Full opinion text
Opinion for the Court filed PER CURIAM. PER CURIAM: In United Steelworkers of America v. Marshall, 647 F.2d 1189 (D.C.Cir.1980), cert. denied, 453 U.S. 913, 101 S.Ct. 3148, 69 L.Ed.2d 997 (1981), this court affirmed most aspects of regulations promulgated by the Occupational Safety and Health Administration (“OSHA”) governing worker exposure to airborne lead. We remanded the record to OSHA, however, for reconsideration of OSHA’s findings on technological and economic feasibility for a number of industries. This proceeding presents challenges from six of those industries to OSHA’s feasibility findings on remand. We affirm OSHA’s conclusions in all respects, except for its finding of economic feasibility for the brass and bronze ingot industry. Because the latter finding was not based on substantial evidence and was made in violation of the notice and comment provisions of the Administrative Procedure Act, we vacate that portion of OSHA’s rulemaking and remand the record to OSHA for further proceedings. I. Background A. Procedural History On November 14, 1978, OSHA, exercising its authority under section 6 of the Occupational Safety and Health Act (“OSH Act”), 29 U.S.C. § 655 (1988), promulgated comprehensive regulations designed to protect American workers from exposure to airborne lead in the workplace. See 43 Fed.Reg. 52,952, 53,007-14 (1978) (codified as amended at 29 C.F.R. § 1910.1025 (1990)). The regulations set a permissible exposure limit (“PEL”) of fifty micrograms per cubic meter of air (50 jag/m3) averaged over an eight-hour period, see 29 C.F.R. § 1910.1025(c), which employers must achieve, to the extent feasible, by a combination of engineering and work practice controls. See 29 C.F.R. § 1910.1025(e)(1). The regulations also contain a number of supporting provisions, including requirements for housekeeping, air monitoring, employee education and training, record-keeping, medical surveillance, and medical removal protection. The prior standard had imposed a limit of 200 /xg/m3 but contained no other protective requirements. See 29 C.F.R. § 1910.1000, Table Z-2 (1978). A number of industry groups, labor unions, and individual companies challenged various aspects of the new standard, leading to this court’s decision in Steelworkers. The Steelworkers court affirmed OSHA’s actions in most respects, including OSHA’s decision to mandate a PEL of 50 ¡ig/m3. See Steelworkers, 647 F.2d at 1311. However, for all but ten industries, the court concluded that OSHA had failed to carry its statutory burden of demonstrating that a 50 jug/m3 PEL was feasible to implement. For the thirty-eight industries where feasibility had not been adequately established, the court remanded the record to OSHA for further proceedings. Pending OSHA’s reconsideration, the court stayed the requirement that these industries comply with the 50 fig/m3 PEL by engineering and work practice controls alone, but mandated that they meet the PEL “by some combination of engineering, work practice, and respirator controls.” Id. In December 1981, OSHA announced that it had found the 50 /ig/m3 PEL technologically and economically feasible for all but nine of the remand industries and stated that further investigation was required to determine whether the standard was feasible for those nine. See 46 Fed.Reg. 60,758, 60,761-62 (1981). OSHA also amended section 1910.1025(e)(1) of the lead standard in several respects. First, consistent with Steelworkers, the amendments made clear that individual employers who are unable to comply with the 50 /ig/m3 PEL may escape penalty by demonstrating that, despite a finding that the standard is feasible for the industry as a whole, they have lowered exposure levels as far as they feasibly can with engineering and work practice controls. In such cases, supplementary use of respirators is permitted. Second, OSHA amended the standard to provide that, regardless of feasibility, respirators may be used with or in place of engineering and work practice controls to achieve the 50 /ig/m3 PEL if employees are exposed to levels higher than that for thirty or fewer days each year. Employers qualifying for this exemption are still required, however, to use engineering and work practice controls to reduce exposures below 200 ¡ig/m3. Finally, OSHA made clear that, in addition to the Steelworkers mandate that the remand industries achieve the 50 ¡ig/m3 with supplementary use of respirators, if needed, during the pendency of the remand proceedings, they also were required to comply with the preexisting PEL of 200 fig/m3 through use of engineering and work practice controls alone during that period. See id. at 60,-759-61, 60,775-76. At OSHA’s request, this court remanded the record again in March 1987 to determine the feasibility of the standard for the remaining industries. In July 1989, OSHA announced that it had found the standard feasible for eight of the nine industries. See 54 Fed.Reg. 29,142 (1989). For the ninth, non-ferrous foundries, OSHA concluded that although the 50 jug/m3 PEL was technologically feasible for the industry, it was not economically feasible because of the severe effects it would have on the small foundry segment of the industry. See id. at 29,245-46. OSHA then moved for another remand of the record to determine whether a PEL between 50 and 200 ¡ig/m3 would be economically feasible for the foundry industry. This court granted the motion, and in January 1990 OSHA concluded that a PEL of 75 jug/m3 was economically feasible for small foundries (those with fewer than twenty employees) and reaffirmed its earlier conclusion that a PEL of 50 ¡ig/m3 was economically feasible for large foundries (those with twenty or more employees). OSHA therefore imposed a bifurcated standard reflecting these figures on the foundry industry. See 55 Fed.Reg. 3146, 3166-67 (1990). This proceeding presents challenges to OSHA’s feasibility findings by six of the industries for which findings were made in the July 1989 and January 1990 rulemakings: leaded steelmaking, lead chemicals manufacturing, independent battery breaking, secondary copper smelting, non-ferrous foundries, and brass and bronze ingot manufacturing. B. Feasibility Determinations Under Steelworkers Section 6(b)(5) of the OSH Act, 29 U.S.C. § 655(b)(5), requires that an OSHA health standard protect workers “to the extent feasible.” Steelworkers marks our path for determining both technological and economic feasibility. To establish technological feasibility, OSHA, after consulting the “best available evidence,” must prove “a reasonable possibility that the typical firm will be able to develop and install engineering and work practice controls that can meet the PEL in most of its operations.” Steelworkers, 647 F.2d at 1272. OSHA can meet this burden by “pointing to technology that is either already in use or has been conceived and is reasonably capable of experimental refinement and distribution within the standard’s deadlines.” Id.; see also American Iron & Steel Inst. v. OSHA, 577 F.2d 825, 832-35 (3d Cir.1978). For example, if “only the most technologically advanced plants in an industry have been able to achieve [the standard] — even if only in some of their operations some of the time,” then the standard is considered feasible for the entire industry. Steelworkers, 647 F.2d at 1264. Because the OSH Act is a “technology-forcing” statute, OSHA can also “force industry to develop and diffuse new