The Environmental Protection Agency (EPA) seeks to recover its response costs under the Comprehensive Environmental Response, Compensation and Liability Act (CERCLA) because of a discharge of chromium waste that contaminated a local water supply. Sequa Corporation appeals from the imposition of joint and several liability, challenges the EPA's decision to provide an alternate water supply system to the area in which the groundwater was contaminated by the chromium discharge, and contests the calculation of prejudgment interest and the application of the proceeds of the EPA's settlement with its co-defendants. We REVERSE the portion of the judgment imposing joint and several liability, and REMAND for further proceedings. Our review of the administrative record has convinced us that the EPA's decision to provide an alternate water supply was arbitrary and capricious; accordingly, we REVERSE the portion of the district court's judgment allowing the EPA to recover the costs of designing and constructing that system, and REMAND for deletion of those amounts and recalculating prejudgment interest.1

* In 1978, a citizen in the Odessa, Texas area complained about discolored drinking water. The Texas Water Commission conducted an investigation. It ultimately focused on a chrome-plating shop that was operated successively from 1971 through 1977 by John Leigh, Western Pollution Control Corporation (hereinafter referred to as Bell), and Woolley Tool Division of Chromalloy American Corporation (which later merged with Sequa), at 4318 Brazos Street, just outside the city limits of Odessa. The investigation showed that during the chrome-plating process, finished parts were rinsed, and the rinse water was pumped out of the building onto the ground.

In 1984, the EPA designated a 24-block area north of the Brazos Street facility as a Superfund site--"Odessa Chromium I." It authorized a response action pursuant to its authority under CERCLA Sec. 104, 42 U.S.C. Sec. 9604, and entered into a cooperative agreement with the State of Texas. The State was to perform a remedial investigation, feasibility study, and remedial design work for the site, with the EPA reimbursing the State for ninety percent of the costs. The remedial investigation revealed that the Trinity Aquifer, the only source of groundwater in the area, contained elevated concentrations of chromium.2

A "focused" feasibility study (FFS) was undertaken to evaluate the need to provide an alternative water supply pending completion of the remaining portion of the feasibility study and implementation of final remedial action.3 The FFS concluded that the City of Odessa's water system should be extended to provide service in the Odessa Chromium I area. On September 8, 1986, the EPA Regional Administrator issued a Record of Decision (ROD), finding that city water service should be extended to the site. Pursuant to the cooperative agreement, the State, through its contractor, designed and constructed the system, which was completed in 1988.

In December 1988, the EPA filed a CERCLA cost-recovery action against Bell, Sequa, and John Leigh, which was consolidated with an adversary proceeding the EPA had filed against Bell in Bell's bankruptcy case. The EPA sought to recover direct and indirect costs it incurred in studying, designing, and constructing the alternate water supply system.

In July 1989, the district court entered a case management order providing that the case would be decided in three phases: Phase I--liability, Phase II--recoverability of the EPA's response costs, and Phase III--"responsibility." In September 1989, the district court granted in part, and denied in part, the EPA's motion for summary judgment as to liability. In its memorandum opinion, it stated that the relative culpability of the parties and the "divisibility of liability" issues would be decided during Phase III. Although the district court ruled that CERCLA did not require the EPA to prove causation, it held an evidentiary hearing and made alternative findings and conclusions addressing causation, holding that "Leigh, Bell and Sequa caused the contamination."4 In March 1990, the district court granted the EPA's motion for clarification of the September 1989 summary judgment, holding that its previous opinion had provided that the defendants were jointly and severally liable. It also entered a declaratory judgment as to the defendants' liability for future response costs.

The Phase II proceeding on recoverability of response costs was handled through cross-motions for summary judgment. 734 F. Supp. 771. The district court held that the defendants had not met their burden of proving that the EPA's decision to implement an alternate water supply was arbitrary and capricious, and held that they were liable for the EPA's direct and indirect response costs, plus prejudgment interest from the date such costs were incurred.

