On April 2, 1982, the Consumer Product Safety Commission (Commission) concluded a six-year investigation and rulemaking with the issuance of a final rule banning urea-formaldehyde foam insulation (UFFI) in residences and schools. Ban of Urea-Formaldehyde Foam Insulation, 47 Fed. Reg. 14,366 (1982) (to be codified at 16 C.F.R. pt. 1306). The Commission found that UFFI presents an unreasonable risk of injury from irritation and cancer and that no feasible product standard exists that would adequately protect the public from these hazards.1

Petitions seeking review of the rule were filed in this court and in the United States Court of Appeals for the District of Columbia Circuit. In an opinion dated July 2, 1982, this court was determined to be the proper venue. Consolidation here of all petitions for review was ordered. Formaldehyde Institute, Inc. v. CPSC, 681 F.2d 255 (5th Cir. 1982).

In all, four petitioners take exception to the UFFI ban. In Nos. 82-4135 and 82-4218, the Formaldehyde Institute and Gulf South Insulation (together “the industry”) challenge the ban generally. In No. 82-4136, C.P. Chemical Company argues that its product, Tripolymer 105, is safer than typical UFFIs and should have been exempted from the ban. Only Public Citizen, in No. 82 — 4311, complains that the ban was too narrow. It contends that the prohibition of UFFI should have extended to all structures, not just residences and schools,2 *1140and that the Commission exceeded its statutory authority in incorporating an exemption feature into the rule.

The record developed before the agency does not contain the substantial evidence necessary to support the Commission’s ban. Accordingly, we grant the industry’s petitions for review and vacate the Commission’s rule. Because of this decision, we do not address the issues raised by C.P. Chemical and Public Citizen.

I. THE PRODUCT

UFFI is' a thermal insulation material used in residences and other buildings. It is manufactured at the job site by mixing a liquid resin containing formaldehyde, a foaming agent, and compressed gas. The resulting liquid foam, which resembles shaving cream, is pumped into the walls of the building being insulated. After a time the mixture solidifies.

According to the industry, UFFI exhibits several characteristics that make it a particularly attractive insulation alternative. Although relatively inexpensive, it has excellent thermal resistance and energy conservation properties. Unlike some other insulation materials, it is resistant to fire.3 Perhaps most important, UFFI is well suited for installations in existing structures, termed by the industry the “retrofit” market.

The attribute of UFFI that spawned this rulemaking is its propensity to emit formaldehyde gas. Formaldehyde is a colorless gas composed of carbon, hydrogen, and oxygen.4 It is present in every cell in the human body and in the atmosphere. For--maldehyde has been in widespread commercial use for almost a century. Seven billion pounds per year are produced in the United States. According to the industry, formaldehyde is a component of products aggregating 8% of this country’s Gross National Product. Approximately half of the formaldehyde produced in the United States is used in preparing bonding resins. These resins are used in producing plywood, particleboard, fiberboard and permanent press products, in addition to UFFI. Such products as shampoo, toothpaste, cosmetics, and paper towels also contain formaldehyde.

II. THE RULEMAKING

The Commission’s involvement with UFFI began in October 1976 when the Metropolitan Denver District Attorney’s Consumer Office filed a petition under 15 U.S.C. § 2059 requesting development of a safety standard for certain home insulation products, including UFFI.5 After reviewing the available literature on the health problems linked to formaldehyde exposure, the Commission on March 5, 1979, instructed its staff to gather additional information about UFFI. 44 Fed.Reg. 12,080 (1979).

A. Formaldehyde Levels

The Commission’s initial step was to determine the average levels of formaldehyde in UFFI homes and non-UFFI homes. The agency obtained results of formaldehyde measurements in 1,164 UFFI homes. The tests showed that formaldehyde levels are highest for several months after installation. They then decrease gradually over a several-year period, eventually approaching ambient levels. The Commission determined that the average level of formaldehyde over a nine-year period after installation of UFFI is .08 parts per million. In *1141tests of 103 non-UFFI homes, an average level of .03 ppm was found.

To obtain more data, the Commission arranged tests on a number of commercially available UFFI products in simulated wall panels. The tests were conducted by the Franklin Institute Research Laboratory (Franklin Lab), the Oak Ridge National Laboratory (Oak Ridge Lab), and the Commission Engineering Laboratory. The Oak Ridge Lab concluded that the panels emitted formaldehyde at levels ranging from .03- 44 ppm sixteen months after foaming. The average level was .13 ppm.