technology.” Id.; see also Society of the Plastics Indus., Inc. v. OSHA, 509 F.2d 1301, 1309 (2d Cir.), cert. denied, 421 U.S. 992, 95 S.Ct. 1998, 44 L.Ed.2d 482 (1975). In applying this standard, the Steelworkers court also noted that “[insufficient proof of technological feasibility for a few isolated operations within an industry, or even OSHA’s concession that respirators will be necessary in a few such operations, will not undermine” a showing that the standard is generally feasible. Steelworkers, 647 F.2d at 1272. OSHA must demonstrate technological feasibility with substantial evidence of all determinable facts and, for “matters having no possible basis in determinable fact, must explain the relevant considerations on which it relied and its reasons for rejecting alternate views.” Id. at 1253. Steelworkers made clear, however, that OSHA need not prove feasibility with “certainty,” nor “even ... identify the single technological means by which it expects industry to meet the PEL.” Id. at 1266. OSHA’s duty, rather, is “to show that modern technology has at least conceived some industrial strategies or devices which are likely to be capable of meeting the PEL and which the industries are generally capable of adopting.” Id. If OSHA makes reasonable predictions based on “credible sources of information” (e.g., data from existing plants and expert testimony), then the court should defer to OSHA’s feasibility determinations. See id. at 1265. Any risk that the standard may prove to be infeasible in practice is counterbalanced by flexibility in the standard’s enforcement, including the ability of firms to raise feasibility issues in enforcement proceedings. See id. at 1266, 1273; Building & Constr. Trades Dep’t v. Brock, 838 F.2d 1258, 1268 (D.C.Cir.1988). See also 54 Fed.Reg. at 29,149 (describing OSHA’s policies in enforcing the lead standard). Although the “test for feasibility cannot be lamely deferential, the possibility of reexamination of the question [in an enforcement proceeding], and the assurance that employers will be able to rely on respirators if OSHA’s predictions ... prove too sanguine, greatly ease OSHA’s preliminary burden of proving feasibility.” Steelworkers, 647 F.2d at 1273. A standard is economically feasible if the costs it imposes do not “threaten massive dislocation to, or imperil the existence of, the industry.” Id. at 1265 (internal quotation marks and citations omitted); see also Industrial Union Dep’t v. Hodgson, 499 F.2d 467, 478 (D.C.Cir.1974). To prove economic feasibility, “OSHA must construct a reasonable estimate of compliance costs and demonstrate a reasonable likelihood that these costs will not threaten the existence or competitive structure of an industry, even if it does portend disaster for some marginal firms.” Steelworkers, 647 F.2d at 1272. As with technological feasibility, OSHA is not required to prove economic feasibility with certainty, but is required to use the best available evidence and to support its conclusions with substantial evidence. See id. at 1267. C. OSHA’s General Approach 1. Technological Feasibility On remand, OSHA began its feasibility-analysis for each industry by examining evidence on existing exposure patterns and controls within the industry. Four of the industries (lead chemicals, lead pigments, non-ferrous foundries, and secondary copper smelting) provided data that came close enough to meeting OSHA’s exacting criteria to permit feasibility assessments “with considerable assurance.” See 54 Fed.Reg. at 29,145. In general (and in particular for the industries that did not provide adequate data), OSHA used its expert judgment to evaluate the data received that did not conform to OSHA’s criteria. OSHA considered evidence that exposures in one or more typical facilities had already been controlled to or below 50 ¡i.g/m3 most of the time to be the best evidence that a 50 jug/m3 PEL was technologically feasible for the industry as a whole. Where exposures were generally but not consistently below or near 50 jag/m3, OSHA concluded that modest improvements or additions to existing controls would likely reduce exposure levels consistently to or below 50 ftg/m3. See id. at 29,146; see also Steelworkers, 647 F.2d at 1264. In most cases, OSHA used a geometric mean analysis to characterize the exposure data, believing that the geometric mean is better suited than the arithmetic mean to finding an average where exposure levels cluster around one point and have only one or a few “outliers” at higher exposures. In these situations, taking an arithmetic mean would result in an average closer to the outliers, which would not be representative of the majority of exposure levels. To determine feasibility, OSHA decided it was most important to know what level was being met most of the time (the cluster of exposure levels), and it therefore concluded that the geometric mean, which is a logarithmic average and tends to fall within the cluster, would better characterize the entire exposure level distribution. See 54 Fed.Reg. at 29,147. Based on its analysis of exposure data and existing controls in place, OSHA then identified engineering and work practice controls that the industries could employ to reduce exposures to or below 50 jig/m3 in operations not yet controlled to that level. Where OSHA could demonstrate an industry’s ability to control the operations of highest exposure to a level near or below 50 jug/m3, it took this as an indication that controlling the other operations likely would follow easily. See id. at 29,146. Consistent with Steelworkers, OSHA found the standard technologically feasible if a typical employer could achieve the PEL in most of its operations most of the time. See id. at 29,148-49. OSHA claims that it did not equate an industry’s or employer’s ability to achieve a geometric mean of 50 yg/m3 with the technological feasibility of the standard, but rather treated it as one important element, along with its analyses of existing controls and practices and feasible improvements to them, in determining the standard’s technological feasibility. A geometric mean of 50 /xg/m3 or below indicated to OSHA that exposures below the PEL were being achieved a substantial portion of the time and that exposures consistently below that level could be achieved with modest additional efforts. Although OSHA did not deem this definitive proof of technological feasibility, it did consider it to be a reliable indicator of an employer’s ability to control most operations to 50 /ig/m3 most of the time, the standard set by Steelworkers. See 54 Fed.Reg. at 29,237. OSHA stated that the most important step plants could take in evaluating the feasibility of complying with the standard was to conduct a plant-wide industrial hygiene survey to identify specific sources of lead exposure, to determine the success of existing controls, and to identify what further controls were needed. OSHA found, however, that most employers in the remand industries had failed to conduct such surveys, and this contributed to OSHA’s discounting of industry assertions that their experiences demonstrated the infeasibility of meeting a 50 yg/m3 PEL. See id. at 29,147. Although the OSH Act is a technology-forcing statute, OSHA concluded with respect to all of the remand industries that conventional controls and practices already in use within at least some segments of the industries would be adequate to meet the standard. See id. at 29,144. 