereinafter cited as Supplemental Response to Comments].
. EPA estimates that total industry use of lead in 1990 would be 16.2 billion grams under a uniform 1.10 gplg standard. 47 Fed.Reg. at 49,329. Current leaded gasoline production by small refiners is 2.714 billion gallons per year, id. at 49,327/3, and a 2.50 gplg standard implies that small refiners can use an additional 1.40 gplg. Thus, if small refinery production remains constant, the 2.50 gplg standard implies additional annual use of as much as 2.714 billion gallons X 1.40 gplg = 3.8 billion grams, and total industry use of 20.0 billion grams. Total small refiner use could be as much as 2.714 billion gallons X 2.50 gplg = 6.8 billion grams of lead per year, or 34% of the total industry use of 20 billion grams.
EPA noted the increasing market share of small refiners as a problem in Supplemental Response to Comments, supra note 38, at III — 4, J.A. at 1332 (under a dual standard, small refiners “would contribute a larger and larger part of total automotive lead emissions”).
.EPA did not state this conclusion explicitly, but it is implicit in, e.g., EPA’s discussion of the need for an 8-month delay before the final standard takes effect, 47 Fed.Reg. at 49,324/2, and its reference to “the scope and profitability of small refinery operations,” id. at 49,324/3. See also Supplemental Response to Comments, supra note 38, at III — 5, J.A. at 1333 (“The Agency has attempted to reach its environmental goal without ... undue cost, but reduction of lead ... must be the driving consideration.”).
. Part IV.D infra discusses the Regulatory Flexibility Act in more detail.
.In Ethyl Corp., 541 F.2d at 54 n. 124, we questioned, without giving reasons and without deciding the matter, whether EPA “is allowed to consider at all the economic effects of [fuel additive] regulations.” But see Union Elec. Co. v. EPA, 427 U.S. 246, 257 n. 5, 96 S.Ct. 2518, 2525 n. 5, 49 L.Ed.2d 474 (1976) (dictum) (§ 211(c)(2)(A) “expressly permit[s]” consideration of economic and technological feasibility). The cumulative weight of several pieces of evidence convinces us that EPA can consider economic factors, notwithstanding the contrary hint in Ethyl Corp.
First, nothing in § 211(c)(2)(A) limits EPA’s discretion to consider all relevant factors, including economic factors, when regulating fuel additives. Quite the contrary, § 211(c)(2)(A) requires EPA to consider “other technologically or economically feasible means of achieving emission standards under [§ 202 (governing automobile emission standards) ].” Thus, EPA must consider economic factors whenever regulation under § 202 can achieve the same health benefits as regulation under § 211.
Second, when EPA regulates a fuel additive for the specific reason that the additive may degrade the performance of an auto emission control device, it must, under § 211(c)(2)(B), conduct “a cost benefit analysis.” While § 211(c)(2)(A), which applies to this case, does not require cost-benefit analysis, it seems unlikely that Congress would mandate cost-benefit analysis for fuel additive regulations under § 211(c)(2)(B) and at the same time forbid EPA from considering cost at all for fuel additive regulations issued under § 211(c)(2)(A).
Third, the legislative history confirms that Congress expected cost to be a factor in EPA’s regulation of fuel additives under § 211(c)(2)(A). See H.R.Rep. No. 1783 (Conf. Rep.), 9Íst Cong., 2d Sess. 53 (1970), reprinted in 1 Congressional Research Service, Environmental Policy Division, A Legislative History of the Clean Air Amendments of 1970, at 151, 203 (Comm.Print 1974) (“1970 Leg.Hist.”), and in 1970 U.S.Code Cong. & Ad.News 5374, 5385 (“Before controlling or prohibiting manufacture or sale [of a fuel additive], the Administrator is required to consider specific technical and cost factors.”); S.Rep. No. 1196, 91st Cong., 2d Sess. 35, 1970 Leg.Hist. 397, 435 (Congress gave EPA power to “control” as well as to “prohibit” fuel additives in order to “achieve maximum control of auto emissions at a minimum cost”).