B. Acute Irritant Effects

The Commission’s primary concern when it began investigating UFFI was the link between formaldehyde exposure and acute irritant symptoms. According to the agency, these symptoms include eye, nose, and throat irritation, persistent cough, respiratory distress, skin irritation, nausea, headaches, and dizziness.

From 1979 to 1981 the Commission investigated 350 homes whose occupants had complained of adverse health effects related to UFFI. The agency obtained information about the particular UFFI product installed, the method of installation, the surrounding environment, and the symptoms suffered by household members. According to the Commission, numerous families were forced to evacuate their homes after installation of UFFI, and in several cases acute symptoms were so severe that hospitalization was required. The agency concluded that “taken as a whole, the complaints do identify a real problem.” 47 Fed.Reg. at 14,382.

The agency commissioned the National Academy of Sciences to determine whether there is a level of formaldehyde exposure below which no acute symptoms will be experienced. After reviewing the available scientific literature, the Committee on Toxicology, a group of experts empaneled by the academy, concluded that there is no threshold for the irritant effects of formaldehyde. According to the committee, even at concentrations of formaldehyde below .25 ppm somewhat less than 20% of the population may experience some degree of irritation.

Based on this finding and its own complaint data, the Commission reached the conclusion that formaldehyde emitted from UFFI poses an unreasonable risk of acute irritant effects.

C. Carcinogenicity

While the investigation of acute effects was ongoing, the industry apprised the Commission of the preliminary results of a Chemical Industry Institute for Toxicology (Chemical Institute) study linking formaldehyde exposure at high levels to nasal cancer in rats. The agency subsequently established the Federal Panel on Formaldehyde, a group of sixteen scientists from various government agencies, to evaluate the Chemical Institute findings. The panel concluded that the Chemical Institute study was valid and that formaldehyde should be presumed to pose a carcinogenic risk to humans.

The Commission’s next step was to extrapolate from the high exposure rat data in order to quantify the risk of cancer to humans at the low levels of formaldehyde exposure associated with UFFI. To perform this task, the agency selected a computerized mathematical risk assessment model called Global 79. Global 79 was chosen over a number of other predictive models because it incorporated several assumptions the Commission believed were applicable to formaldehyde carcinogenesis. Unlike some other models, Global 79 does not predict an actual or most likely risk. Rather it predicts a range of risk within which there is a 95% possibility the actual risk will fall. Based on the actual formaldehyde levels found in the tests of 1,164 homes, the model predicted that the increased risk of cancer to a person living in a UFFI home for 9 years, 16 hours a day, would range from 0 to 51 in 1,000,000. From the results of the Franklin and Oak Ridge Labs studies, Glob*1142al 79 predicted a risk range of between 0 and 37 in 1,000,000.6

From the Global 79 analysis of the Chemical Institute rat study, the Commission concluded that UFFI poses an unreasonable risk of cancer to humans. This determination, coupled with the agency’s finding of an unreasonable risk of acute irritant effects, is the reason for the ban. The Commission acknowledges that unless substantial evidence supports both its carcinogenicity and acute effects findings the ban was not proper.

The industry challenges both of these findings and the procedures employed by the Commission in reaching them. Before addressing these contentions in detail, we outline the standard by which we review rules promulgated under the Consumer Product Safety Act (Act).

III. STANDARD OF REVIEW

The Act contains detailed procedures the Commission must follow in issuing consumer product safety rules. Numerous findings are mandated. See 15 U.S.C. § 2058(f)(1) and (f)(3). Under section 2058(f)(3)(A), the Commission is expressly prohibited from promulgating a safety rule unless it finds that the product that will be subject to the rule poses an unreasonable risk of injury. The central focus of our analysis is on the Commission’s finding that UFFI poses such a risk.

Jurisdiction to review rules issued under the Act is vested in the courts of appeals. 15 U.S.C. § 2060(a).7 Section 2060(c) provides that a consumer product safety rule shall not be affirmed “unless the Commission’s findings ... are supported by substantial evidence on the record taken as a whole.” Our specific inquiry, then, is whether the Commission’s finding that UFFI poses an unreasonable risk of injury is supported by substantial evidence.