2. Economic Feasibility OSHA approached the question of economic feasibility by estimating the probable costs of industry compliance and comparing those costs to the industry’s financial profile in order to determine the likely effect of the costs on the prices of the industry’s product (if the costs were able to be passed through to customers) or on the viability of the industry (if the costs had to be absorbed). Economic feasibility was shown if the industry could either pass on the costs or absorb the costs without threatening the competitive structure of the industry. See 54 Fed.Reg. at 29,150. Compliance cost estimates for each industry were developed by OSHA’s consultant, Meridian Research, Inc., and were revised in response to comments. See id. OSHA then compared annualized compliance costs to industry sales figures to determine the percentage the industry would have to raise prices in order to maintain existing profits. See id. at 29,220. Where costs could not be passed on, OSHA compared annualized costs to annual profits to determine the impact on the industry of absorbing those costs. Qualitative information was also considered (e.g., evidence of modernization or new construction indicated future profitability of the industry). See id. at 29,150. D. Standard of Review On judicial review of an OSHA standard, “[t]he determinations of the Secretary [of Labor] shall be conclusive if supported by substantial evidence in the record considered as a whole.” 29 U.S.C. § 655(f). Our function is “not to decide what assumptions or findings we would make were we in the Secretary’s position,” but rather to “scrutinize the record to ensure that the Secretary has made his findings of fact on the basis of substantial evidence and has provided a reasoned explanation for his policy assumptions and conclusions.” Building & Constr. Trades Dep’t, 838 F.2d at 1266; see also Steelworkers, 647 F.2d at 1207. Deference to OSHA’s findings and policy decisions is particularly appropriate when the regulatory action involves complex technical issues: [W]e do not pretend to have the competence or the jurisdiction to resolve technical controversies in the record or, where the rule requires setting a numerical standard, to second-guess an agency decision that falls within a “zone of reasonableness[.]” Rather, our task is to “ensure public accountability” by requiring the agency to identify relevant factual evidence, to explain the logic and the policies underlying any legislative choice, to state candidly any assumptions on which it relies, and to present its reasons for rejecting significant contrary evidence and argument. Id. at 1207 (citations omitted); see also Building & Constr. Trades Dep’t, 838 F.2d at 1266 (“When called upon to review technical determinations on matters to which the agency lays claim to special expertise, the courts are at their most deferential.”). With these preliminaries out of the way, we now turn to OSHA’s findings with respect to each industry and the charges levied by the industries against them. II. Leaded Steel Industry In its initial effort to demonstrate feasibility of the lead standard, OSHA analyzed the leaded steel industry along with approximately forty other industries under one category called “other industries.” See Steelworkers, 647 F.2d at 1299-1300. Despite the uniqueness of each of the “other industries,” OSHA concluded that the lead standard was feasible for all of them “[b]ecause these industries generally have very low lead exposure, [hence], any compliance activities will require very simple engineering controls.” Id. at 1301 (quoting 43 Fed.Reg. 54,494 (1978)). The Steelwork ers court, however, rejected OSHA’s feasibility findings with respect to most of the “other industries,” including the leaded steel industry, and remanded to OSHA for reconsideration and treatment on an industry-by-industry basis. See id. OSHA has now completed its task on remand of individually analyzing each industry. The leaded steel industry has renewed its challenge to OSHA’s findings as to both the technological and economic feasibility of the 50 /xg/m3 PEL for the leaded steel industry. A. Technological Feasibility Employees in the leaded steel industry are generally exposed to airborne lead during three phases of the production process: casting or “teeming” (lead shot is added to the molten steel at very high temperatures and the molten metal is then poured or cast into molds to create the leaded steel), rolling (the leaded steel is cut into the proper dimensions and cheeked for defects), and surface conditioning (the surface of the leaded steel is refined and conditioned to create the finished product according to customer specifications). See 54 Fed.Reg. at 29,209. OSHA concluded that the 50 /xg/m3 PEL could be achieved in the leaded steel industry in all three of these operations by implementing readily available engineering and work practice controls. 1. OSHA’s Findings OSHA based its feasibility analysis on raw exposure level data submitted by LTV Corporation. See id. at 29,211; Exhibit (“Ex.”) 688A. The American Iron and Steel Institute (“AISI”) also submitted data of ranges of exposure levels industry-wide. See Ex. 694-41 at 21. OSHA discounted the AISI data because the data were submitted in an aggregated industry-wide form and did not include critical annotations correlating the exposure levels to the types of plants and operations sampled, the nature of the engineering and work practice controls in place during sampling, and the conditions of sampling that may have caused abnormally high or low exposure levels. See 54 Fed.Reg. at 29,210-11. OSHA determined that the LTV submission was the best available evidence because the samplings were carried out when leaded steel was actually being produced. See id. In addition, the data were not aggregated or averaged — they were individual sampling results of recent air lead monitoring, which enabled OSHA to make its own evaluation of the data. The data also included annotations describing the conditions of sampling. See id. at 29,210. OSHA determined that the LTV operations were typical in nature and in size of the equipment used for the industry and that its data thus were applicable to the industry as a whole. See id.; Hearing Transcript at 893-94. Based on its own analysis of LTV’s raw exposure data, OSHA concluded that the airborne lead standard of 50 /xg/m3 was technologically feasible for the entire industry because LTV was already meeting that standard most of the time in all of its operations. See 54 Fed.Reg. at 29,212. Under Steelworkers, which requires that OSHA prove “a reasonable possibility that the typical firm will be able to develop and install engineering and work practice controls that can meet the PEL in most of its operations,” OSHA has met its burden by “pointing to technology that is ... already in use” by LTV. Steelworkers, 647 F.2d at 1272. The teeming area at LTV is currently equipped with general and local ventilation systems, which capture and evacuate the volatile lead fumes. See 54 Fed.Reg. at 29,212-13. Work practice controls have also been implemented to ensure that the ventilation system is operated properly. In addition, crews are rotated and the basic teeming operation of adding lead to the molten steel is done only intermittently (only once per eight-hour shift) to minimize individual worker exposure. See id. at 29,-213. As a result of these existing engineering and work