Fourth, Congress took economic feasibility into account in granting small refiners a looser sliding-scale standard from 1977 to 1982. This further suggests that it considered economic factors to be relevant under § 211 generally.
. SRTF Brief at 25-36, 39-47; SRTF Reply Brief at 9, 24 — 31, 37-41; see also Plateau Brief at 27-30; Plateau Reply Brief at 18-23.
. State Farm Mut. Auto. Ins. Co. v. Department of Transp., 680 F.2d 206, 221 (D.C.Cir.), cert. granted, — U.S. —, 103 S.Ct. 340, 74 L.Ed.2d 382 (1982).
. Greater Boston Television Corp. v. FCC, 444 F.2d 841, 852 (D.C.Cir.1970) (footnote omitted), cert. denied, 403 U.S. 923, 91 S.Ct. 2233, 29 L.Ed.2d 701 (1971); see also Hatch v. FERC, 654 F.2d 825, 833-35 (D.C.Cir.1981); Natural Resources Defense Council, Inc. v. SEC, 606 F.2d 1031, 1049 n. 23 (D.C.Cir.1979); Pacific Legal Found v. Department of Transp., 593 F.2d 1338, 1343-44 (D.C.Cir.), cert. denied, 444 U.S. 830, 100 S.Ct. 57, 62 L.Ed.2d 38 (1979).
. Figure 1 is taken from Statement by Dr. Vernon Houk, Acting Director, Center for Environmental Health, Centers for Disease Control, on Regulations on Fuel and Fuel Additives 10 (Apr. 15, 1982), reprinted in Natural Resources Defense Council Brief, Supplemental Appendix (“NRDC App.”) item 4. The acronymNHANES II stands for Second National Health and Nutrition Examination Survey.
This and the other studies cited in the opinion were criticized by various commenters. EPA responded to the criticisms in a number of ways: explaining why some critiques were unfounded; commissioning new studies to examine others; accepting still others as partially valid. See generally 47 Fed.Reg. at 38,073-77; Supplemental Response to Comments, supra note 38, at VIII — 1 to -21, IX-1 to -9, J.A. at 1371-1400. We have reviewed the criticisms and EPA’s responses and find that EPA has amply discharged its duty to respond “to each of the significant comments, criticisms, and new data submitted ... during the comment period.” Clean Air Act § 307(d)(6)(B), 42 U.S.C. § 7607(d)(6)(B).
. Statement by Dr. Vernon Houk, supra note 46, at 8.
. See Memorandum by Dr. Joel Schwartz, Office of Policy and Resource Management, Energy Policy Division, EPA, on Health Effects of Gasoline Lead Emissions 9-10 (May 11, 1982), J.A. at 274, 286-91; I. Billick, Prediction of Pediatric Blood Levels from Gasoline Consumption (unpublished HUD report, Apr. 21, 1982).
. See ICF, Inc., The Relationship Between Gasoline Lead Emissions and Blood Poisoning in Americans 11-12 (unpublished EPA report, Oct. 1982), J.A. at 1255, 1269-70 (food lead); D. Hsia, Blood Lead of Chicago Children and Its Sources, 1967-1980, at 9 (unpublished EPA report, July 1982) (concluding that gasoline is responsible for 55% of blood lead levels in Chicago children, lead paint for only 25%). Also, adults consume little lead paint, and infants less than a year old consume little canned food (an important source of food lead), yet both adults and small infants show drops in mean blood lead levels similar to other age groups. ICF, Inc., supra, at 12-13, J.A. at 127ÍL-71; Memorandum by Dr. Joel Schwartz, supra note 48, at 8-9, J.A. at 285-86.
. EPA surveyed the major studies in its statement of basis and purpose for the lead ambient air quality standard. See 43 Fed.Reg. 46,246, 46,254'(1978).
. Memorandum by Dr. Joel Schwartz, supra note 48, at 1, J.A. at 274.
. Id at 7, J.A. at 281.
. Memorandum by Dr. James Pirkle, Center for Environmental Health, Centers for Disease Control, on Percentage Distributions of NHANES II Blood Lead Data By Age, at 4 (May 12, 1982), J.A. at 305, 308.