The leading decision in this circuit on the meaning of the substantial evidence test in the context of this Act is Aqua Slide ‘N’ Dive v. CPSC, 569 F.2d 831 (5th Cir. 1978), which involved the first product standard the Commission promulgated. There we found that “Congress put the substantial evidence test in the statute because it wanted the courts to scrutinize the Commission’s actions more closely than an ‘arbitrary and capricious’ standard would allow.” Id. at 837 (citing H.R.Rep. No. 1153, 92d Cong., 2d Sess. 38 (1972); S.Rep. No. 749, 92d Cong., 2d Sess. 35 (1972); -118 Cong.Rec. 31378 (1972) (remarks of Cong. Moss)).8 The facts that detract from the agency as well as those that support it are *1143to be considered. Id. at 838 (citing Universal Camera v. NLRB, 340 U.S. 474, 478, 71 S.Ct. 456, 459, 95 L.Ed. 456 (1951)). “The ultimate question is whether the record contains ‘such relevant evidence as a reasonable mind might accept as adequate to support a conclusion.’ ” Id. (quoting Consolidated Edison Co. v. NLRB, 305 U.S. 197, 229, 59 S.Ct. 206, 216, 83 L.Ed. 126 (1938)). If it does, the Commission has sustained its burden of adducing “substantial evidence on the record as a whole” and the rule must be affirmed. Id. If not, the rule must be vacated.

In attacking the UFFI ban, the industry and amici9 leave very few stones unturned. They take issue with numerous aspects of the Commission’s findings with respect to acute irritant effects. The bigger salvo, however, is aimed at the finding that UFFI poses an unreasonable risk of injury from cancer. We therefore address the Commission’s cancer risk assessment first.

IV. THE RISK OF CANCER FROM UFFI

As we have noted, the Commission incorporated the results of the Chemical Institute high-exposure rat study into its Global 79 computerized risk assessment model in order to predict the cancer risk to humans at the low levels of formaldehyde exposure associated with UFFI. Based on actual formaldehyde levels found in 1,164 homes, the Commission found that “[a]ny individual living in a home with U.F. foam insulation for 9 years after installation would, according to the assessment, have an additional risk of 51 in a million of developing cancer from the formaldehyde released by the insulation.” 47 Fed.Reg. at 14,372. The Franklin/Oak Ridge Lab studies predicted an additional risk of 37 in a million.

According to the industry, these predictions are flawed for a number of reasons. The industry argues that: (1) neither the formaldehyde levels found in the 1,164 test homes nor the Franklin/Oak Ridge Labs test results are accurate indicators of the formaldehyde levels in average UFFI homes; (2) the Commission erred in relying exclusively on the Chemical Institute rat data in its risk assessment model and ignored numerous epidemiologic studies indicating that formaldehyde is not a human carcinogen; (3) the Commission ignored the real explanation for the incidence of tumors at the high levels of formaldehyde exposure involved in the Chemical Institute study; (4) no substantial evidence supports the Commission’s assumption that the effective formaldehyde dose for humans is the same as that for rats; (5) Global 79 incorporates several assumptions about formaldehyde carcinogenicity that are not supported by substantial evidence; (6) Global 79 predicts only an upper limit of risk and does not constitute substantial evidence that “it is at least more likely than not that [UFFI] presents a significant risk of [cancer],” Industrial Union Dept. AFL-CIO v. American Petroleum Inst., 448 U.S. 607, 653, 100 S.Ct. 2844, 2869, 65 L.Ed.2d 1010 (1980); and (7) other federal agencies have determined that formaldehyde does not pose a substantial health risk to man. We need examine only the first two contentions in detail.

A. Formaldehyde Levels

At the outset, the industry asserts that the average formaldehyde level attributed by the Commission to UFFI homes is inaccurate and grossly exaggerates the true level. In fact, the industry contends that there is no significant difference between formaldehyde levels in UFFI homes and non-UFFI homes.

According to the industry, the in-home measurements, which resulted in a predicted additional risk of cancer of 51 in a million, were scientifically invalid and do not constitute substantial evidence. The industry points out that the great majority of the tests were conducted by state agencies, not *1144Commission personnel. Measurement methodologies differed from state to state. Twenty percent of the test homes were located in Massachusetts and Minnesota, where measurement techniques of questionable accuracy were used. More important, the test homes were not randomly chosen. A large percentage of the measurements taken were in residences whose occupants had complained about UFFI-related health problems.10 According to the industry, this tended to overstate the actual average level. Additionally, the high formaldehyde levels that were recorded in many of the “complaint” homes were the result of faulty installation.11 These problems could be eliminated by a product standard. The industry concludes that the Commission should have conducted its own tests of randomly selected UFFI homes before taking the draconian step of a product ban.