practice controls, 65% of all samples taken at LTV were below 50 jug/m3. See id. at 29,211. Based on the LTV data, OSHA determined that the standard could be achieved in the teeming process for the industry as a whole if a limited number of additional engineering and work practice controls were implemented. See id. at 29,215. For example, OSHA recommended automating the injection of lead, enclosing and ventilating the teeming area, and improving housekeeping and work practice controls. See id. at 29,214-16. In the rolling process, most of LTV’s operations are performed by remote control from inside an enclosed and ventilated booth. However, sometimes the operator must go outside the booth to maintain or clean the equipment. OSHA found that 90% of the samples taken at LTV were below 50 /xg/m3 using existing controls such as ventilation and semi-automation. See id. at 29,211. OSHA determined that the rest of the industry could meet the 50 /xg/m3 level in the rolling process by enclosing the operators’ booths, automating the milling equipment, installing local exhaust ventilation systems positioned over and alongside the milling equipment, and improving housekeeping. See id. at 29,215. In the surface conditioning process OSHA found that 90% of the LTV samples were below 50 jig/m3 using existing controls such as an enclosed, air-conditioned cab, and automation or semi-automation. See id. at 29,211. There are no controls used for manual surface conditioning, except for work practice controls such as rotating the employees. See id. at 29,210. OSHA recommended that the industry employ semi-automated conditioning equipment operated by remote control from enclosed and ventilated work stations, and local exhaust and ventilation systems for manual surface conditioning operations. In addition, OSHA recognized that supplemental respirators may be necessary for manual surface conditioning. See id. 2. Industry Challenges AISI challenges OSHA’s technological findings, asserting that OSHA did not rely on the best available data and that the controls recommended by OSHA are not technologically feasible because they either are unavailable or cannot achieve the PEL. AISI argues that the LTV data were not the best available evidence because they reflected only one producer, a producer that AISI claims is atypical for the industry because some of the samples were taken during an experimental continuous casting process, a process that results in lower exposure levels than the more widely used teeming process. However, the record makes clear that LTV’s data included at most seven out of twenty-seven samples taken from an experimental continuous casting operation. The vast majority of the LTV samples were taken when lead was added to molten steel in teeming operations typical of the industry. See id. at 29,212; Ex. 688A. AISI contends that the current industry-wide data it submitted best reflects the industry as a whole. OSHA, however, could not use AISI’s data because it was aggregated (the data was submitted in the form of ranges of exposure levels from various mills, instead of individual air lead measurements), and the data did not include any annotations indicating what operations and controls correlated with the samples. See 54 Fed.Reg. at 29,210. We find that OSHA acted reasonably in rejecting the AISI data for those reasons; the annotations were critical to analyzing the data for purposes of determining what controls would be necessary to reduce air lead exposure. We also find that it was reasonable for OSHA to consider LTV’s exposure level data as the best available evidence on the record. AISI also asserts that the controls recommended by OSHA either cannot be installed because of the nature of the production process, or have already been implemented and have not achieved the 50 ¡ig/m3 PEL. However, the record belies AISI’s claim: at least one producer, LTV Corporation, has already controlled lead levels to or below 50 jtig/m3 most of the time in all operations using existing engineering and work practice controls. See id. at 29,211-12. Under Steelworkers, OSHA can demonstrate the technological feasibility of the lead standard by showing that “only the most technologically advanced plants in an industry have been able to achieve [the standard] — even if only in some of their operations some of the time.” Steelworkers, 647 F.2d at 1264. See also American Iron & Steel Inst. v. OSHA, 577 F.2d at 832-35. We find that OSHA has met that burden here by showing that LTV, a typical plant in the industry, has met the 50 gg/m3 PEL most of the time in all of its operations with existing controls. B. Economic Feasibility 1. OSHA’s Findings OSHA determined that a PEL of 50 gg/m3 is economically as well as technologically feasible for the industry. OSHA based its economic feasibility finding on financial data for firms found in SIC 3312, which indicated that profits of approximately $900 million were expected for the industry in 1987. See 54 Fed.Reg. at 29,-218. Although the leaded steelmaking segment of the industry is experiencing difficulty, the steel industry as a whole shows signs of profitability and a willingness to withstand the losses of the leaded steel-making operations; for example, the steel industry has entered into long-term contracts for producing leaded steel and is investing in modernization of its leaded steelmaking operations. See id. at 29,218-19. OSHA estimated the costs of the additional controls necessary to achieve the PEL, and compared the costs to the overall profit and sales levels for both teeming and continuous casting producers. Annual costs to enclose and ventilate the four teeming facilities would range from $375,-604 to $622,404. See id. at 29,220. Annual costs for the one plant that uses a continuous casting process were estimated at $736,799. See id. OSHA concluded that the industry could pass through to its customers the costs of compliance because there is no general all-purpose substitute for leaded steel. See id. at 29,221. Profit impacts at the corporate level for costs that could not be passed through were estimated to be quite small and could be absorbed by the industry. See id. Impacts at the leaded steelmaking production level within the steelmaking facility (based on information for three of the five producers) were also determined to be small, although greater than the impacts estimated for the industry as a whole. See id. 2. Industry Challenges AISI asserts that OSHA should have compared the costs of compliance to the profits generated solely by the leaded steel production within each steel manufacturing facility rather than with the sales and profits generated by the facility as a whole, which includes both leaded and unleaded steel production. However, leaded steel-making operations constitute only one small part of the steel manufacturing industry; for instance, leaded steel is cast in the steel manufacturing facility no more than twice per 24-hour day. See 54 Fed. Reg. at 29,221. As AISI itself stated in its brief, “exposures [in the steel manufacturing facility resulting from leaded steel production] are intermittent because leaded steel is only a small part of the steel industry’s bar production and no facility produces leaded steel on a continuous basis.” AISI Brief at 19. Therefore, leaded steel-making is a small, yet integrated part of the steelmaking industry; it is not a separate entity. Moreover, although OSHA agreed with AISI’s