Full opinion text
OPINION OF THE COURT Before SEITZ, Chief Judge, ROSENN, Circuit Judge and MEANOR, District Judge. SEITZ, Chief Judge. These are petitions to review Environmental Protection Agency regulations governing certain manufacturing processes within the iron and steel industry. The regulations, which were published in the Federal Register on March 29, 1976, establish maximum permissible quantities of pollutant which may be discharged by operations performing the designated processes. The regulations also survey the available pollution control techniques, and specify techniques — the “best practicable control technology currently available” (BPCTCA) —which might be used in meeting the prescribed effluent limitations. In issuing the regulations, the EPA was exercising its statutory mandate to establish effluent limitations requiring the application of the BPCTCA by July 1, 1977. 33 U.S.C. § 1311(b)(1)(A). It is these regulations as to BPCTCA which are under review here. The regulations promulgated by the EPA also contain proposed: a) effluent limitations and guidelines as to the application of the “best available technology economically achieveable” by July 1,1983 b) standards of performance for new point sources and c) pretreatment standards for existing sources and for new sources. The regulations governing the application of the BPCTCA are in “interim final” form, which is to say that, while they were to take effect immediately upon promulgation, the EPA is now in the process of considering “final” regulations covering the same subject matter. The regulations state that the reasons for this novel procedural form is that “[t]he Agency is subject to an order of the United States District Court for the District of Columbia entered in Natural Resources Defense Council v. Train, et a 1. (Cv. No.1609-73) which requires the promulgation of regulations for this industry category no later than March 15, 1976.” 41 Fed. Reg. 13004. The regulations deal only with certain processes of the iron and carbon steel and specialty steel (ferroalloy and stainless steel) industries, namely forming and finishing processes, and with certain steelmaking processes within the specialty steel industry. An earlier phase (“phase I”) of the regulations, published “on June 28, 1974, dealt with steelmaking processes within the iron and carbon steel industries. This court considered the latter regulations in American Iron and Steel Institute v. EPA (AISI I), 526 F.2d 1027 (3d Cir. 1975), and remanded them to the EPA for reconsideration in certain respects. I. Some petitioners have argued that the “interim final” regulations are invalid because they were not promulgated in accordance with the Administrative Procedure Act. The relevant procedural history may be outlined as follows. Since November 15, 1973, the EPA has been under order of the district court for the District of Columbia, in Natural Resources Defense Council v. EPA to promulgate regulations governing the processes of the iron and steel industry covered by the present regulations. The EPA was originally under order to promulgate these regulations in 1974, but it has asked for and received several extensions of the deadlines. On August 21,1975, the EPA published in the Federal Register “advance notice of intent to propose and promulgate effluent limitations and guidelines for existing sources.” 40 Fed.Reg. 36708. The notice of “proposed rulemaking” (ANPR) then proceeded to list the subcategories of the industry for which tentative regulations were being set forth. The EPA engaged in extensive reevaluation and revision of the tentative limitations and guidelines after they were published, partly in response to substantial sentiment within the agency that they were too lenient. The EPA did not, however, propose the revised regulations for further notice and comment. It also stipulated that the interim final regulations would be effective immediately, and thus did not allow the 30 day interval between publication date and date of effectiveness usually required by the APA. 5 U.S.C. § 553(d). The regulations contain a statement that, due to the pendency of the court order and the need to expedite the effectuation of the Act, the Agency had determined that it was “impracticable and contrary to the public interest” “to develop and publish regulations ... in proposed form [and] to provide a 30 day comment period.” The Agency also stated that there was “good cause ... for these regulations to become effective immediately upon publication.” 41 Fed.Reg. 13004. Petitioners in Nos. 76-1386, 76-1757 and 76-2176, (“the Companies”) — who are steel companies and the American Iron and Steel Institute — have made two challenges to the validity of the “interim final” regulations under the APA which require discussion. They first argue that any regulations governing specialty steel are invalid because the ANPR failed to give sufficient indication that the Agency was considering regulations as to the specialty steel segment of the industry. This court delineated -the purposes of the APA’s notice and comment requirement in Texaco, Inc. v. FPC, 412 F.2d 740, 744 (1969): “Section 553 was enacted to give the public an opportunity to participate in the rule-making process. It also enables the agency promulgating the rule to educate itself before establishing rules and procedures which have a substantial impact on those regulated.” In evaluating the Companies’ contention that the ANPR was insufficient to enable the public to effectively participate in the rule-making process, we must determine whether the notice given was “sufficient to fairly apprise interested parties” of all significant subjects and issues involved. Senate Rept. No. 752, 79th Congress, 1st Session, at 16 (1945). The adequacy of the notice as to specialty steel really encompasses two questions: whether there was adequate notice that the making of specialty steel would be covered by the regulations, and whether there was adequate notice that the forming and finishing of specialty steel would be covered by the regulations. With respect to the forming and finishing of specialty steel, the ANPR set forth “Proposed Effluent Guidelines and Standards” for the “Iron and Steel Manufacturing Point Source Category.” As noted above, the EPA was making regulations under order of the court in Natural Resources Defense Council v. EPA. In ordering the EPA to make regulations, the court listed various “categories”: while category 19 was “Iron and Steel Manufacturing”, category 23 was “Ferroalloy Manufacturing” — what we have called specialty steel. Thus, an interested person who read that the EPA was giving advance notice of intent to promulgate regulations for the “Iron and Steel Manufacturing Point Source Category” would be misled into thinking that regulations for “Ferroalloy Manufacturing” were not involved. This misimpression would be fostered by the fact that the ANPR said that “[i]n developing the requisite data to support effluent limitations, guidelines and standards EPA commissioned a study and report entitled ‘Development Document for Effluent' Limitations Guidelines and New Source Performance Standards — Iron and Steel Industry: Hot Forming and Cold Finishing Segment’ prepared by Cyrus Wm. Rice Division, NUS Corporation.” The EPA had commissioned an entirely separate study of the specialty steel industry — by Datagraphics, Inc. Our conclusion that the ANPR would not apprise an interested person that the specialty steel industry would be covered by the regulations on forming and finishing processes applies a fortiori with respect to the portions of the interim final regulations which deal with the making of specialty steel. The ANPR gave no indication that the Agency intended to make any regulations on steelmaking, whether in the carbon steel or specialty steel industry: all the enumerated subcategories of the “iron and steel manufacturing” industry pertained to processes other than steelmaking. The interim final regulations, however, do contain regulations governing three processes by which specialty steel is made. Since the ANPR gave no indication that the regulations would deal with steelmaking, an interested person would not be able to make comments which could assist the EPA in formulating these regulations. EPA argues that even if the ANPR was insufficient to apprise an interested person that specialty steel would be covered by the regulations, the regulations may be upheld under an exception to the APA’s notice and comment requirement. 