The industry also asserts that the Franklin/Oak Ridge Labs test data do not accurately predict formaldehyde levels in actual homes. It notes that only nine panels were tested and that the results varied widely around the mean of .13 ppm.12 The industry also argues that storage of the panels in extreme conditions invalidated the test results.13 Additionally, the industry contends that the test did not account for the air flow and exchange that occurs in actual homes and that Oak Ridge Lab recognized that this failure could lead to errors of two-to threefold. According to the industry, even the Commission recognized that in-home measurements could not be extrapolated from the Franklin/Oak Ridge Labs data.

The industry bases its claim that formaldehyde levels in UFFI homes do not differ significantly from levels in non-UFFI homes on three studies. A University of Iowa study conducted by Dr. Frank found no difference in formaldehyde levels in randomly selected UFFI and non-UFFI homes.14 Studies conducted by the Canadian government and by Dr. Firstman of the Georgia Institute of Technology detected only negligible differences.15 The industry also relies on Borden v. Commissioner of Pub. Health, No. 38,478, slip. op. (Mass.Super.Ct. Jan. 19, 1982), which overturned as arbitrary and capricious a state ban of UFFI. The Borden court specifically found that “there has been no showing that the ambient level of formaldehyde concentration in houses in which UFFI has been properly installed is significantly more appreciable or different than the level of formaldehyde in houses without UFFI.” Slip op. at 20.

The Commission contends that the Frank, Firstman, and Canadian studies were invalid. None of the three considered the length of time that had elapsed since installation-of the UFFI, a vital factor in the level of formaldehyde emitted. Therefore, they do not support the claim that UFFI adds only trace amounts of formaldehyde to the home environment.

The agency also defends its use of Franklin/Oak Ridge Labs data. It argues that the conditions to which the panels were exposed before testing were no worse than would exist in a home without heating or air-conditioning. The Commission does not address other concerns raised by the indus*1145try about the Franklin/Oak Ridge Labs studies, but nevertheless concludes that the study was valid because its results were similar to those obtained in the in-home tests. Two significant omissions with respect to the in-home tests cloud this conclusion. The Commission does not explain its reliance on a data base comprised largely of complaint houses. Nor does the agency justify its failure to conduct a study of randomly selected UFFI homes before issuing the product ban.

We do not reach the conclusion the Borden court did. The studies relied on by the industry do not demonstrate conclusively that formaldehyde levels in UFFI and non-UFFI homes are essentially the same. As the Commission notes, age is a vital factor in the level of formaldehyde emitted by UFFI. The in-home and Franklin/Oak Ridge Labs studies do suggest that UFFI appreciably raises in-home formaldehyde levels.

But the Commission did not use the studies only to support such a generalized finding. They were also incorporated into an exacting, precise, and extremely complicated risk assessment model. The goal of the model was to determine the risk of cancer to a consumer living in an average UFFI home. The difficulty in reaching this goal is that neither the in-home nor the Franklin/Oak Ridge Labs studies were consistent with this aim. The in-home study focused on complaint residences, not average residences, not randomly selected residences. The Franklin/Oak Ridge Labs studies reflected conditions similar to an unheated, unair-conditioned home, not an average home. The similar results achieved by the two studies validate neither. The studies were inadequate to serve as a data base for the Global 79 risk assessment.

B. Evidence of Carcinogenicity: Epidemi-ologic Data and the Chemical Institute Study

The industry next contends that the numerous studies of humans exposed to formaldehyde in the workplace discredit the Commission’s finding that formaldehyde is carcinogenic at low levels. Eleven epidemiologic studies involving a total of 10,000 workers were introduced into the record. None of the studies’ authors found a statistically significant increase in the number of cancers among workers exposed to formaldehyde compared to the general population. The largest studies were conducted by Dr. Marsh, who found no nasal cancer and no dose-response relationship between formaldehyde exposure and other respiratory cancers among 2,490 formaldehyde workers; Drs. Walrath and Fraumeni of the National Cancer Institute, who found no nasal cancer mortality and no unusual respiratory cancer mortality in their study of 1,106 morticians; and Dr. Wong of Tabershaw Occupational Medical Associates, who found no nasal cancer mortality or excess respiratory cancer mortality among 2,026 formaldehyde workers. Dr. Wong concluded that “at this point there is no epidemiologic evidence that formaldehyde is a human carcinogen.”