assertion that the leaded steel operations operated at a loss in 1987, OSHA found that the steel manufacturers are absorbing losses for the leaded steel operations and are continuing to invest in that segment of the industry. See 54 Fed.Reg. at 29,221. OSHA determined that the costs of compliance would constitute only a small percentage of the leaded steelmaking segment’s losses, and that the steelmaking industry would absorb the compliance costs rather than shift away from leaded steel-making because of those costs. Therefore, OSHA concluded, and we agree, that leaded steelmaking is not a “stand alone” sector of the industry and that profit data from steelmaking production could be relied upon here to find that the steel industry could absorb those costs of compliance that it was unable to pass through to its customers. See id. C. Conclusion OSHA has adequately demonstrated that the 50 jug/m3 level can be met with engineering and work practice controls that are currently available and in use, and in fact the level has already been achieved in at least one plant most of the time in every operation. See Steelworkers, 647 F.2d at 1272. OSHA has also shown that the costs of compliance will not threaten the existence or competitive structure of the industry. Thus, OSHA has satisfied its burden of demonstrating both technological and economic feasibility under Steelworkers. III. Lead Chemioals Industry This court in Steelworkers found that OSHA failed to present substantial evidence to support the feasibility of the lead standard for the lead chemicals industry. See Steelworkers, 647 F.2d at 1311. On remand, OSHA was directed to perform a more individualized and detailed technological and economic feasibility analysis. OSHA performed such an analysis and concluded that a PEL of 50 fig/m3 is both technologically and economically feasible for the lead chemicals industry. See 54 Fed.Reg. at 29,186, 29,194. That industry now challenges OSHA’s findings and conclusions. A. Technological Feasibility Employees in the lead chemicals industry are exposed to airborne lead during three stages of the production process: manufacturing of lead oxide, packaging of the final product, and maintenance. Each of these three phases of the production process has unique emission control problems. In particular, the packaging of lead chemical product into specialized, or “non-bulk,” packages according to customer specifications (the product is manually measured into bags and drums), and the maintenance of production and control equipment, are processes burdened with airborne lead emissions that are difficult to control. 1. OSHA’s Findings OSHA primarily relied on raw exposure level data submitted by various producers in conjunction with site visits to the producers’ plants by OSHA’s panel of certified industrial hygiene experts. See id. at 29,-174-75. OSHA relied most heavily on Plant A’s data because the data were complete, recent, and annotated to indicate the causes of high exposure levels and the job tasks performed in certain work areas during the sampling. See id. at 29,174. OSHA determined that Plant A was a typical older plant that used conventional controls. See id. Data submitted by other producers were also extensive and recent, but they were not annotated; therefore, OSHA did not rely on such data as extensively as on the Plant A data. OSHA discounted two other data sets submitted — data submitted by the industry group, the Lead Industries Association (“LIA”), and data submitted by Plant C. OSHA disfavored the LIA data because it was aggregated industry-wide and was not annotated; it did not specify the plants, the operations within each plant, or the controls in place during the samplings. Such annotations are necessary to determine the exposure levels that are associated with particular operations and sampling conditions. Plant C’s data were rejected for the same reasons. Although OSHA discounted LIA’s data, it did consider LIA’s comments on OSHA’s data and adjusted its estimates accordingly. See id. at 29,184. During the processing of lead oxides, lead emissions are currently controlled by enclosed dust collection systems, ventilation and exhaust systems, and housekeeping. See id. at 29,178. These existing controls reduce exposure levels to 50 p.g/m3 most of the time in most operations. See id. at 29,177, 29,188. For example, at Plant A, the geometric mean for most of the production jobs was already below 50 /ig/m3 using existing controls. See id. at 29,177. Furthermore, a total of six plants in the industry have already achieved the 50 /^g/m3 standard in almost all operations, with the possible exception of the maintenance and non-bulk packaging operations. See id. OSHA found that modest improvements in housekeeping, preventative maintenance, ventilation, and enclosing open equipment would be sufficient to bring air lead levels to below 50 /xg/m3 consistently. Therefore, OSHA concluded that for most operations, the lead chemicals industry could achieve a 50 jag/m3 PEL with available engineering and work practice controls. See id. at 29,180. The most difficult operations to control are the packaging and maintenance operations. The worst exposure problem in packaging is specialized packaging of the lead chemical product in paper bags and drums according to customer specified measurements. Specialized packaging requires extensive manual intervention to adjust the weight of the package to customer specifications. OSHA determined that the PEL could be achieved in packaging if more of the product could be packaged with automated or semi-automated bulk packagers. Several plants have successfully converted entirely or mostly to bulk shipments or semi-bulk shipments. It is much easier to control exposure levels in bulk and semi-bulk packaging operations because of dust collection and exhaust ventilation systems attached to the automated equipment that fills the bulk packages. Automation can, however, be installed on smaller non-bulk packaging equipment to reduce emissions where bulk packaging is not feasible. The industry complains that the automated equipment cannot meet the rigid demands of customers for precisely measured quantities of lead chemical product. However, automation of packaging has been developed to a precision level of 0.25%-0.1%, and even further improvements are on the horizon. See id. at 29,-183; Ex. 582-90, App. C, at 3. Still, OSHA recognized that not all specialized packaging can be converted to automated bulk packaging. Therefore, for those packaging operations that cannot be converted to bulk packaging, OSHA recommended improving existing controls, such as enclosing and ventilating the existing packaging, weighing, and palletizing operations; isolating and enclosing other operations that are contamination sources, such as the cleaning of old drums; implementing rigorous housekeeping performed by a separate housekeeping staff; and improving work habits. See 54 Fed.Reg. at 29,183-84. These recommended controls are already available and OSHA has determined that the controls will reduce emissions. For example, as a result of ventilated booths alone, a 43% reduction in exposures is anticipated. See id. OSHA also conceded that supplemental respirator use may be required to meet the PEL in a limited number of specialized manual packaging operations. Another difficult control problem is presented by maintenance operations, which by their very nature involve high exposure