. The percentage of children screened by CDC with lead poisoning dropped from 4.3% in 1980 to 3.55% in 1981. Memorandum by Dr. Joel Schwartz, supra note 48, at 5, J.A. at 280. These percentages are higher than the NHANES II percentages discussed in text because the CDC screening program concentrates on high-risk groups.
. Annest, O’Connell, Roberts & Murphy, Blood Lead Levels from the Second National Health and Nutrition Examination Survey, 1976-1980, in American Public Health Ass’n, Silver Anniversary of the National Health Survey Act: Contributed Papers Session 31, 37 (Hyattsville, Md., Oct. 1981), reprinted in NRDC App., supra note 46, item 5.
. Needleman, Gunnoe, Levitón, Reed, Peresie, Maher & Barrett, Deñcits in Psychologic and Classroom Performance of Children with Elevated Dentine Lead Levels, 300 New Eng.J. Med. 689 (1979). While this study did not correlate dentine lead levels with blood lead levels, the tested children presumably had blood lead levels in line with national norms; thus, most had blood lead levels well below 30 ug/dl.
The National Academy of Sciences concluded that the Needleman study was “largely successful in overcoming the methodological deficiencies of earlier studies” and was a “major contribution” that provided “significant evidence that some ‘average’ children may be at risk.” National Academy of Sciences, National Research Council, Committee on Lead in the Human Environment, Lead in the Human Environment 73 (1980).
. Yule, Lansdown, Millar & Urbanowicz, The Relationship Between Blood Lead Concentrations, Intelligence and Attainment in a School Population: A Pilot Study, 23 Dev.Med.Child. Neurology 567, 569 (1981).
. Otto, Benignus, Muller & Barton, Effects of Age and Body Lead Burden on CNS Function in Young Children I: Slow Cortical Potentials, 52 Electroencephalography & Clinical Neurophysiology 229 (1981). See generally Low Level Lead Exposure: The Clinical Implications of Current Research (H. Needleman ed. 1980).
. See Memorandum by Dr. James Pirkle, supra note 53, at 2, J.A. at 306 (in NHANES II study, 2.1% of children tested during 1978-1980 had blood lead levels of 30 ug/dl or higher; 3.7% had blood lead levels of 25 ug/dl or higher). Ten ug/dl was the mean blood lead level for all persons in 1980, according to NHANES II data. See Figure 1 supra. Mean blood lead levels for small children were somewhat higher. Annest, Mahaffey, Cox & Roberts, Blood Lead Levels for Persons 6 Months-74 Years of Age: United States, 1976-80, in National Center for Health Statistics, Advance Data from Vital and Health Statistics 4-5 (No. 79, May 12, 1982).
. See 47 Fed.Reg. at 38,074/1 (60 of 64 comments opposed loosening); Memorandum from Richard Kozlowski, Director, EPA Field Operations and Support Division, to Kathleen Bennett, Assistant EPA Administrator for Air, Noise and Radiation, at 2 (June 3, 1982), reprinted in NRDC App., supra note 46, item 3 (four doctors sponsored by the Lead Industries Association favored rescission). (Mr. Kozlowski’s memorandum refers to a fifth doctor who supported rescission; we assume that this was an error that was corrected in the later Federal Register notice.)
. Letter from Dr. Glenn Austin, President, American Academy of Pediatrics (Apr. 9,1982), J.A. at 77; see Statement of the American Medical Association 2 (July 6, 1982), J.A. at 254, 255 (“further decreases in blood lead levels are necessary”).
. See 47 Fed.Reg. at 38,074/1; Memorandum from Richard Kozlowski to Kathleen Bennett, supra note 60, at 2.
. National Academy of Sciences, supra note 56, at 236-37.
. Id. at 254.
. ICF, Inc., supra note 49, at 38-39, J.A. at 1296-97.
. See text accompanying notes 38-39 supra.
. See SRTF Comments, Oct. 8, 1982, at 21, J.A. at 738, 759.
. EPA Brief at 34-36; see NRDC Brief at 5, 21.