On March 2, 1990, the EPA sought approval of a proposed consent decree, in which it settled its claims against Bell for all costs, past and future, for $1,000,000. Sequa objected to the settlement, contending that Bell was not being required to pay its fair share. The district court granted Sequa's request for a hearing on the fairness of the proposed consent decree, and entered an order providing that a Phase III hearing regarding apportionment of liability was to be conducted before it ruled on the motion for entry of the consent decree. In response to the EPA's motion for clarification of the scope of the hearing, the court ruled that the hearing would be limited to determining the relative contributions of Bell, Sequa, and Leigh to the contamination. After the Phase III hearing in June 1990, Sequa filed a motion for reconsideration on the issue of joint and several liability. On July 24, the district court denied that motion, and approved the consent decree. It held that the evidence at the Phase I and Phase III hearings demonstrated that there was no method of dividing the liability among the defendants which would rise to any level above mere speculation, because each of the proposed apportionment methods involved a significant assumption factor, inasmuch as records had been lost, and because each of the apportionment methods differed significantly. In the alternative, it concluded that, based on equitable factors, responsibility should be divided as follows: Bell--35%; Sequa--35%; and Leigh--30%.

In December 1990, the district court entered an order approving another consent decree, pursuant to which the EPA settled its claims against Leigh for past and future costs--for $100,000.

In sum, the district court held that Sequa is jointly and severally liable for $1,866,904.19, including the costs of studying, designing, and constructing the alternate water supply system. In addition, Sequa is jointly and severally liable for all future costs incurred by the EPA in studying, designing, and implementing a permanent remedy.5

CERCLA was enacted in 1980, and amended in 1986 by the Superfund Amendments and Reauthorization Act (SARA). Its purpose is to facilitate the prompt clean-up of hazardous waste sites. See, e.g., United States v. R.W. Meyer, Inc., 889 F.2d 1497, 1500 (6th Cir.1989). CERCLA Sec. 104, 42 U.S.C. Sec. 9604, authorizes the President (who has delegated most of his authority under CERCLA to the EPA) to use Superfund money to respond to any threatened or actual release of any hazardous substance that may pose an imminent and substantial public health threat. CERCLA Sec. 107, 42 U.S.C. Sec. 9607, provides for the recovery of response costs from all persons responsible for the release of a hazardous substance. Response actions include both "remedial" and "removal" actions. Removal actions generally are immediate or interim responses, and remedial actions generally are permanent responses. See Voluntary Purchasing, Inc. v. Reilly, 889 F.2d 1380, 1382 n. 4 (5th Cir.1989).

The National Contingency Plan ("NCP"), 40 C.F.R. Part 300, promulgated by the EPA as mandated by CERCLA Sec. 105, 42 U.S.C. Sec. 9605, guides federal and state response activities. The NCP identifies methods for investigating the environmental and health problems resulting from a release or threatened release and criteria for determining the appropriate extent of response activities. See Daigle v. Shell Oil Co., 972 F.2d 1527, 1533 n. 4 (10th Cir.1992); United States v. R.W. Meyer, Inc., 889 F.2d at 1500.

Since CERCLA's enactment, the federal courts have struggled to resolve the complicated, often confusing, questions posed by the concept of joint and several liability, and its application under a statute whose provisions are silent with respect to the scope of liability, but whose legislative history is clear that common law principles of joint and several liability may affect liability.6 The issue is one of first impression in this Circuit.

* Common Law: The Restatement of Torts

Although joint and several liability is commonly imposed in CERCLA cases,7 it is not mandatory in all such cases. United States v. Monsanto Co., 858 F.2d at 171. Instead, Congress intended that the federal courts determine the scope of liability in CERCLA cases under traditional and evolving common law principles, guided by the Restatement (Second) of Torts. Alcan-Butler, 964 F.2d at 268; O'Neil v. Picillo, 883 F.2d at 178; Allied Corp. v. Acme Solvents Reclaiming, Inc., 691 F. Supp. 1100, 1116 (N.D.Ill.1988); Chem-Dyne, 572 F. Supp. at 810.