to lead emissions. However, maintenance work is intermittent, and some of the equipment that requires constant maintenance and repair work is being phased out by the industry and replaced; for example, mechanical conveyor systems are being replaced by pneumatic systems, which have fewer parts to break and fewer joints to allow leakage of the lead chemical product. Although OSHA did find that in one plant the exposure levels for maintenance workers were below 50 ¡ig/m3 approximately 60% of the time, and the geometric mean was 22 gg/m3, see id. at 29,-187, OSHA conceded that certain maintenance tasks cannot be controlled to 50 ¡ig/m3 and those tasks may require supplementary respirators. See id. at 29,185. OSHA determined that at Plant A, a “typical” older plant with existing controls, a majority of all samples were below 50 gg/m3 and the geometric mean exposure levels in six out of eight job classifications were below 50 ;xg/m3. In addition, in six other plants, exposure levels already have been or in the foreseeable future will be controlled with existing controls to 50 /ig/m3 in most operations most of the time. See id. at 29,176-77. 2. Industry Challenges The industry challenges OSHA’s technological feasibility determination on the grounds that (1) the judgments of OSHA’s panel of three certified industrial hygiene experts, as well as the data and the site visits on which the panel based its judgments, were not the best available evidence; (2) OSHA’s recommendation that the industry convert to bulk packaging and/or automated packaging is not an “engineering and work practice control”; (3) OSHA’s finding that maintenance operators may need supplemental respirators invalidates the feasibility finding for the entire industry; and (4) OSHA’s use of the geometric mean in its feasibility analysis is unfair because OSHA does not use the geometric mean in its enforcement. We shall address each challenge in turn. First, the industry criticizes OSHA for using the “judgment’' of the panel rather than raw data; however, Steelworkers clearly permits OSHA to rely on expert judgments: reasonable predictions based on “ ‘credible sources of information,’ whether data from existing plants or expert testimony,” constitute substantial evidence in the record on which a court may uphold OSHA’s findings. See Steelworkers, 647 F.2d at 1266 (quoting AFL-CIO v. Marshall, 617 F.2d 636, 657-58 (D.C.Cir.1979)). Accordingly, we find that OSHA may rely here upon the “judgment” of “experts.” The industry also challenges OSHA’s reliance on data obtained on site visits. However, the industry did not offer any additional raw data accompanied by the annotations OSHA would need to determine exposure levels and the effect of additional controls on these exposure levels; instead, the industry offered only aggregated data without any annotations. Though this type of data was not useful to OSHA, OSHA did take into account the industry’s comments on OSHA’s own data and adjusted its analysis of that data accordingly. In any case, we reject the industry’s challenges to the suitability of OSHA’s data and find that in this case the judgment of the panel of experts, along with the data from several existing plants, constitutes substantial evidence supporting OSHA’s technological feasibility determination. The industry also claims that OSHA cannot require the industry to convert specialized product packaging to bulk packaging because it will cause them to lose the specialized packaging market to foreign competition. The industry contends that conversion to bulk packaging is a business marketing strategy dictated by the needs and demands of lead chemical consumers, not an engineering or work practice control, and therefore OSHA does not have authority under the OSH Act to require conversion. We are not persuaded by the industry’s argument. First, OSHA’s feasibility finding does not require that the industry convert entirely to bulk packaging in order to meet the lead standard, but only requires that the industry meet the standard using a combination of controls that the company selects, which may include automation of specialized packaging or improved ventilation controls on the existing specialized packaging machines. Second, even if substantial conversion proves necessary to meet the standard, the agency would be within’ its authority under the OSH Act to so require because, as this court made clear in Steelworkers, the OSH Act is a technology-forcing statute. See id. at 1230, 1264. The industry claims that OSHA’s “innovative method” of recommending conversion to bulk packaging is not an “engineering and work practice control” within the meaning of 29 C.F.R. § 1910.1025(e). Under the OSH Act and its regulations, however, “engineering controls are intended to alter industry’s machines, processes, materials, or products to reduce lead exposure at its source.” 43 Fed.Reg. at 52,989. See also Steelworkers, 647 F.2d at 1205 n. 12. We agree with OSHA that alterations in packaging and shipping methods fall within the definition of “engineering controls” and therefore are encompassed in the meaning of the Act. Although OSHA does not require conversion to bulk packaging, the agency does encourage as much automation of the packaging process as possible, either by conversion to bulk packaging or by automating the smaller packaging processes. The industry claims that full automation is infeasible because of the many different sizes of packages demanded by consumers. Yet, as noted earlier, the technology is available to automate, or at least semi-automate, packaging for a variety of smaller sizes to meet customer specifications to within 0.25%-0.1% error. See 54 Fed.Reg. at 29,183; Ex. 582-90, App. C, at 3. Again, we note that OSHA does not demand full automation: OSHA concedes that some specialized packaging operations will continue to be manual and may require supplementary respirators to meet the lead standard. Switching gears, the industry asserts that OSHA’s feasibility analysis is flawed by the finding that respirators may be necessary; however, as Steelworkers clearly states, “the standard can be feasible generally for an industry even where it is, either on the facts or by OSHA’s concession, infeasible for certain operations within that industry.” Steelworkers, 647 F.2d at 1297. Indeed, the specialized packaging operations, which require manual intervention and therefore result in high exposure levels unless supplementary respirators are used, are not “primary” operations and involve only 15% of the industry’s shipments. See 54 Fed.Reg. at 29,179. The industry also challenges OSHA’s feasibility analysis on the grounds that OSHA failed to demonstrate the feasibility of meeting the 50 gg/m3 PEL in maintenance operations without supplemental respirators. The industry claims that the need for respirators in maintenance operations invalidates the feasibility analysis for the entire industry. That assertion, however, is unfounded, as this court previously found in Steelworkers: As for maintenance, since no one could logically expect industry to always meet the PEL for workers one of whose main tasks will be to maintain and repair the control devices designed to achieve the PEL generally, we believe that OSHA need not prove that industry will not have to rely on respirators in this operation. Steelworkers, 647 F.2d at 1281 n. 138. Steelworkers made clear that the ability to control maintenance exposure levels, while