5 U.S.C. § 553(b) says that: “General notice of proposed rule making shall be published in the Federal Register . . Except when notice or hearing is required by statute, this subsection does not apply . (b) when the agency for good cause finds that notice and public procedure are impracticable, unnecessary, or contrary to the public interest.” This exception is to be narrowly construed. As is stated in the Senate Report on a precursor of the final APA: “Impracticable” means a situation in which the due and required execution of the agency functions would be unavoidably prevented by its undertaking public rule-making proceedings “Public interest” supplements the terms “impracticable” or “unnecessary”; it requires that public rule-making procedures shall not prevent an agency from operating and that, on the other hand, lack of public interest in rule making warrants an agency to dispense with public procedure. Senate Rept. No. 752, 79th Congress, 1st Session at 16. (1945) In arguing that it was “impracticable and contrary to the public interest” to provide notice and an opportunity for comment, the EPA relies chiefly on the fact that it had been ordered to promulgate regulations by March 15, 1976 by the'court in Natural Resources Defense Council v. EPA. But we find this argument unconvincing. The EPA had been under order to promulgate regulations for the specialty steel industry since November 15,1973, and it received its contractor’s study of the specialty steel industry in January, 1974. So EPA was aware of the necessity for rapid promulgation of the limitations on specialty steel in November of 1973, and had the basic contractor’s study in January of 1974. But the interim final regulations were not published until March of 1976. Moreover, the ANPR did in fact contain detailed tentative regulations on the forming and finishing of iron and carbon steel, a subject on which EPA had obtained its contractor’s report in July of 1974 — 6 months after it received its contractor’s report on the carbon steel industry. Under these conditions, EPA has failed to show that it may be exempted from the APA’s usual requirement of notice and comment. In sum, EPA may not be exempted from the APA’s rulemaking requirement of notice and comment, and it has failed to give adequate notice that the interim final regulations would govern the specialty steel industry. The interim final regulations are thus invalid insofar as they apply to the specialty steel industry. Apart from their challenge to the regulations concerning specialty steel, the Companies also allege that the ANPR did not give sufficient indication of the following two issues: “(1) whether filtration and tight recycle (the ‘10/10/5 model’) are practicable for the steel industry to construct and operate by July 1, 1977 and (2) if so, the treatment levels achievable by such technology.” The history behind the adoption of the 10/10/5 model is as follows. After the ANPR was published, the EPA, in its internal review of the tentative regulations found that: “there were substantial errors in the limitations. The advance notice was improperly biased; costing inaccuracies caused an invalid comparison between clarifier and filter technologies; plants had already installed both filtration and recycle technology; and, generally, both BPT and BAT were set at improperly low levels, levels at which a quarter of the Nation’s steel production would already be at BAT, with many more facilities at BPT ...” EPA’s Brief at 90. The Agency’s response was to make the regulations substantially more stringent. The Companies especially question the use of “filtration and tight recycle” technology, which is used in Subparts M (hot forming-primary subcategory), N (hot forming-section subcategory), 0 (hot forming-flat subcategory and P (pipe and tubes subcategory). We agree with the court in International Harvester Co. v. Ruckelshaus, 155 U.S.App.D.C. 411, 428, 478 F.2d 615, 632 n.51 (1973), that the submission of a proposed rule for comment does not of necessity bind an agency to undertake a new round of notice and comment before it adopts a rule which is different — even substantially different — from the proposed rule. As we have noted above, the adequacy of the notice must be tested by determining whether it would fairly apprise interested persons of the “subjects and issues” before the Agency. Judged by this standard, we conclude that the ANPR was sufficient to apprise interested persons of the issues involved in the EPA’s decision to adopt more stringent treatment technologies than had been embodied in the ANPR. The Companies’ concern is focussed on the EPA’s determination that recycling and filtration technology is BPCTCA with respect to some forming and finishing subcategories. These techniques, as we have noted, are prescribed as BPCTCA for Subparts M through P of the industry, and the control technology specified for these subcategories is similar. The chief pollutants produced by all four processes are suspended solids, oil and grease. The interim final regulations specify that the first step of the BPCTCA for all four processes is a primary scale pit — a settling unit — with oil skimming equipment applied to the effluent in the scale pit. This was also specified as the first step of the BPCTCA in the ANPR. The interim final regulations specify that after the scale pit, part of the effluent is recycled for use in the steel making process; this is one of petitioners’ points of concern. The interim final regulations state that the remaining effluent in the scale pit goes to a clarifier; the use of a clarifier was also specified in the ANPR. The underflow from the clarifier is to be vacuumed filtered, as the ANPR had also stated. Finally, the overflow from the clarifier is filtered and discharged. The use of filtration is a second point of concern to petitioners, and is peculiar to the interim final regulations. We conclude that the ANPR was sufficient to apprise interested persons that there was an issue as to whether recycling of partially clarified effluent was BPCTCA. In the first place, the suggested BPCTCA in the ANPR regulations for three of the four subcategories of the industry here involved — N, O and P — include some recycling of effluent. Moreover, in surveying the treatment technology currently in place in existing plants, the ANPR notice mentioned recycling with