. See, e.g., Asamera Oil (U.S.), Inc. Comments, Sept. 22, 1982, at 8, J.A. at 381, 389 (giving one example); Environmental Defense *227Fund Comments, Oct. 8, 1982, at 18, J.A. at 776, 794 (“a large proportion” of small refineries are located near pipelines). The only numerical estimate of sales to pipelines (other than SRTF’s) is by another small refiner association. See American Petroleum Refiners Association Comments, Oct. 4, 1982, at 12, J.A. at 490, 502 (“small” refiners sold 34,000 bpd of leaded gasoline to the “six pipelines serving major metropolitan areas”).
This and similar conclusions about whether there is record evidence on a particular issue must be qualified by the disclaimer “to the best of our knowledge.” We have perused the almost 2000 pages of joint appendix and 350 pages of public hearing transcript with a fair amount of care. We have read closely the almost 550 pages of briefs. Where these items were insufficient, we have asked EPA to supply us with specific additional portions of the record. We do not pretend, however, to have read every page of the record nor could we reasonably be expected to do so. Ultimately, we must depend on the parties to tell us where in that record to look. Evidence that they have not cited and that we have not found on our own must, of practical necessity, not be weighed in the balance.
.The American Petroleum Refiners Association admitted that “[sjmall refiners do sell into a number of [other] pipelines,” but asserted that those pipelines do not serve “major metropolitan areas.” American Petroleum Refiners Association Comments, supra note 69, at 12 n. *, J.A. at 502 n. *. Without attempting to parse what areas the American Petroleum Refiners counted as “major metropolitan areas,” a term it did not define, we feel safe in assuming that most pipelines predominantly serve urban areas, for a pipeline’s profits generally depend on high volume shipments.
. “Processing” refineries refine gasoline from crude oil or other basic feedstocks. “Blenders” mix already refined high-octane components to produce marketable gasoline. SRTF, needless to say, represents only processing refineries.
. Leaded gasoline production by the 64 small processing refineries from July 1, 1981 through March 30, 1982 was approximately 33,738,000,-000 barrels or 124,000 bpd. See Letter from Jimmie Peterson, Director, Office of Oil and Gas, Energy Information Administration, to Martin Wagner, Acting Director, EPA Energy Policy Division, at 2 (Sept. 29, 1982), J.A. at 1301, 1302.
. See 47 Fed.Reg. at 38,085. EPA found in its notice of proposed rulemaking that small refineries had a “substantial” market share in only one urban area. It concluded that there was not a “significant problem” of small refiner lead use causing specific urban areas “to experience significantly higher lead concentrations.” Id. at 38,085/1. However, small refiner sales to urban areas can be a significant percentage of small refiner output even if (as EPA concluded in its notice of proposed rule-making) they are not a large percentage of total gasoline sales by all refiners in any one urban area.
. Sobotka & Co., Inc., Estimated Gasoline Disposition Between Urban and Nonurban Areas for Selected Group of Small Refineries and Blending Plants (Oct. 7, 1982), J.A. at 1215.
. Sobotka & Co., Inc., Disposition of Gasoline Production By Small Refiners — Rural vs. Urban Areas 2 (May 7, 1982), J.A. at 327, 328 (an earlier report using a different definition of small refinery).
. Letter from L.L. Hank Hankla, Attorney for SRTF, to Kathleen Bennett, Assistant EPA Administrator for Air, Noise & Radiation Programs, at 4 (July 30, 1982), J.A. at 215, 218. The 23% figure was based on an earlier EPA definition of small refinery; a revised SRTF figure would likely have been somewhat lower.
. For EPA’s 1990 estimate, see note 39 supra and accompanying text.
. See text accompanying note 55 supra.
. Memorandum from Dr. James Pirkle, Center for Environmental Health, Centers for Disease Control, on Sampling of the Urban Population and the Downward Trend in NHANES II Blood Lead Levels, figs. 14 — 15 (Aug. 24, 1982). On reflection, this result is not hard to understand. First, most people live and work in towns, even in rural areas. Second, lead settles quickly to the ground; thus, gasoline lead exposure depends only on the density of traffic in the immediate vicinity. Therefore, gasoline lead exposure in a town in, say, Wyoming will be comparable to lead exposure in a similar-sized town in the suburbs of a major city.