important, will not invalidate the feasibility of controls for the industry as a whole. See id. at 1286. Thus, we find that OSHA’s concession that respirators may be necessary for certain maintenance operations does not invalidate the standard where, as here, most of the operations in the industry can meet the standard with engineering and work practice controls. Finally, the industry criticizes OSHA’s use of the geometric mean to establish technological feasibility. The industry challenges OSHA’s reliance on the geometric mean as unfair because the geometric mean virtually ignores outlying exposure samples, while the. “point source” method of sampling during enforcement does not ignore such samples and may result in high exposure outlier samples that are not representative of the average exposure levels. OSHA found that the geometric mean was the best statistical method to summarize the routine exposure samples. Feasibility of compliance turns on whether exposure levels at or below 50 /xg/m3 can be met in most operations most of the time; therefore, it is the routine exposure levels that determine feasibility, and atypical outliers cannot invalidate a feasibility finding. As discussed earlier, an arithmetic mean provides little insight into the distribution of exposures that are lognormally distributed. Lognormally distributed exposure data can be described as a cluster of data points with a few abnormally high or low samples. If OSHA took an arithmetic average of such data, the average would be skewed higher than the cluster of samples and therefore would not reflect the majority of the exposure levels, which fall within the cluster. Because the geometric mean falls within or close to the cluster of samples, it best reflects the majority of the exposure levels. See 54 Fed.Reg. at 29,-177. The point source samples taken during the enforcement phase most likely will fall within the cluster of routinely sampled exposure levels; therefore, the geometric mean, which best reflects the majority of exposure levels, is a good indicator of the feasibility of compliance. Additionally, OSHA’s enforcement policy takes into account the possibility that a single measurement over 50 gg/m3 can be due to random variability, and OSHA will not issue a citation on that basis alone. See id. at 29,149. Instead, OSHA will examine the employer’s historical exposure patterns to determine whether conditions on the day of inspection were representative. See id. The employer has the opportunity to demonstrate that the “one-time” OSHA sample is not representative and to establish that it has already reduced exposures to the lowest feasible limit. Several courts, including this court in Steelworkers, have upheld feasibility determinations by OSHA while recognizing that the general feasibility of the standard may still need to be counterbalanced by flexible enforcement, including variance proceedings. See Steelworkers, 647 F.2d at 1273; Society of the Plastics Indus., 509 F.2d at 1310; Building & Constr. Trades Dep't, 838 F.2d at 1268. Thus, we do not agree with the industry’s contention that the use of the geometric mean, which attempts to summarize routine exposure levels, is inherently inconsistent with the flexible enforcement policy espoused by OSHA before this court. Neither the lead standard nor OSHA’s enforcement policy ties compliance to achievement of any particular mean and Steelworkers requires nothing more than a showing of a reasonable probability that the lead standard can be met. See Steelworkers, 647 F.2d at 1272. We agree with OSHA’s determination that the geometric mean is an appropriate method of determining the reasonable probability of meeting the 50 /rg/m3 PEL in most of the industry’s operations. We find that OSHA has met that burden using the geometric mean of 50 pg/m3 as an indicator of a reasonable probability of compliance. Finally, to the extent that the industry claims that the controls recommended by OSHA will not achieve the 50 jug/m3 PEL, or cannot be installed for technological reasons, their assertion is contradicted by the record, which shows that several plants are already meeting, or are close to meeting, the 50 p.g/m3 PEL with existing controls that are readily available to the rest of the industry. See 54 Fed.Reg. at 29,183. B. Economic Feasibility The single most difficult, and most costly, operation to control in this industry is packaging. OSHA has recommended many possible controls and combinations of controls, including an increase in automated bulk packaging. However, the individual companies are not required to choose the option of conversion in order to comply with the 50 /¿g/m3 PEL; under Steelworkers, OSHA merely suggests options and leaves it to the companies to decide which ones to implement. See Steelworkers, 647 F.2d at 1266. Still, for purposes of determining economic feasibility, OSHA calculated the cost of adding at least two automated packagers to each plant as part of the compliance costs. See 54 Fed.Reg. at 29,192. 1. OSHA’s Findings OSHA based its economic feasibility analysis on data for firms in SIC 2819. See id. at 29,194. Profits were estimated using the 1986 Dun & Bradstreet rate of return on sales for SIC 2819 of 4.9%. See id. The industry’s net profit after tax was $137,700 (1985) and $84,500 (1986). OSHA calculated the costs of controls for older plants at a total incremental annual cost of approximately $200,000 (including two melting pot ventilation systems, two packaging enclosure and ventilation systems, two automated packaging systems, and improved housekeeping). See id. at 29,193. For relatively new plants, the total incremental annual cost was estimated at $81,074 (including two enclosed ventilated packaging stations, two enclosed ventilated drossing systems, and improved housekeeping). See id. And for new or modernized plants, OSHA determined that there would be no costs for the plants packaging exclusively in bulk (three out of six plants), and for the plants packaging in bags, the total annual cost was estimated at $32,066. Aggregate industry compliance costs were estimated to be $937,000. See id. OSHA determined that the lead chemicals industry would be able to absorb the costs of compliance in their profits. See id. The impact on profits for older plants would be greater than the impact for new plants; however, the rate of return on sales for both types of plants would still be positive after the compliance costs are absorbed. In sum, OSHA determined that, with an extended compliance schedule of five years to allow for modernization and increased profits, the industry would be able to absorb the costs of compliance without threatening its existence or competitive structure. See id.; Ex. 686E at 36-38. 