respect to all subcategories of the iron and steel industry for which the interim final regulations later specified recycling as part of the BPCTCA. An interested person thus should have been aware that recycling was potentially a significant component of the ultimate regulations as to BPCTCA. The use of filtration upon the overflow from the clarifier presents a closer question, since the ANPR did not suggest that filtration of the clarifier overflow would be part of the BPCTCA for any of the hot-forming subcategories (Subparts M, N and 0) or the pipe and tubes subcategory (Subpart P). Nevertheless, we conclude that the notice was sufficient, since, as with the use of recycle, the ANPR did inform the public that high rate filtration technology was already in place in plants in those of three subcategories (M, N and 0). With respect to the meaning of BPCTCA, the Senate Report to the bill which ultimately became the Federal Water Pollution Control Act Amendments of 1972 said that: “[T]he Administrator should establish the range of best practicable levels based upon the average of the best existing performance by plants of various sizes, ages, and unit processes within each industrial category.” Senate Report No. 92-414, 92d Congress, 1st Session (1971), p. 50. The statements in the ANPR that “a few plants will have high rate filters to treat either scale pit or clarifier effluents”, and that the “range of treatment technology currently practiced in existing plants includes all the items discussed above” — including the use of high rate filters — may not be a full statement that the plants now employing filter technology within each subcategory of the industry are “of various sizes [and] ages.” But this is little more than to say that EPA had not yet determined that filtration was BPCTCA when it published the August notice. The notice does fairly apprise interested parties that there was an issue as to whether filtration was BPCTCA because it states that some of the best existing plants use filters. In sum, we conclude that the use of the 10/10/5 model did not violate the APA’s requirement of notice and hearing. Pennsylvania’s Department of Environmental Resources, (“Pennsylvania”) the Petitioner in No. 76-1751, questions the sufficiency under the APA of the Agency’s decision to exempt plants in the Mahoning River Valley region of Eastern Ohio “from the effluent limitations based on best practicable control technology currently available.” We need discuss only two of Pennsylvania’s arguments: 1) that the EPA did not provide notice and an opportunity for comment on the special treatment afforded plants in the Mahoning Valley 2) that adequate judicial review of the Mahoning Valley exemption is precluded by the EPA’s failure to include a “concise general statement of [the] basis and purpose” of the action as required by the APA. 5 U.S.C. § 553(c). 1. The interim final regulations state that: The relief granted from severe economic impact in the Mahoning River Valley region, which impact is likely to occur absent such relief, is the exemption of point sources located within that region from the effluent limitations based on best practicable control technology currently available. Nevertheless, the Agency fully expects that authorities granting permits, pursuant to section 402 of the Federal Water Pollution Control Act, as amended, shall not allow point sources in that region to discharge pollutants in any greater amounts than are currently being discharged by those sources. 41 Fed.Reg. 12995. The only indication in the ANPR that such an action might be taken is the following: One commenter has claimed that the proposed guidelines will result in the loss of 12,000 jobs from the steel employment in the Mahoning River Valley region. Furthermore, the commenter asserts that “there is ample justification for adding to the guidelines, a subcategory based on the age of the facility.” The Agency intends to secure and evaluate additional information on possible economic impacts in this region and would consider revision of the regulations if the information appears to warrant this action. 40 Fed.Reg. 36723. Pennsylvania argues that in reading the language “would consider revision of the regulations” an interested member of the public might well come to the conclusion that the EPA would not consider whether to give Mahoning Valley special treatment until the rulemaking then noticed had already been completed. Thus, the public could well have been misled into thinking that comments on the Mahoning Valley question were not appropriate during the present rulemaking proceedings. Pennsylvania’s argument hinges on the meaning of the word “regulations” in the ANPR. If the word refers to the regulations to be promulgated at the end of the noticed rulemaking, then a member of the public might indeed be misled into thinking that he should defer comment on the Ma-honing Valley question. On the other hand, if the word refers to the regulations then proposed, a member of the public would be on notice that immediate comment was appropriate. We note that the ANPR states that “[t]he regulations set forth below when promulgated will amend 40 CFR, part 420.” 40 Fed.Reg. 36708 (emphasis added). This, read in conjunction with the statement that the EPA would consider revision of the “regulations” to take account of any special conditions in the Mahoning Valley would indicaté that immediate comment was appropriate. Thus, we conclude that there was sufficient notice to “fairly apprise” the public that the Agency might afford special treatment to facilities in the Mahoning Valley. We confess, however, that the language in the ANPR is hardly a model of clarity. 2. Pennsylvania has also argued that the Mahoning Valley exemption is invalid because the EPA did not give a “concise general statement of [the] basis and purpose” of the action as required by the APA. 5 U.S.C. § 553(c). This argument is meritless. The interim final regulations state that: Tentative analysis of the available data leads to the conclusion that conditions in the Mahoning River Valley region are unique with respect to the physical and geographical characteristics of the region, physical and operating characteristics of the facilities located therein, and the importance of the facilities to the economy of the region. Tentative analysis of the available data and the consultant’s evaluation thereof appears to support the contention that mandatory compliance with effluent limitations guidelines which do not take into account these factors is likely to result in severe economic dislocation within the Mahoning River Valley region. In addition to similar economic disadvantages resulting from age and size characteristics, facilities in the region appear to share economic disadvantages caused by locational characteristics. These include the movement of markets away from the region, constrained access to raw materials due to the unavailability of waterborne transportation and required transshipment by rail, and space limitations which prohibit major expansion of existing facilities. 