. See note 40 supra and accompanying text. The analysis in the text of whether the record supports the conclusion that most small refiners can meet the 1.10 gplg standard should not be read to suggest that § 211 precludes EPA from setting a standard that most small refiners cannot meet. Even if EPA must consider small refiners as a discrete class (as we concluded, see text accompanying note 41 supra) and must also consider the economic effects of fuel additive regulations (a question that we avoided, see note 42 supra and accompanying text), nothing in § 211 requires EPA, after duly considering economic impact on small refiners, to grant them special benefits.
. Sobotka & Co., Inc., Potential for Cost Reduction and Energy Savings Realized Through Modiñcation of Lead Phasedown Regulation II — 9 to -10 (Nov. 20, 1981); J.A. at 12, 21-22.
. EPA’s conclusion that most small refiners can meet the 1.10 gplg standard will also be justified if variation among small refiners is small enough so that the model fairly depicts what most small refiners can achieve without interrefinery trading. There is conflicting evidence in the record on how many small refineries have enough octane-enhancing equipment to meet the new standard. Compare Crown Central Petroleum Corp. Comments, Oct. 6, 1982, at 5-6, J.A. at 872, 876-77 (only 15 small refineries lack the “catalytic reformers” needed to produce high octane gasoline) with Sobotka & Co., Inc., Diversity Among Small Refiners ' Eligible for Higher Lead Use Concentrations 1-2 (Oct. 6, 1982), Í.A. at 1231, 1231-32 (17 small refineries have no and 28 have only limited octane-enhancing capability; unless these refiners can purchase blending components or lead credits, they “will be at a disadvantage in the manufacture of gasoline (or will be unable to make gasoline)”). In light of this conflict, which EPA did not try to resolve, we are unable to conclude that most small refineries can meet the 1.10 gplg standard without interrefinery trading.
. Sobotka & Co., Inc., Diversity Among Small Refiners, supra note 82, at 1, J.A. at 1231.
. See Environmental Protection Agency, Final Regulatory Flexibility Analysis for Lead Phase-down Regulations 11, 21-22 (Oct. 1982), J.A. at 1418,1429, 1439-40 (making this argument and noting that the Herfindahl market concentration index for the petroleum refining industry is only 0.04, indicating a highly competitive industry).
. See id. at 25, J.A. at 1443 (“since the marginal value of lead to large refineries is 0.9 cents per gram in 1983, [small refiners] should be able to purchase lead rights ... for about this pjice”).
. See Sobotka & Co., Inc., Potential for Cost Reduction, supra note 81, at VI — 4 to -6, J.A. at 26-28 (describing precedent for lead credit trading).
. See, e.g., United States Department of Justice Comments, Oct. 8, 1982, at 3 — 1, J.A. at 599, 602-03 (“incentives provided by basic market forces” will create a market for lead credits); Texaco U.S.A. Comments, Oct. 8, 1982, at 18, J.A. at 665, 682 (averaging will provide small refiners “ample flexibility” in meeting the large refiner limit); Mid-Continent Energy Co. Comments, Oct. 7, 1982, at 2, J.A. at 578, 579 (averaging “would be a necessity for small companies to compete in the market”).
. See 47 Fed.Reg. at 49,326 (summarizing objections to averaging).
. Both the Department of Justice and the Department of Energy supported a uniform standard on efficiency grounds. See United States Department of Justice Comments, May 17, 1982, at 10-12, J.A. at 113, 123-25; Department of Energy Comments, Oct. 8, 1982, enclosure 3, at 2, J.A. at 798, 811.
. See H.R.Rep. No. 519, 96th Cong., 1st Sess. 11-12 (1979); S.Rep. No. 878, 96th Cong., 2d Sess. 9-10 (1980), reprinted in 1980 U.S.Code Cong. & Ad.News 2788, 2796-97.
. Neither SRTF nor EPA noticed this history. They therefore engage in a pointless debate on the significance of language in the Senate Report suggesting that judicial review is available.