2. Industry Challenges The industry challenges only OSHA’s finding that converting to automation or automated bulk packaging is economically feasible for the industry. The industry claims that complete conversion is not feasible because of customer demand for specialized packaging, which requires manual measurement, and because the cost of 100% conversion would be much greater than OSHA estimated and could not be absorbed by the industry. OSHA made clear, however, that it does not require complete conversion to automated bulk packaging. See 54 Fed.Reg. at 29,192. In fact, OSHA recommended a number of possible controls for those specialized packaging operations that cannot be automated or converted to bulk packaging, thereby recognizing that 100% conversion is not required for compliance with the lead standard. Furthermore, the industry itself admits that chemical buyers request specialized packaging only “on an infrequent basis” and that this specialized packaging is “relatively small in terms of the total market for lead chemicals generally.” LIA Brief at 25. Only 15% of the chemicals produced in the industry are specially packaged. See 54 Fed.Reg. at 29,179. Even assuming that conversion to automated bulk packaging is the only means of compliance, and that such conversion may cause a 15% drop in demand for the industry’s product, still there is not sufficient evidence to challenge OSHA’s determination that such a loss would not threaten the competitive structure of the industry as a whole. In the absence of such evidence, we cannot find that OSHA was incorrect in concluding, based on the best available evidence, that the lead chemicals industry can absorb the estimated costs of compliance without any threat to the industry’s existence or to its competitive structure, and therefore, that the PEL of 50 /xg/m3 is economically feasible. C. Conclusion In sum, we find that OSHA’s conclusions are reasonable and supported by substantial evidence in the record; therefore, we affirm OSHA’s findings that the 50 ¡ig/m3 PEL is technologically and economically feasible for the lead chemicals industry. IV. Independent Battery Breaking The independent battery breaking industry extracts lead from used batteries for sale to secondary smelters. Lead is recovered by cutting or crushing industrial, automobile, and other batteries and then separating the lead compounds from the other battery materials, such as sulfuric acid, cell separators, and cell casings. See 54 Fed.Reg. at 29,162. The cutting or crushing processes release lead particles into the air, and these operations present the greatest difficulties in suppressing the airborne lead. In addition, handling of the batteries before they are broken up, and of the recovered lead itself, may also give rise to airborne emissions. See id. at 29,162-63. Among the controls used by the industry to minimize exposure to lead are wetting of the materials during all phases of the process to reduce the lead particles’ tendency to become airborne; enclosure of cutting, crushing, and processing operations to limit the escape of the particles that do become airborne; and automation of all these processes to limit the presence of employees near lead sources. See id. at 29,165. Battery breaking is conducted either by integrated industrial operations (battery breaking is often integrated into secondary lead smelters plants that use the output of the battery breakers) or by independent operators; it is the challenge of the latter group which is presented here. There is little dispute that the independent battery breaking industry is in a state of severe decline: OSHA believes that, out of 250 independent battery breaking companies in existence in 1978, only two or three remained when the rulemaking under review was completed, and Meridian Research, an OSHA contractor, identified only two which remained in business at the time of its supplemental study in 1988. See id. at 29,171; Ex. 576 at 2; Ex. 686F at 1-2. These companies are Ashland Metals Company (“Ashland”) and Ace Battery (“Ace”). The precipitous decline in membership might overstate the extent of the industry’s troubles — at least some of the companies have acquired related businesses or were acquired by companies in other industries, creating integrated lead recovery operations which are no longer considered a part of the independent battery breaking industry. Still, OSHA does not contest that the industry is in a severe downturn— OSHA simply attributes the decline to factors other than the costs of compliance with the previous standard of 200 /¿g/m3. Petitioner, the Institute of Scrap Recycling Industries, Inc. (“ISRI”) (successor to the National Association of Recycling Industries, Inc.), does not challenge the technological feasibility of the 50 /ig/m3 standard for the battery breaking industry, raising instead only arguments that attack OSHA’s conclusion that the standard is, economically feasible. A. Economic Feasibility 1. OSHA’s Findings OSHA found that the independent battery breaking industry cannot pass any increase in its operating costs either forward or backward, and the industry naturally does not challenge this finding. The industry cannot increase the price to its consumers because it is competing with other sources of lead, primarily raw ore producers, as well as other sources of recovered secondary lead such as the metal scrap industries. On the supply side, more than 50% of used batteries are not recovered for recycling at current prices, indicating to OSHA that decreasing the price paid for the used batteries is also infeasible. These uneontested determinations mean that all of the increased costs of compliance will have to come out of the industry’s profits. OSHA further determined that the industry is highly volatile and that its continued existence depends on lead prices. As evidence of the industry’s unstable nature, OSHA relied on data indicating very large swings in profitability of the companies in the industry. For example, one of the remaining companies (Ace) had a return on equity of — 23.75% in the 1986 fiscal year and a return of + 56.05% in the 1987 fiscal year. See Ex. 694-1 at 14. Because no regulatory changes affecting airborne lead occurred in that period, OSHA reasoned that the wide swings in profit levels were entirely market-driven. OSHA concluded that the industry could easily accommodate the costs of complying with the standard because it found that the only two companies active in the industry, Ace and Ashland, did not substantially exceed the 50 /ig/m3 limit. In 1986, the year for which the most recent data was available, the arithmetic average exposure at Ashland was 51 jtig/m3, and the highest sampling result was 63 /ig/m3. See 54 Fed.Reg. at 29,163. Ace had an arithmetic average of 62 jug/m3. See id. at 29,164. Ace already limits airborne lead by wetting the entire breaking area in order to reduce lead’s mobility, see Ex. 694-1 at 6-7, and Ashland uses wetting along with automation of its operations and enclosure of the breaking area. See 54 Fed.Reg. at 29,165. OSHA found that additional controls could reduce the exposure levels to 50 jag/m3 and that such additional controls would require minimal costs and no technological advances because both companies are already close to the PEL of 50 /ig/m3. OSHA determined that in the best case scenario the PEL might be attained by Ace through ventilation of its industrial battery cutting operation and by Ashland through the installation of an additional water spray system. See id. at 29,171. In the worst case scenario, which OSHA considers “unlikely,” both companies would be required to enclose, automate and ventilate certain operations and to engage in additional cleaning of their facilities to lower the incidence of airborne lead. See id. Total annual costs for the Ace factory were estimated at between $4,319 and $29,885, depending on whether the best case or the worst case scenario proved accurate. This would cause a reduction in Ace’s profits of between 2.3% and 16.1%. See id. at 29,-171-72. Total annual costs for Ashland would range between $3,453 and $41,619, which would reduce Ashland’s profits by