The industry also relies on studies that were not completed until after the close of the record. The Federal Panel on Formaldehyde, which was convened by the Commission to assess the carcinogenicity of formaldehyde, stated in its recommendations to the agency that three pending studies of embalmers would “provide a good estimate of potential risk from formaldehyde exposure.” 16 Results in two of the three have now been released.17 In neither were nasal cancers or excess respiratory cancers found.

Finally, the industry points to statements made by Dr. Higginson, former director of the International Agency for Research on Cancer, in a letter submitted to the Commission shortly before issuance of the ban. Dr. Higginson stated that the epidemiologic *1146data “are insufficient to exclude a minimal risk, but certainly they weigh heavily against the view that formaldehyde constitutes any considerable risk for nasal cancer to man.” He also wrote that “[e]xact estimates as to the number of cases of a cancer that might be expected to occur in man based on a single experiment [the Chemical Institute rat study] are silly and simply ignore biological realities.”

The Commission defends its reliance on the Chemical Institute rat data by reminding us that the Federal Panel found the study valid and concluded that formaldehyde should be presumed to pose a carcinogenic risk to humans. The agency rejects the industry’s contention that the available epidemiologic evidence demonstrates that formaldehyde is safe. The Commission points to the small number of workers involved in the epidemiologic studies and their failure to account for the duration and concentration of formaldehyde exposure.

Again, the truth appears to lie somewhere between the positions taken by the Commission and the industry. We agree with the agency that the epidemiologic studies cited by the industry do not demonstrate conclusively that formaldehyde poses no cancer risk to man. The Commission concluded that the increased risk of cancer from formaldehyde exposure (at the levels it attributed to UFFI) was up to approximately 1 in 20,000 (51 in a million). It is highly unlikely that studies involving a total of 10,000 workers would detect such a small risk. Additionally, the failure to consider either the length of time the workers were exposed to formaldehyde or the levels to which they were exposed diminishes the studies’ usefulness.

While the Commission correctly notes that the epidemiologic evidence is not conclusive, its exclusive reliance on the Chemical Institute study in its Global 79 risk assessment is equally unsupportable. In the study 240 rats were exposed to an average of 14.3 ppm formaldehyde for six hours a day, five days a week.18 After 24 months 103 of the rats developed nasal carcinomas. This was the only empirical datum with respect to formaldehyde carcinogenicity that was incorporated into the Global 79 model. But in a study as small as this one the margin of error is inherently large. For example, had 20 fewer rats, or 20 more, developed carcinomas, the risk predicted by Global 79 would be altered drastically.

The element of doubt present here is similar to that with respect to formaldehyde levels. The Federal Panel’s findings that the Chemical Institute study was valid and that formaldehyde should be presumed to pose a cancer risk to man do not authenticate the use of the study’s results, and only those results, to predict exactly the cancer risk UFFI poses to man. As Dr. Higginson aptly stated, it is not good science to rely on a single experiment, particularly one involving only 240 subjects, to make precise estimates of cancer risk.

This problem is exacerbated by concerns about the Chemical Institute study raised by the industry. Although the average level of formaldehyde exposure in the experiment was 14.3 ppm, the rats in fact were exposed regularly to much higher doses. Measurements of between 17 and 20 ppm were not uncommon. The highest recorded level was a near-lethal 32.4 ppm. We do not have to agree with the industry that this disparity renders the study invalid for all purposes to conclude that the Commission could not properly use the study as it did. To make precise estimates, precise data are required.

C. Other Contentions

The industry attacks the Commission’s cancer findings on several other grounds. Most of these relate to assumptions made *1147by the Commission in its risk assessment. Although several of these contentions present substantial questions,19 we do not address them. The predictions made by the risk assessment model are no better than the data base. We have concluded that this base was inadequate. The Commission improperly relied on in-home data gathered largely from complaint homes. It failed to conduct a controlled study of randomly selected residences. The result is that the Commission’s finding that UFFI poses an unreasonable risk of cancer is not supported by substantial evidence on the record as a whole.20

V. ACUTE IRRITANT EFFECTS

The Commission relied on its findings with respect to both cancer and acuté' irritant effects in promulgating the UFFI ban. With one of the two supports gone, the ban cannot stand.21 Because of this our brief discussion of the agency’s finding that UFFI poses an unreasonable risk of acute irritant effects is intended only to indicate that we would also find this conclusion to be unsupported by substantial evidence.