41 Fed.Reg. 12994. This statement is clearly sufficiently detailed to allow for “searching judicial scrutiny.” Amoco Oil v. EPA, 163 U.S.App.D.C. 162, 179, 501 F.2d 722, 739 (1974). II. We now turn to the issues raised by the various petitioners as to whether the EPA has properly carried out its functions under the Water Pollution Control Act. The Companies argue that the regulations are invalid because they specify only a maximum permissible level of éffluent discharge, and do not indicate a range of permissible effluent levels by providing guidelines as to how much more severe controls may be required of any particular plant by the authority granting a pollution discharge permit under § 402 of the FWPCA. The argument thus raises questions as to the proper relation between § 301, which is the Agency’s authority for promulgating specific effluent limitations requiring application of BPCTCA, and § 304, which directs the Agency to “specify factors to be taken into account in determining the control measures and practices applicable to point sources.” The Companies rely chiefly on this court’s decision in AISI I which held EPA’s regulations on iron and carbon steelmaking processes invalid because they failed to provide guidelines “to guide the permit-issuing authorities in deciding whether, and by how much, the limitation to be applied to any individual point source is more stringent than the base level.” 526 F.2d at 1045. Since briefing and oral argument in this case, however, the panel in the earlier case has granted a motion to recall its mandate, and has held that the EPA may specify a maximum permissible level of effluent discharge without providing guidelines as to how permit issuers are to select more stringent levels of control. American Iron and Steel Institute v. EPA, slip opinion (July 29, 1977). This court now stands with the other courts of appeals which have considered this question. See Hooker Chemicals v. Train, 537 F.2d 620 (2d Cir. 1976); American Frozen Food Institute v. Train, 176 U.S.App.D.C. 105, 539 F.2d 107 (1976); FMC Corporation v. Train, 539 F.2d 973 (4th Cir. 1976); American Petroleum Institute v. E. P. A., 540 F.2d 1023 (10th Cir. 1976). The holding of the prior panel on the motion to recall its mandate is binding upon us, and requires that we reject contentions. Consequently, we need not engage in any discussion of the merits of these contentions. III. The Companies have also argued that EPA did not properly evaluate the costs and benefits relevant to identifying the BPCTCA. In evaluating these arguments, we must follow the familiar standards governing judicial review of agency action under the APA. § 706 of the APA provides that “[t]he reviewing court shall hold unlawful and set aside agency action, findings, and conclusions found to be . arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law . . . ” As the Supreme Court said in Citizens to Preserve Overton Park v. Volpe, 401 U.S. 402, 415, 91 S.Ct. 814, 823, 28 L.Ed.2d 136 (1971), an agency’s “decision is entitled to a presumption of regularity, (citations omitted). But that presumption is not to shield [its] action from a thorough, probing, in-depth review.” In conducting this review, we will not overturn agency action because the underlying reasoning is not fully set forth, if we can fairly discern the basis for the action. Bowman Transportation, Inc. v. Arkansas-Best Freight System, Inc., 419 U.S. 281, 285-6, 95 S.Ct. 438, 42 L.Ed.2d 447 (1974). But the touchstone of our review, both as to the Agency’s consideration of the issues and the factual predicates of this consideration must be the administrative record. In deciding to accept or reject petitioners’ contentions we must look “to the record that was considered by the agency — not to post hoc rationalizations of counsel or even agency members and not to evidentiary materials that were not considered by the agency.” Dry Color Mfrs. Assn., Inc. v. Department of Labor, 486 F.2d 98, 104 n.8 (3d Cir. 1973). Section 304(b) of the FWPCA, as amended, states that, in assessing the BPCTCA, the EPA must consider: . the total cost of application of technology in relation to the effluent reduction benefits to be achieved from such application, and shall also take into account the age of equipment and facilities involved, the process employed, the engineering aspects of the application of various types of control techniques, process changes, non-water quality environmental impact . ., and such other factors as the Administrator deems appropriate. As this court noted in AISI I, Senator Muskie, the Chairman of the Senate Subcommittee on Air and Water Pollution, stated in support of the House-Senate Conference Committee Report: . The balancing test between total cost end effluent reduction benefits is intended to limit the application of technology only where the additional degree of effluent reduction is wholly out of proportion to the costs of achieving such marginal level of reduction for any class or category of sources. The Conferees agreed upon this limited cost-benefit analysis in order to maintain uniformity within a class and category of point sources subject to effluent limitations and to avoid imposing on the Administrator any requirement to consider the location of sources within a category or to assess water quality impact of effluent controls, or to determine the economic impact of controls on any individual plant in a single community. Another factor which lead the court in AISI I, 526 F.2d at 1052, to conclude that “the Administrator must have broad discretion in weighing the costs and benefits” is the following passage from the Senate Report which discusses § 302 of the Act — where a more specific cost-benefit analysis is required than with § 304: The Committee recognizes that no mathematical balance can be achieved in considering relative costs and benefits nor would any precise formula be desirable, but in each case the Administrator or the State will be able to determine whether there is any reasonable connection at all between the costs which a particular effluent limitation would impose and any benefits (including the attainment of natural water quality) which might be derived. 1. The Companies have argued that the interim final regulations are invalid because the EPA failed to reach any final conclusion as to whether the costs of compliance are justified by the benefits. But the record shows that the EPA did investigate the harmful effects of the various pollutants from iron and steel facilities. Moreover, the Agency calculated the contribution of selected additional levels of technology to eliminating the discharge of the various pollutants, and compared this contribution to the costs of these technologies. The Companies claim that this procedure was an “intellectual fraud” since, as the Agency expressly warned, inferences could not be drawn as to the cost-effectiveness of levels of treatment other than those which the Agency had selected to present. But given the EPA’s “broad discretion” in weighing costs and benefits, and the fact that the Agency did evaluate several treatment technologies within each category, we conclude that the regulations are not invalid on the grounds that the Agency did not properly evaluate the cost effectiveness of possible treatment technologies. 2. The Companies have also argued that the EPA failed to properly consider the factors relevant to determining how the iron and steel industry should be broken down into subcategories, prior to identifying the BPCTCA for each subcategory. EPA’s methodology may be described as follows. The Agency took a survey of existing treatment facilities and their performances to develop a list of the best plants: the list was ultimately narrowed to 68 plants with 111 operations of various ages, sizes, unit processes and in-place treatment technologies. The industry was studied “for the purpose of determining whether separate limitations are appropriate for different segments within the category.”