.126 Cong.Rec. S 10,939 (daily ed. Aug. 6, 1980). The italicized words “and consciously” were not present in Senator Culver’s original analysis, but Senator Culver proposed adding them in a letter to House Speaker O’Neill (dated after the Senate passed the bill) “in response to [Speaker O’Neill’s concerns] with regard to the judicial review provisions of [the Senate bill].” See id. at H8466 (daily ed. Sept. 8, 1980) (reprinting the letter).
. See also id. at H8461 (statement of Rep. Bedell) (“regulatory flexibility analyses can be examined by the courts when the validity of final rules is being determined”); id. at H8462 (statement of Rep. Moorhead) (“the regulatory analysis ... will be part of the record on review of the rule itself’); id. at H8471 (statement of Rep. Roth) (“If the Agency fails to do the required analysis or fails to take its own study into account when the final rule is published, the regulation is subject to being struck down by the courts.”).
One speaker believed that the Act provided for full-scale review of the regulatory analysis, see id. at H8463 (statement of Rep. McDade) (“the court should strike down the regulation” if the regulatory analysis is inadequate), and one believed that it barred all review, see id. at H8470 (statement of Rep. Danielson) (“the language of the bill itself” precludes review).
. SRTF’s and Plateau’s remaining objections to EPA’s flexibility analysis merely rehash their objections to the rule in general. For example, SRTF complains that EPA used an aggregate model and failed to consider the supposed rural nature of small refineries. These issues are fully addressed elsewhere in this opinion.
. Plateau also claims that it had no notice of the possibility that EPA might provide only eight-months lead time before imposing a uniform final standard. Plateau Brief at 15-18. However, in its August proposal, EPA specifically solicited comments on “whether the special standard for small refineries should be limited to some definite time period.” 47 Fed.Reg. at 38,085/3. This request was sufficient to put Plateau on notice that it ought to submit comments on how much lead time it needed and why. Plateau chose not to submit comments on this issue.
To be sure, EPA did not proffer a particular time period or periods that it was considering, but Plateau has not suggested that its hypothetical comments would have been altered had it known that EPA was specifically contemplating an eight-month time period. See BASF Wyandotte Corp. v. Costle, 598 F.2d 637, 644 (1st Cir.1979) (petitioners “have not shown us that the content of their criticisms would have been different” had the agency given more detailed notice), cert. denied, 444 U.S. 1096, 100 S.Ct. 1063, 62 L.Ed.2d 784 (1980).
. See, e.g., SRTF Brief at 42 n. 37 (objecting to Sobotka’s calculation of the cost to small refiners of meeting a 1.10 gplg standard); notes 74-76 supra and accompanying text (discussing SRTF’s objection to Sobotka’s calculation of direct small refiner sales to urban areas).
. Department of Energy Comments, supra note 89, enclosure 3, at 2, J.A. at 811 (small refiner costs at 1.4 gplg are equal to large refiner costs at 1.17 gplg).
. The sliding scale standard was 2.65 gpg for refineries producing 5000 bpd or less and 2.15 gpg for refineries producing 5,001 to 10,000 bpd. Clean Air Act § 211(g)(2), 42 U.S.C. § 7545(g)(2). In the first quarter of 1982, small processing refineries produced 72% leaded gasoline. See Letter from Jimmie Peterson, supra note 72, at 2, J.A. at 1302. A refinery that was subject to a 2.65 gpg standard and produced 72% leaded gasoline could use up to 3.7 gplg; a similar refinery subject to a 2.15 gpg standard could use up to 3.0 gplg.
Actual average lead use was somewhat lower — roughly 2.5 gplg for small processing refineries. See notes 103-04 infra and accompanying text.
. EPA Brief at 58.
. The only suggestion in the record that EPA was considering a strict interim standard is a single question asked at the public hearing by Dr. Joel Schwartz of EPA’s Office of Policy Analysis. Dr. Schwartz inquired of a panel of large refiners:
Now you’ve already testified that you think that if there is a separate standard for small refiners it should phase out quickly over time. I wonder if you could ... give us some indication what you think would be an appropriate level that would not unduly burden the refiners over the transition period ... ?