Section 433 of the Restatement provides that:

(1) Damages for harm are to be apportioned among two or more causes where

(a) there are distinct harms, or

(b) there is a reasonable basis for determining the contribution of each cause to a single harm.

(2) Damages for any other harm cannot be apportioned among two or more causes.

Restatement (Second) of Torts, Sec. 433A.

The nature of the harm is the key factor in determining whether apportionment is appropriate. Distinct harms--e.g., where two defendants independently shoot the plaintiff at the same time, one wounding him in the arm and the other wounding him in the leg--are regarded as separate injuries. Although some of the elements of damages (such as lost wages or pain and suffering) may be difficult to apportion, "it is still possible, as a logical, reasonable, and practical matter, ... to make a rough estimate which will fairly apportion such subsidiary elements of damages." Id., comment b on subsection (1).

The Restatement also discusses "successive" harms, such as when "two defendants, independently operating the same plant, pollute a stream over successive periods of time." Id., comment c on subsection (1). Apportionment is appropriate, because "it is clear that each has caused a separate amount of harm, limited in time, and that neither has any responsibility for the harm caused by the other." Id.

The final situation discussed by the Restatement in which apportionment is available involves a single harm that is "divisible"--perhaps the most difficult type of harm to conceptualize. Such harm, "while not so clearly marked out as severable into distinct parts, [is] still capable of division upon a reasonable and rational basis, and of fair apportionment among the causes responsible.... Where such apportionment can be made without injustice to any of the parties, the court may require it to be made." Id., comment d on subsection (1). Two examples of such harm are described in the comment. The first is where cattle owned by two or more persons trespass upon the plaintiff's land and destroy his crops. Although "the aggregate harm is a lost crop, ... it may nevertheless be apportioned among the owners of the cattle, on the basis of the number owned by each, and the reasonable assumption that the respective harm done is proportionate to that number." Id. The second example involves pollution of a stream by two or more factories. There, "the interference with the plaintiff's use of the water may be treated as divisible in terms of degree, and may be apportioned among the owners of the factories, on the basis of evidence of the respective quantities of pollution discharged into the stream." Id.8

Apportionment is inappropriate for other kinds of harm, which, "by their very nature, are normally incapable of any logical, reasonable, or practical division." Id., comment on subsection (2). Examples of such harm are death, a single wound, the destruction of a house by fire, or the sinking of a barge. "Where two or more causes combine to produce such a single result, incapable of division on any logical or reasonable basis, and each is a substantial factor in bringing about the harm, the courts have refused to make an arbitrary apportionment for its own sake, and each of the causes is charged with responsibility for the entire harm." Id.

Apportionment is also inappropriate in what the Restatement describes as "exceptional" cases, "in which injustice to the plaintiff may result." Id., comment h on subsection (1). For example, "one of two tortfeasors [may be] so hopelessly insolvent that the plaintiff will never be able to collect from him the share of the damages allocated to him." Id. Where the court deems it unjust to require the innocent plaintiff to bear the risk of one of the tortfeasors' insolvency, it may refuse to apportion damages in such a case. Id.

In sum, the nature of the harm is the determining factor with respect to whether apportionment is appropriate. Ultimately, the decision whether to impose joint and several liability turns on whether there is a reasonable and just method for determining the amount of harm that was caused by each defendant (or, in some cases, by an innocent cause or by the fault of the plaintiff). The question whether the harm to the plaintiff is capable of apportionment among two or more causes is a question of law. Restatement (Second) of Torts, Sec. 434(1)(b). Once it has been determined that the harm is capable of being apportioned among the various causes of it, the actual apportionment of damages is a question of fact. Id., Sec. 434(2)(b) & comment d.