. 41 Fed.Reg. 12990. The following factors were considered in determining possible subcategorization of the industry: manufacturing processes and equipment, products, raw materials, wastewater characteristics, waste treatability, gas cleaning equipment, size and age, land availability (location), and process water usage. The EPA ultimately determined that “[t]he individual processes, products, and the waste-water characteristics comprise the most significant factors in the categorization” of the industry, but also concluded that some of the other factors were helpful in supporting its subcategorization on the basis of the above factors. For example, with respect to the relevance of process water usage, the interim final regulations state that: Examination of the available data indicates that within well defined ranges process water usage can be directly correlated to the various manufacturing operations. This correlation verifies the basis of the subcategorization scheme by manufacturing processes. Differences in scale (see size factor) of a categorized manufacturing process was considered. The results indicated that on a per ton of steel basis, process water usage is not dependent upon the scale of the manufacturing operation. It was observed though, that much larger volumes of process cooling water are generally required to cool the hot forming machinery than that which is needed for the cold forming operations, thus further substantiating the subcategorization by manufacturing process. Considerations of age, location and raw materials revealed no discernible differences in process water usage. Process water usage parallels the subcategorization by final product considerations (see final products factor) where data revealed that for particular product requirements well defined manufacturing processes must be employed. 41 Fed.Reg. 12993-4. On the other hand, the EPA found that some of the factors considered were not useful bases for subcategorization. As to size and age, the interim final regulations state: Plant size and age, per se, are not viable factors for subcategorization of the iron and steel industry. Information compiled during this study and previous steel industry investigations do not reveal any discernible relationship between these factors and raw waste loads, effluent quality, treatability, or any other basis for subcategorization. Although specialty steel plants do tend to be smaller than carbon steel plants, the type of steel produced has a greater impact on the waste loads and water use than does the size per se. Size was considered as a possible factor for subcategorization but from analysis of the compiled data size, per se, does not justify categorization. Throughout the steel industry mills vary greatly in physical size, layout and product size. However, these considerations revealed no relationship to process water usage, discharge rate, effluent quality or any other pertinent factors. Age as a factor might be expected to be at least amenable to quantitative identification and interpretation, but the extensive investigation of the industry does not indicate that age alone is a factor. The steel industry is old. Some of the old mills still incorporate early operating ideas and practices. However, other old mills are very new in that they have incorporated the latest operating ideas and practices. Nevertheless, most older mills have been updated by internal changes in process, design, and equipment. Therefore, to say that a mill was built 50 years ago and is 50 years old is not particularly meaningful in terms of interpreting mill practices. In particular, no consistent pattern between mill age and raw waste characteristics was found. 41 Fed.Reg. 12993. Contrary to the Companies’ contentions, we believe that these passages in the interim final regulations, and others appearing at 41 Fed.Reg. 12992-5, demonstrate that the EPA did consider subcategorizing the iron and steel industry on the basis of factors other than manufacturing processes. Moreover, the Companies have not suggested that the record shows that there is an insufficient factual basis for the EPA’s treatment of the subcategorization question. As EPA concedes, the chosen subcategorization — like any subdivision of the industry — does not take into account all of “the innumerable differences within the particular subcategories.” AISI I, 526 F.2d at 1046. But this does not in itself imply that the EPA has failed to properly exercise its functions under the FWPCA. We must, however, agree with the Companies’ contention that, to the extent it considered age as a relevant factor, the EPA did not consider age as it might pertain to the cost or feasibility of retrofitting plants with new pollution control technology, as required by this court’s decision in AISI I, 526 F.2d at 1048. As in the previous case, the interim final regulations recite that the EPA has found that age per se is not relevant to factors such as type of pollutants produced. But the EPA has not specifically found that age has no bearing on the cost or feasibility of retrofitting plants with pollution control technology, and a finding on this issue is mandated by our earlier decision. The interim final regulations recite that the concept of plant age is not amenable to interpretation because “most older mills have been updated by internal changes in process, design, and equipment. Therefore, to say that a mill was built over 50 years ago and is 50 years old is not particularly meaningful in terms of interpreting mill practices.” In its brief, EPA suggests that this is a sufficient answer to the retrofit question: age has little or no bearing on retrofit because the concept of plant age itself is difficult to define. Were we writing on a clean slate, we might find this argument convincing. But since the facts in this case cannot be properly distinguished from the facts in the earlier case we must reject EPA’s contention. The discussion of the bearing of age and size in the interim final regulations does not differ in material respects from EPA’s discussion of this issue in the regulations which were reviewed in the previous case. Compare 39 Fed.Reg. 24116 with 41 Fed. Reg. 12993, 13002. EPA argues that new data submitted to it after the earlier regulations were remanded “substantiates [its] conclusion and obviates the need for further investigation.” But again, the interim final regulations do not expressly address the retrofit problem. EPA also argues that a study by Arthur D. Little, Inc., which was commissioned by AISI, reinforces the Agency’s findings, since this study states that: “[i]n the iron and steel industry it is difficult to define the age of a plant, because many of the unit operations were installed at different times and are also periodically rebuilt on different schedules. Thus, by definition, the age of steel facilities should offer only limited benefits as a means of categorizing plants for purposes of standard setting or impact analysis.” But the Companies are not bound to accept the conclusion that age is not relevant to cost or feasibility of retrofit merely because of any conclusion contained in a report commissioned by AISI. They have a right, under our earlier decision, to a finding on the retrofit question by the EPA. We note, however, that we have not dismissed the EPA’s resolution of the retrofit question on the merits: we merely require that the Agency reexamine the relevance of age specifically as it bears on retrofit. As this court said in AISI I, 526 F.2d at 1048: “[i]f [EPA] concludes on remand that age is not relevant to the cost or feasibility of retrofitting plans [sic], as well as to the processes and treatment system, that conclusion will be valid to the extent it is supported by the record.” We note that the reconsideration of the bearing of age on the cost or feasibility of retrofitting plants shall include consideration of standards setting within subcategories as well as altered subcategorization. 