From its in-depth investigations, the Commission concluded that inhabitants of UFFI homes suffer eye, nose, and throat irritation, persistent cough, respiratory distress, skin irritation, nausea, headaches, and dizziness. The industry asserts that “in-depth” is a misleading misnomer. It points out that very few of the reported symptoms, which resemble those of influenza and colds, were diagnosed by physicians to be related to formaldehyde. Additionally, the industry notes that formaldehyde measurements were taken in very few homes. The industry takes the position that the complaint data are worthless and do not establish a causal relationship between UFFI and the symptoms reported. The Commission responds to this assessment with the following facts: Symptoms (1) were reported by more than one family member in many homes; (2) normally manifested themselves immediately after installation and often persisted for months or even years; (3) frequently dissipated outside the home; and (4) in several cases were verified to be linked to formaldehyde by physicians.

Certainly the investigations do not establish that all the reported symptoms were caused by formaldehyde emitted by UFFI. But the Commission’s defense of its investigations persuades us that UFFI is not completely innocent. We agree that “taken as a whole, the complaints do identify a real problem.” 47 Fed.Reg. at 14,382.

*1148The problem is the complaints do no more. Specifically, they do not answer the question whether the risk of injury from acute irritant effects is unreasonable. This inquiry involves “a balancing test like that familiar in tort law: The regulation may issue if the severity of the injury that may result from the product, factored by the likelihood of the injury, offsets the harm the regulation imposes upon manufacturers and consumers.” Southland Mower v. CPSC, 619 F.2d 499, 508-09 (5th Cir. 1980) (quoting Aqua Slide ‘N’ Dive, 569 F.2d at 839). See also H.R.Rep. No. 1153, 92d Cong., 2d Sess. 33 (1972). What the complaints fail to demonstrate is how likely it is that acute symptoms will occur.

To fill this gap in the equation, the Commission points to the conclusion reached by the National Academy of Sciences’ Committee on Toxicology. That committee found that somewhat less than 20% of healthy adults may respond to the irritant effects of formaldehyde at .25 ppm. For the Commission’s purposes and ours, this finding is worth exactly nothing. First, it does not state whether the response at .25 ppm is expected to be slight or severe. All record evidence indicates the former.22 Second, with the exception of the first few months after installation, .25 ppm is much higher than the formaldehyde level associated with UFFI. The committee did not address the risk at the appropriate level. Third, the agency states that the committee predicted that “somewhat less than 20%” of the population may respond at .25%. This, like the Commission’s cancer prediction of “up to 51 in a million,” provides us no basis for review under the substantial evidence standard.

We conclude that the Commission’s finding that UFFI poses an unreasonable risk of injury from acute irritant effects is not supported by substantial evidence. The failure to quantify the risk at the exposure levels actually associated with UFFI is the finding’s Achilles heel. Predicting how likely an injury is to occur, at least in general terms, is essential to a determination of whether the risk of that injury is unreasonable.

VI. THE CONSUMER PRODUCT SAFETY ACT vs. THE FEDERAL HAZARDOUS SUBSTANCES ACT

One procedural issue is tangential to our present review of the rule. It may nevertheless become important if the Commission wishes to take action with respect to UFFI in the future. The industry contends that the Consumer Product Safety Act is an improper vehicle for this rulemaking; that the rule should have been promulgated pursuant to the Federal Hazardous Substances Act, 15 U.S.C. §§ 1261-1276. The rulemak-ing procedures mandated by the two acts differ substantially. The Consumer Product Safety Act provides for an informal rulemaking.23 In contrast, the Federal Hazardous Substances Act requires a formal hearing, complete with rules of evidence and the right to confront and cross-examine witnesses. See 15 U.S.C. § 1262(a)(2). The Commission’s failure to conduct such a hearing is the substantive reason for the industry’s procedural objection.

Congress has mandated which act’s procedures the Commission must follow. In 15 U.S.C. § 2079(d), the Consumer Product Safety Act provides:

A risk of injury which is associated with a consumer product and which could be eliminated or reduced to a sufficient extent by action under the Federal Hazardous Substances Act ... may be regulated under this [Act] only if the Commission by rule finds that it is in the public interest to regulate such risk of injury under this [Act].