Section 433B of the Restatement sets forth the burdens of proof. As a general rule, the plaintiff must prove that the defendant's tortious conduct caused the harm. Id., Sec. 433B(1). As we have already noted, however, this rule does not apply in CERCLA cases. See note 4, supra. Nevertheless, subsection (2) of Sec. 433B, which sets forth the burdens of proof with respect to apportionment, does apply and provides as follows:

Where the tortious conduct of two or more actors has combined to bring about harm to the plaintiff, and one or more of the actors seeks to limit his liability on the ground that the harm is capable of apportionment among them, the burden of proof as to the apportionment is upon each such actor.

As explained in the comment, this rule applies only to "a proved wrongdoer who has in fact caused harm to the plaintiff." Id., comment d on subsection (2). Thus, the rule stated in subsection (2) will not permit a defendant to escape liability altogether, but only to limit its liability, if it can meet its burden of proving the amount of the harm that it caused. If it is unable to do so, it is liable for the full amount of the harm. According to the Restatement, the typical case to which this rule applies "is the pollution of a stream by a number of factories which discharge impurities into it." Id., comment c on subsection (2).

Comment e notes that there is a possibility that the rule stated in subsection (2) may cause disproportionate harm to defendants where each of a large number of them contributes a relatively small and insignificant part to the total harm. For example, "if a hundred factories each contribute a small, but still uncertain, amount of pollution to a stream, to hold each of them liable for the entire damage because he cannot show the amount of his contribution may perhaps be unjust." Id., comment e on subsection (2). The comment, however, expresses no conclusion with respect to the applicability of this illustration, noting that such a case had not arisen.

CERCLA is a strict liability statute, one of the purposes of which is to shift the cost of cleaning up environmental harm from the taxpayers to the parties who benefited from the disposal of the wastes that caused the harm. See, e.g., Chem-Dyne, 572 F. Supp. at 805-06. "The improper disposal or release of hazardous substances is an enormous and complex problem of national magnitude involving uniquely federal interests." Id. at 808. Often, liability is imposed upon entities for conduct predating the enactment of CERCLA, and even for conduct that was not illegal, unethical, or immoral at the time it occurred. We recognize the importance of keeping these facts in mind when attempting to develop a uniform federal common law for CERCLA cases. We also recognize, however, that CERCLA, as a strict liability statute that will not listen to pleas of "no fault," can be terribly unfair in certain instances in which parties may be required to pay huge amounts for damages to which their acts did not contribute. Congress recognized such possibilities and left it to the courts to fashion some rules that will, in appropriate instances, ameliorate this harshness. Accordingly, Congress has suggested, and we agree, that common-law principles of tort liability set forth in the Restatement provide sound guidance. In applying those principles to this CERCLA case, we think that it will be helpful to examine briefly some of the relevant CERCLA jurisprudence.

The first published case to address the scope of liability under CERCLA is United States v. Chem-Dyne Corp., 572 F. Supp. 802 (S.D.Ohio 1983), which was cited approvingly in the legislative history of the SARA amendments to CERCLA. In that case, 24 defendants, who allegedly generated or transported hazardous substances located at Chem-Dyne's treatment facility, sought "an early determination" that they were not jointly and severally liable for the EPA's response costs. Id. at 804. After examining the statute and its legislative history, the court concluded that provisions for joint and several liability were deleted from CERCLA "in order to avoid its universal application to inappropriate circumstances." Id. at 810. It relied on the Restatement for guidance in applying federal common law. Id.

The court described the nature of the "fairly complex factual determination" involved in deciding whether the defendants were jointly and severally liable as follows:

The Chem-Dyne facility contains a variety of hazardous waste from 289 generators or transporters, consisting of about 608,000 pounds of material. Some of the wastes have commingled but the identities of the sources of these wastes remain unascertained. The fact of the mixing of the wastes raises an issue as to the divisibility of the harm. Further, a dispute exists over which of the wastes have contaminated the ground water, the degree of their migration and concomitant health hazard. Finally, the volume of waste of a particular generator is not an accurate predictor of the risk associated with the waste because the toxicity or migratory potential of a particular hazardous substance generally varies independently with the volume of the waste.