3. The Companies have also questioned the EPA’s approach to determining BPCTCA within the subcategories. After EPA had decided upon a subcategorization of the industry, it determined the BPCTCA for each of these subcategories by calculating the average waste load at the exemplary plants, and “visualizing]” a treatment model for attaining the calculated effluent load. In visualizing a treatment scenario, the EPA arrived at a projected effluent flow, given an “achievable” concentration level. But it is important to emphasize that the interim final regulations do not require the use of treatment technologies with these particular flow and concentration levels. Indeed, the industry is free to employ any treatment technology it chooses, as long as this technology enables the industry to attain the prescribed level of effluent load. The Companies insist that determining BPCTCA by arriving at a maximum permissible load of effluents and “visualizing” a control technology which would meet these limitations is working backward: they argue that EPA should have first examined the technology in existence and then estimate how much this technology can contribute toward cleaning up effluent. We must reject this argument. In the first place, in visualizing control technologies, EPA specified measures which it had found were used in plants in each subcategory. We do not mean to imply that the EPA is always bound to select technologies already in place for BPCTCA: indeed, the legislative history of the FWPCA would seem to indicate that the EPA is not so restricted. But the fact remains that the EPA did select technology already in place in some plants for BPCTCA. Moreover, the EPA performed a cost-benefit analysis on the technology it selected through the challenged procedure, and this is ultimately what is required by the FWPCA. As long as EPA has performed the cost-benefit analysis required by the Act, and has selected a type of technology which may, under the Act, be a BPCTCA, it is irrelevant that it selected the BPCTCA by visualizing a control technology that would achieve desired effluent limitations. The Companies also argue that the EPA’s approach of determining the BPCTCA after calculating the average effluent load at the best plants in each subcategory “permitted [it] to derive limitations from effluent samples without analyzing how these particular treatment levels could be achieved at other plants, or investigating why they were not being [met] by other plants.” But this argument really amounts to a claim that it was “necessarily an abuse of discretion to base the regulations on results obtained from a few plants which were using the best technology” — a claim which was rejected in AISI I, 526 F.2d at 1057. It is true that as long as EPA focuses only on some plants in each subcategory there is a possibility that some differences between plants in the subcategory may be ignored: not all plants within a subcategory are perfectly typical of all plants within the subcategory — even once the industry has been subcategorized in a permissible manner. But petitioners have not made an adequate showing that the plants EPA surveyed do not constitute an adequate sampling of the subcategories they represent. 4. The Companies assert that the EPA’s cost estimates are defective because they fail to take into account costs for such factors as land acquisition, site clearance, utility interconnection between battery limits and process equipment, sales and use taxes, freight charges, supporting utility requirements such as boilers, instrumentation other than that for pH and fluoride control, and safety systems. They also stress that the Agency’s assumptions as to the characteristics of the sites on which plants are located were unrealistic, and that these assumptions thus caused the Agency’s cost figures to be inaccurate. EPA argues that the Companies’ contention is satisfactorily disposed of by the following passage from this court’s opinion in AISI I: A troublesome question is presented by petitioners’ contention that the costs estimated by the Administrator were artificially low because certain factors were excluded. Petitioners point to the fact that the Final Development Document listed several factors which admittedly affected costs, but which were excluded from the cost analysis. Petitioners contend that these factors, which include such things as land acquisition and site clearance costs, generate costs which are as large as those which the Administrator included in his estimates. However, petitioners have pointed to no data which would support their allegation as to the magnitude of these excluded costs factors. Furthermore, many of these factors were excluded because they were inherently site-specific or because they could not be evaluated. For these reasons, we do not believe we can conclude that the exclusion of these factors from the costs analysis was arbitrary or capricious. 526 F.2d at 1053. (footnotes omitted). The present case, however, is different from the situation presented in AISI I in two significant respects. First, the Companies have pointed to record data in the EPA’s study of plants in Ohio’s Mahoning River Valley which substantiates their claim that these costs are indeed significant. Second, as to whether the costs in question can be evaluated, EPA has presented in its brief a procedure for comparing its “generous” estimates of cost factors other than site-specific factors with the actual total (including site-specific) costs of reporting plants which meet or exceed the BPCTCA standards. • The EPA also argues that its treatment of site-specific costs was adequate because it did in fact include an adequate allowance for these costs. The ANPR states that: The most accurate way to determine the water pollution control facility costs and the site specific costs would be to calculate these costs for each operation at each of the many mills and plants. Such a determination is far beyond the scope of this ’effort and thus a method approximating these costs was developed. The contractor was conservative in estimating both the treatment facilities required to achieve the effluent quality and the cost of these facilities. This is borne out by the data submitted by the companies, relative to the plants visited which were meeting the proposed guidelines (Forming and Finishing Segment), shows that these plants on the average experienced total installation costs (Facility costs plus site-specific costs) and annual operating costs less than the facility costs alone (i. e., not including site specific costs) or the operating costs as estimated by the study contractor. For this analysis actual costs reported were adjusted to August 1971 dollars to be consistent