The Commission thus can proceed under the less stringent procedural provisions of the *1149Consumer Product Safety Act if it satisfies one of two tests. It must show either that the risk could not be regulated sufficiently under the Federal Hazardous Substances Act or that it is in the public interest to proceed under the Consumer Product Safety Act rather than the Federal Hazardous Substances Act. The Commission asserts that it has met both tests.

In support of its fiiiding that UFFI could not be regulated adequately under the Federal Hazardous Substances Act, the agency points out that that Act is limited in scope to products for use in the household. 15 U.S.C. § 126(q)(l)(B). See Barnes v. Litton Indus. Prod., Inc., 555 F.2d 1184 (4th Cir. 1977). Unlike the final rule here, a regulation promulgated under the Federal Hazardous Substances Act could not have banned UFFI in schools. Additionally, the Commission notes that its original proposed rule extended to all buildings, not just homes and schools. 464 Fed.Reg. 11,188 (1981). The Commission argues that the intended scope of the rule when the proceeding was initiated should be determinative. Because the proposed rule was unlimited, the agency asserts that the Federal Hazardous Substances Act was wholly inadequate and its decision to proceed under the Consumer Product Safety Act cannot be challenged.

We reject the Commission’s reliance on the need to regulate UFFI in schools to justify proceeding under the Consumer Product Safety Act. The record is totally devoid of evidence to support extension of the ban to schools. From the beginning the Commission focused its attention exclusively on homes. The cancer risk assessment used the formaldehyde levels found in the in-home tests24 to predict the risk to a person present in a UFFI home for sixteen hours a day, seven days a week. No academic institution was shown to have a curriculum requiring such constant attendance. The in-depth investigations of acute effects also ignored schools. Only one complaint was received regarding a school. It was not investigated. We cannot understand why the Commission decided to extend the ban to schools unless it was to justify its decision to proceed under the Consumer Product Safety Act. Because the ban in schools was completely unsubstantiated, we find that it was insufficient for this purpose.

The agency’s argument that a Consumer Product Safety Act rulemaking was justified because the proposed ban was unlimited is also unpersuasive. The initial selection of procedures must be based on assumptions made at the outset of the rule-making. But if it immediately becomes apparent, as it did here, that other procedures should control, those procedures must be used. The record clearly demonstrates that the Commission never made any real attempt to address the risk from UFFI in buildings other than homes. This belies the major premise underlying the initial decision to proceed under the Consumer Product Safety Act, a decision that denied the industry important procedural rights guaranteed by the Federal Hazardous Substances Act. In our view the differences between the acts demand an ongoing evaluation of the facts as they develop. We need not decide here the extreme case where last minute proof indicates a change of procedures. Because the record makes clear that from the outset the Commission’s entire effort was channelled to investigating the risk of UFFI in homes, we reject the finding that the risk could not have been regulated sufficiently under the Federal Hazardous Substances Act.

The Commission’s finding that it was in the public interest to regulate UFFI under the Consumer Product Safety Act was based solely on a desire to avoid “the complex and lengthy nature of the rulemaking proceeding that would be required under the FHSA ....” 47 Fed.Reg. at 14,369. We reject this reasoning. If the due proc*1150ess procedures mandated by the Federal Hazardous Substances Act justify discarding it, the Commission would never have to use it. Congress has provided otherwise. Rulemaking under the Consumer Product Safety Act is to be the exception, not the rule. We do not say that it could never be in the public interest to regulate an extremely dangerous product under the Consumer Product Safety Act for the sake of speed and efficiency. We say only that UFFI is not such a product.

The Commission’s decision to continue its rulemaking under the Consumer Product Safety Act was improper. The selection of procedures is too important to be based on unexplored theories and desires for administrative convenience. Any future regulatory effort directed at UFFI must be made pursuant to the Federal Hazardous Substances Act unless the Commission substantiates its decision to proceed under the Consumer Product Safety Act with the findings required by section 2079(d).

VII. CONCLUSION

We are not unmindful that regulating in the face of scientific uncertainty within ever-tightening budgetary constraints presents the Commission with a difficult task. We nevertheless cannot abdicate our role in the regulatory process. Congress and our circuit’s precedents require us.to take a “harder look”25 to determine whether rules adopted under the Consumer Product Safety Act are supported by substantial evidence. That look discloses that the evidence is lacking here. The Commission’s rule banning UFFI is

VACATED.