Id. at 811. The court concluded that the defendants had not met their burden of demonstrating the divisibility of the harm and the degree to which each was responsible, and denied their motion for summary judgment. Id.

United States v. Ottati & Goss, Inc., 630 F. Supp. 1361 (D.N.H.1988), was a cost recovery action against operators and former operators of drum reconditioning businesses, property owners, and generators of wastes contained in the drums that were sent to the site for reconditioning. The evidence showed that chemical substances leaked or spilled from drums and were mixed together. Although the generators satisfied their burden of proving approximately how many drums each brought to the site, the court nevertheless imposed joint and several liability, because "the exact amount or quantity of deleterious chemicals or other noxious matter [could not] be pinpointed as to each defendant[, and] [t]he resulting proportionate harm to surface and groundwater [could not] be proportioned with any degree of accuracy as to each individual defendant." Id. at 1396.

A similar situation existed in O'Neil v. Picillo, 883 F.2d 176 (1st Cir.1989). The site at issue there was a Rhode Island pig farm that had been used as a waste disposal site. The site was described as having "massive trenches and pits 'filled with free-flowing, multi-colored, pungent liquid wastes' and thousands of 'dented and corroded drums containing a veritable potpourri of toxic fluids.' " Id. at 177. The defendants argued that it was possible to apportion the removal costs, because there was evidence of the total number of barrels excavated during each phase of the clean-up, the number of barrels in each phase attributable to them, and the cost of each phase. Id. at 181. There was testimony that, of the approximately 10,000 barrels excavated, only 300-400 could be attributable to a particular defendant. Id. at 182. The court concluded that because most of the waste could not be identified, and the defendants had the burden of accounting for the uncertainty, the imposition of joint and several liability was appropriate.9

On the other hand, the Third Circuit reversed a summary judgment in favor of the EPA, and remanded the case for further factual development on the scope of liability, in United States v. Alcan Aluminum Corp. (Alcan-Butler), 964 F.2d 252, 255 (3d Cir.1992). This case involved the Butler Tunnel Site, a network of approximately five square miles of underground mines, tunnels, caverns, pools, and waterways, drained by the Butler Tunnel into the Susquehanna River in Pennsylvania. During the 1970s, millions of gallons of liquid wastes containing hazardous substances were disposed of through a borehole that led directly into the mine workings. In 1985, 100,000 gallons of contaminated water were released from the site into the river.

The government filed a cost-recovery action against 20 defendants; all but Alcan settled. The district court granted summary judgment for the government, holding that Alcan was jointly and severally liable for the response costs. The Third Circuit held that the "intensely factual nature of the 'divisibility' issue" highlighted the district court's error in granting summary judgment without conducting a hearing. Id. at 269. It remanded the case in order to give Alcan the opportunity to limit or avoid liability by attempting to prove its personal contribution to the harm to the Susquehanna River. Thus, under the Third Circuit's approach, Alcan could escape liability altogether if it could prove that its "emulsion did not or could not, when mixed with other hazardous wastes, contribute to the release and the resultant response costs." Id. at 270.

The Third Circuit noted that the analysis involved in apportioning several liability is similar to that involved in apportioning damages among jointly and severally liable defendants in an action for contribution, because both focus on what harm was caused by the defendant. Id. at 270 n. 29. However, it stated that the issue of joint and several liability should be resolved at the initial liability stage, rather than at the contribution stage.10 It noted that drastic consequences could result from delaying that determination, because "a defendant could easily be strong-armed into settling where other defendants have settled in order to avoid being held liable for the remainder of the response costs." Id. It also noted that contribution would not be available from settling defendants, pursuant to CERCLA Sec. 113(f)(2).11 Id.