with the costs used by the study contractor. An adjustment was made also to achieve consistency in the production capacity of the actual facility and the typical facility for which the treatment models were costed out. 40 Fed.Reg. 36724. But while EPA asserts that its contractor’s cost estimates were, on the average, adequate to cover site-specific factors, we are not directed to record evidence which would permit us to verify that the Agency actually employed the procedure set forth in its brief or another procedure to compare its model cost estimates with the actual total (including site-specific) costs of reporting plants which meet or exceed the BPCTCA standards. Indeed, EPA’s brief demonstrates that in several subcategories, the costs predicted from the model diverge significantly from the actual cost data from reporting plants. The case will thus be remanded for appropriate consideration of the factors originally not included by the Agency. See EPA’s Development Document for the Interim Final Regulations at 526-8. To the extent that the sufficiency of the allowance for certain factors which were excluded may not be adequately checked using the procedure presented in EPA’s brief or another procedure, the Agency may continue to exclude these factors. 5. The Companies contend that the EPA did not adequately consider the industry’s ability to find the funds necessary to install and operate the pollution control equipment required to achieve the effluent limitations specified in the interim final regulations. As EPA states in its brief, its analysis of the economic impact of its regulations is based upon its analysis of a study by Temple, Barker and Sloane, Inc. (TBS). The following data from the TBS report indicates the magnitude of the burdens EPA’s regulations will impose on the industry. Long run (1975-83) capital expenditures required to comply with the BPCTCA regulations now under review total $1.39 billion [all figures in 1975 dollars]; $0.96 billion will be required for operations and maintenance of this equipment over the same period. On the other hand, the capital for capacity expansion and modernization sufficient to meet anticipated steel shipment requirements and to be able to operate at planned levels of utilization comes to $27.50 billion over the period 1975-83. Revenues required to comply with the BPCTCA effluent limitations in the present regulations come to $2.59 billion for the period 1975-83, while $338.61 billion will be needed for baseline requirements associated with the production of iron and steel. TBS anticipates that $1.11 billion of the revenues required over the 1975-83 period to finance pollution control must come from external sources of funds, while $12.99 billion in external financing will be needed for baseline expenditures associated with iron and steel manufacturing. The extent of the burden imposed upon the industry is underscored by TBS’ statement that: “[d]uring the last decade the steel industry has depended upon external sources for less than $1.0 billion . . . Thus, the steel industry must attract significant amounts of capital to finance expansion and modernization of steel operations after more than a decade of negligible financing activity. Environmental regulations will increase these financing requirements.” Moreover, TBS noted that the industry’s ability to attract sufficient capital was diminished by the fact that: . the industry’s level of profitability has seriously lagged the average of all manufacturing with the exception of 1974 when the relaxation of price controls and worldwide shortages of both steel and basic raw materials enabled the steel industry to raise its prices and fully utilize its productive capacity. Overall steel industry rates of return in 1974 equalled the all-manufacturing average, with natural resources extraction and some non-steel activities (e.g., chemicals) far exceeding the returns on steel operations. Over the last decade, however, the steel industry’s rate of return on common equity has averaged one-half of that for all manufacturing firms when profitability is adjusted to reflect inflationary trends. The Companies have made several challenges to the Agency’s consideration of the financing problem. But we need consider only one: that the Agency relied on a February draft of the TBS report, which contains significant differences from the report dated as of March — the month in which the interim final regulations were promulgated. The February draft states that: [I]t is TBS’s opinion that the steel industry would be unable to finance its expansion, modernization and pollution control requirements without earning an adequate return on common equity at least equal to that realized by steel firms overall. Difficulties in financing these needs would be eased further if steel firms could achieve a return equal to other manufacturing firms, (emphasis added) On the other hand, the March report states that: “[i]t is TBS’s opinion that the steel industry will be unable to finance its expansion, modernization and pollution control requirements without earning a rate of return on common equity at least equal to those realized by other manufacturing firms.” (emphasis added) As EPA concedes before this Court, the “projected average nominal return on equity for the steel industry equals 72% of the projected average nominal return on equity by the manufacturing sector as a whole.” Thus, we cannot consider the difference between the February draft of the TBS study and the March report merely as a change in emphasis or of verbal shading. Nor can we consider the subject of the conditions under which the industry can meet all of its financing needs as peripheral to TBS’ basic conclusions. Thus, unless the differences between the drafts of the TBS report are irrelevant because EPA made a sufficient independent assessment of the financing problem, or unless we can infer that the EPA was aware of and considered the March report, we must remand the matter to the Agency for reconsideration of the financing problems raised by the interim final regulations. If the Agency based its assessment of the financing problem in substantial part on the assumption that it had the benefit of TBS’ expertise in evaluating the problem, and it did not in fact have the benefit of TBS’ ultimate conclusions, the regulations must be viewed as having been promulgated under a misimpression going to the heart of the Agency’s exercise of discretion. As to whether the EPA did make a sufficient independent assessment of the financing problem, it is important to note the following disclaimer in the March TBS report: This report is being released and circulated at approximately the same time as publication in the Federal Register of a notice of proposed and interim final rule making under sections 304(b) and 306 of the Act for the subject point source category. The study is not an official EPA publication. It will be considered along with the information contained in the Development Document and any comments received by EPA on either document before or during rule making proceedings necessary to establish final regulations. Prior to final promulgation of regulations, the accompanying study shall have standing in any EPA proceeding or court proceeding only to the extent that it represents the views of the contractor who studied the subject industry. It cannot be cited, referenced, or represented in any respect in any such