Transcript of Lead Phasedown Hearing, Sept. 7, 1982, at 41. As discussed in part VI.C.3 infra, neither the APA nor the Clean Air Act requires parties to a rulemaking to scrutinize the record for tiny clues as to what the agency might do. This single question does not come close to being adequate notice of a possible interim standard.
.Texaco Brief at 48-49; Environmental Defense Fund Brief at 18-22.
. As we further concluded in part III.D, the additional requirement in § 307(d)(9)(D)(i), 42 U.S.C. § 7607(d)(9)(D)(i), that the procedural error is grounds for reversal only if “arbitrary or capricious” applies only to EPA decisions on how to implement the § 307(d) procedures. It cannot excuse failure to give adequate notice of a final rule.
. Small processing refineries produced 72% leaded gasoline, see Letter from Jimmie Peterson, supra note 72, at 2, J.A. at 1302, and used an average of 1.7 gpg during the second quarter of 1982, see Sobotka & Co., Inc., Comparison of Recent Lead Use By Small Refiners with Lead Use Under the Current Rule and Under the New Rule 1 (Oct. 27, 1982), J.A. at 1464, 1464. They therefore used 1.7 gpg/.72 = 2.36 gplg.
. SRTF Comments, supra note 67, at 10, J.A. at 748.
. This assumption was apparently wrong. At oral argument, counsel for SRTF stated that only two or three small refiners had been able to purchase lead credits and that apparently only one large refiner was as yet selling these credits. Transcript of Oral Argument, Jan. 17, 1983, at 7.
. At oral argument, EPA noted that its regulations permit small refiners to exceed 1.90 gplg for short periods of time, so long as they average 1.90 gplg or less for the five-month “compliance period” from November 1, 1982 to March 31, 1983. It argued that small refiners therefore had five months, not two days, to purchase lead credits. Transcript of Oral Argument, supra note 105, at 55-56. We believe, however, that small refiners could not reasonably choose to use illegal amounts of lead in the hope they could cure the violation by buying lead credits in the future.
. In light of this conclusion, we do not reach SRTF’s claim that much of the record evidence concerning the interim standard was docketed after the close of the comment period. Cf. part IV.F supra (EPA’s late docketing of evidence to support the final standard was improper, but not grounds for reversal).
We also do not reach Plateau’s complaint that EPA lacked proper reason under the APA, 5 U.S.C. § 553(d), to issue the interim standard without 30-days advance notice. We note a puzzle, however. Both Plateau and EPA assume that § 553(d) applies to this case. It does not, and Clean Air Act § 307(d), which does apply, does not require 30-days advance notice. On the other hand, nothing in the legislative history of § 307(d) explains this omission, and it may well have been inadvertent. We leave to another day the question whether Clean Air Act § 307(d) should be read to incorporate the APA’s 30-day notice period.
. We need not address in this case the existence or scope of any possible exceptions to this general rule. A court can, of course, give the agency specific guidance on some issues that will arise on remand. For example, in this case we suggested in' our order of January 26, 1983 that EPA could issue immediately a temporary lead standard not “significantly below previous lead-use levels.” See also National Ass’n of Recycling Indus. v. ICC, 704 F.2d 638, 640 (D.C.Cir.1983) (per curiam) (ICC improperly approved rail rate; court leaves the rate in effect pending ICC determination of the correct rate and instructs the ICC to include a refund provision in its final order).
. EPA may also, of course, institute regular rulemaking procedures with a view to adopting a more stringent interim standard, and may (within reason and statute) expedite such a rulemaking.
. See also PPG Indus. v. Costle, 659 F.2d 1239, 1250 (D.C.Cir.1981); Weyerhaeuser Co. v. Costle, 590 F.2d 1011, 1031 (D.C.Cir.1978); South Terminal Corp. v. EPA, 504 F.2d 646, 659 (1st Cir.1974) (the first case to use the “logical outgrowth” formula).