proceeding as a statement of EPA’s views regarding the subject industry. Moreover, the Agency has argued before us that it “not only reviewed the TBS text but also analyzed the underlying data which remained unchanged.” However, several factors lead us to conclude that EPA did rely in substantial part on TBS’ expertise in evaluating the industry’s financing problem, and not simply on the data contained in the report. In the first place, EPA itself belittles the “standard” disclaimer contained in the March TBS report. Moreover, after completing a survey of the conclusions (rather than merely the data) contained in the March report, EPA’s brief states that “[t]he soundness of the TBS report and the Administrator’s reliance thereupon are clearly supported in the record.” We also note that the “Summary Economic Analysis of Effluent Guidelines for the Integrated Iron and Steel Industry” dated February 6,1976, which was the last full discussion of the financing problem actually prepared by the Agency, states that: “[t]he summary of economic effects that follows is prepared from preliminary results [of the TBS study] . This report summarizes the study in terms of the effects of the effluent guidelines on internal and external costs.” The “Summary Economic Analysis” states that its “baseline assumption” is that the industry will be able to earn “approximately a 9% return on stockholder equity.” As the Agency itself notes, the 9% figure is the projected rate of return on equity for the steel industry, (the rate stressed in the February TBS draft) rather than the manufacturing sector as a whole. In sum, we conclude that the EPA did rely on TBS’ expertise in evaluating the extent of the financing burden imposed by the interim final regulations. But we cannot infer that EPA ever received or considered the contents of the March draft of the TBS report. The Agency’s own “Summary Economic Analysis,” as we have noted, is dated February 6, 1976. Moreover, the interim final regulations state that: Studies of the economic impact of these regulations are under way and will be reported in the near future as separate reports entitled “Economic Analysis of Effluent Guidelines, Iron and Steel Industry” and “Economic Analysis of Effluent Guidelines, Specialty Steel Industry”. 41 Fed.Reg. 13001. The March TBS report is entitled “Economic Analysis of Proposed and Interim Final Effluent Guidelines, Integrated Iron and Steel Industry.” The matter is remanded to the Agency for reconsideration of the financing problems raised by the interim final regulations. 6. The Companies argue that the EPA did not adequately consider the evidence that combined treatment of waste streams (“central treatment”) should be adopted as BPCTCA. The interim final regulations state that: The comment has been made that these limitations require individual waste treatment facilities at each operating unit and prohibit the use of central waste treatment facilities which are more economical to construct and to operate. Central treatment facilities typically provide for equalization, neutralization, solid removal, and sludge dewatering. Other pollutants requiring removal are usually more efficiently and economically controlled or recovered by a pretreatment step applied to the segregated stream. These regulations have been constructed so as to permit a discharger to apply either approach. In the event that waste streams from various sources are combined for treatment or discharge, the quantity of each pollutant or pollutant property attributable to each controlled waste source (subcategories M through V) shall not exceed the specified limitation for that particular waste source. For example, if a plant’s production allows it to discharge 5 lbs. a day of tin from its terne plating operation, it would not be allowed to discharge 10 lbs. a day of tin because the terne line wastes were combined with cold rolling wastes for treatment. However, in the instance of the use of pickling wastes to assist in the breaking of emulsified oils from cold rolling wastes, some added waste load discharges are permitted by the regulation. The cost savings that could be achieved by the use of one waste stream to treat another waste stream was considered sufficient to justify permitting additional loads to be discharged. 41 Fed.Reg. 13002. The nature of EPA’s decision with respect to central treatment can fairly be inferred from this passage. EPA recognized that the use of central treatment facilities could result in significant savings, but was concerned that combining different waste streams would sometimes increase the total load of pollutants discharged because some pollutants would be diluted and thus made more difficult to treat. The Agency consequently provided that when waste streams are combined, no more pollutant load can be discharged than if the streams had been segregated. However, the Agency recognized that in one instance — combined treatment of cold rolling and pickling wastes— some added waste load discharge should be allowed. The Agency also concluded that, in some eases, some of the benefits of central treatment could be attained by incorporating a pretreatment step to remove such wastes as would be unduly diluted before combining waste streams for central treatment. While we believe that the Agency could have set forth its reasoning on the cost-benefit question with greater clarity, we conclude that the record shows that it did give adequate consideration to the use of central treatment. The Companies argue that many steel facilities are so crowded and have such limited land available that it is simply impracticable to employ separate treatment. They cite one of the plants studied by EPA’s contractor as an example of a plant where central treatment could be impracticable. There may be some individual plants with extraordinary space limitations or other problems; these plants may petition the permit granters for a variance from the nationwide standards under the variance clause of EPA’s regulations. But the companies have failed to show that the EPA’s conclusions as to central treatment are unreasonable with respect to the industry as a whole. They have not argued that EPA’s cost estimates assume central treatment, and thus make insufficient allowances for the measures which would have to be taken to institute separate treatment. Moreover, with respect to land availability, the interim final regulations state that: “[i]t is recognized that at older mills, the mill buildings may be crowded together, so the technologies suggested for ... BPCTCA minimize land requirements.” 41 Fed.Reg. 12993. 7. The Companies also claim that the EPA based its regulations on effluent flow rates which are too restrictive and unrealistic. Again, we must note that the interim final regulations do not require a particular flow rate: industry may employ any combination of concentrations and flow rates which achieves the prescribed effluent loads. Nevertheless, it is true that the EPA’s assumption about the effluent flow rates was relevant to its choice of BPCTCA, and thus to the ultimate § 301 effluent limitations as well, since the flow rate is relevant to the type of technology to be employed, and thus to the costs. The Companies’ first challenge to the EPA’s treatment of flow rates is that the Agency ignored comments to the contractor’s report and to the ANPR which indicated that the data base was unduly narrow and that the data from some of th