. See, e.g., Transcript of Lead Phasedown Hearing, supra note 100, at 177-81 (oral comments of United Refining Co.); United Refining Co. Comments, Oct. 4, 1982, at 2-9, J.A. at 413, 415-22; Environmental Defense Fund Comments, supra note 69, at 12, J.A. at 788; Crown Central Petroleum Corp. Comments, supra note 82, at 8, J.A. at 879; United States Department of Justice Comments, supra note 87, at 8-11, J.A. at 607-10; Farmland Industries, Inc. Comments, Sept. 24, 1982, at 2, J.A. at 406, 407.
. Part III.A supra discusses notice requirements under Clean Air Act § 307(d)(3).
. It would have been better yet for EPA to issue a supplemental notice describing the loopholes it intended to close. Such a notice, if issued early enough to permit responsive comments, need not have extended the overall length of the rulemaking. See Sierra Club v. Costle, 657 F.2d 298, 358-60 (D.C.Cir.1981) (finding no error in EPA’s giving notice in the middle of a rulemaking that it had reassessed certain assumptions).
. See, e.g., notes 74-76 supra and accompanying text (describing SRTF responses to a Sobotka study).
. Asamera Oil (U.S.) Inc. Comments, supra note 69, at 11-13, J.A. at 392-94. The Department of Justice also noted this problem, but EPA has not adverted to the Justice Department’s comments either in its explanation of the past ownership requirement in the Federal Register or in its briefs to this court. United States Department of Justice Comments, supra note 87, at 9, 11, J.A. at 608, 610.
.See also American Iron & Steel Inst. v. EPA, 568 F.2d 284, 291 (3d Cir.1977) (agency notice must fairly apprise interested persons “of all significant subjects and issues involved”; EPA’s notice of proposed effluent regulations for “Iron and Steel Manufacturing” did not fairly apprise makers of specialty alloy steel that they would be subject to separate regulation); Kollett v. Harris, 619 F.2d 134, 144 (1st Cir.1980) (general reference to income attribution in proposal for social security regulation was not adequate notice that agency might impute certain parental income to disabled child).
. See Platt’s Oil Marketing Bulletin, Sept. 27, 1982, at 3, col. 1.
. Simmons Reply Brief at 4 n. 3.
. See 47 Fed.Reg. at 49,326/1 (“Since January 1982, two of these refineries have been sold and the sale of a third announced, according to [Asamera].”). The other small refinery that Asamera believed had been sold by a large refiner is now owned by Giant Industries, Simmons asserts that Giant’s facility does not qualify as a “small refinery” for reasons other than the past ownership requirement, Simmons Brief at 24-25, and Giant and EPA concede the point, see EPA Brief in Response to Simmons at 6 n. 6; Giant Industries Brief at 4 n. 1.
.Clean Air Act Amendments of 1977, supra note 6, sec. 222(a), § 211(f)(5), 91 Stat. at 764.
. See S.252, 95th Cong., 1st Sess. § 36 (1977), 3 Leg.Hist.p75, 679-80.
. In earlier debate, Senator Muskie explained *247the need for § 211(f) at greater length. Apparently, oil companies had replaced tetraethyl lead with the additive “MMT” in order to produce high-octane gasoline without using lead. The auto companies complained that this new additive was damaging their catalytic converters. Senator Muskie explained:
[0]il companies have recently started adding MMT to unleaded gasoline to raise its octane rating. This was done without regard to MMT’s effect on the over 25 million catalysts in cars on the road.
I consider this a total breach of good faith by the oil companies, and an outrageous act against the American consumer.
The committee' ... required] the removal of all additives introduced into unleaded fuel since 1975 when catalysts were first introduced. This ban can be waived by the Administrator if ... the additive will not impair the emission performance of vehicles produced in 1975 or afterward.
123 Cong.Rec. 18,034 (1977), 3 Leg.Hist. 759.
. We have earlier noted the apparent haste with which the Conference Committee considered the 1977 amendments to the Clean Air Act. See note 35 supra.
. EPA may also, of course, institute regular rulemaking procedures with a view to adopting a more stringent interim standard, and may (within reason and statute) expedite such a rulemaking,
. The Comments of the Department of Justice, at 9, 11 (Oct. 8, 1982), J.A. at 599, 608, 610, suggested a similar change, but EPA has not adverted to those comments either in its explanation of the past ownership rule in the Federal Register or in its briefs to this court.