The Second Circuit essentially adopted the Third Circuit's approach to joint and several liability in another case involving Alcan, United States v. Alcan Aluminum Corp. (Alcan-PAS), 990 F.2d 711 (2d Cir.1993). That case involved a waste disposal and treatment center operated during the 1970s by Pollution Abatement Services (PAS). Alcan used PAS for the disposal or treatment of 4.6 million gallons of oil emulsion. The government brought a cost-recovery action against 83 defendants. As in Alcan-Butler, all of the defendants except Alcan settled. The Second Circuit reversed a summary judgment in favor of the government, stating that "Alcan should have the opportunity to show that the harm caused at PAS was capable of reasonable apportionment." Id. at 722. It held that Alcan was entitled to "present evidence relevant to establishing divisibility of harm, such as, proof disclosing the relative toxicity, migratory potential, degree of migration, and synergistic capacities of the hazardous substances at the site." Id.

The court stated that Alcan could escape liability if it could prove that its oil emulsion, when mixed with other hazardous wastes, did not contribute to the release and resulting clean-up costs. It acknowledged that "causation is being brought back into the case--through the backdoor, after being denied entry at the frontdoor--at the apportionment stage." Id. However, it pointed out that causation was "reintroduced only to permit a defendant to escape payment where its pollutants did not contribute more than background contamination and also cannot concentrate." Id.

With respect to the timing of the joint and several liability inquiry, the Second Circuit stated that it preferred the Third Circuit's "common sense approach." Id. It ultimately concluded, however, that "the choice as to when to address divisibility and apportionment are questions best left to the sound discretion of the trial court in the handling of an individual case." Id. at 723.

A "moderate" approach to joint and several liability was adopted in United States v. A & F Materials Co., Inc., 578 F. Supp. 1249 (S.D.Ill.1984). That case involved a disposal site at which over 7,000,000 gallons of waste were deposited. The court concluded that a rigid application of the Restatement approach to joint and several liability was inappropriate. Under the Restatement approach, a defendant who could not prove its contribution to the harm would be jointly and severally liable. The court thought that such a result would be inconsistent with congressional intent, because Congress was "concerned about the issue of fairness, and joint and several liability is extremely harsh and unfair if it is imposed on a defendant who contributed only a small amount of waste to a site." Id. at 1256.

The court concluded that six factors delineated in an unsuccessful amendment to CERCLA proposed by Representative (now Vice President) Gore could be used to "soften" the modern common law approach to joint and several liability in appropriate circumstances. Under this "moderate" approach, a court has the power to impose joint and several liability upon a defendant who cannot prove its contribution to an injury, but it also has the discretion to apportion damages in such a situation according to the "Gore factors":

(i) the ability of the parties to demonstrate that their contribution to a discharge[,] release or disposal of a hazardous waste can be distinguished;

(ii) the amount of the hazardous waste involved;

(iii) the degree of toxicity of the hazardous waste involved;

(iv) the degree of involvement by the parties in the generation, transportation, treatment, storage, or disposal of the hazardous waste;

(v) the degree of care exercised by the parties with respect to the hazardous waste concerned, taking into account the characteristics of such hazardous waste; and

(vi) the degree of cooperation by the parties with Federal, State, or local officials to prevent any harm to the public health or the environment.

Id. at 1256. The court stated that its moderate approach would promote fairness by allowing courts to be sensitive to the inherent unfairness of imposing joint and several liability on minor contributors, and to make rational distinctions based on such factors as the amount and toxicity of a particular defendant's contribution to a waste site. Id. at 1257.

In Allied Corp. v. Acme Solvents Reclaiming, Inc., 691 F. Supp. 1100 (N.D.Ill.1988), a private cost recovery action in which the government was not a party, the court adopted the A & F moderate approach to joint and several liability. However, it expressed no opinion on the propriety of that approach in cost recovery actions involving the government as plaintiff. Id. at 1118 & n. 12.

The A & F moderate approach, to the extent it is inconsistent with the Chem-Dyne approach to joint and several liability, was rejected in United States v. South Carolina Recycling and Disposal, Inc., 653 F. Supp. 984 (D.S.C.1986), aff'd in part and vacated in part, United States v. Monsanto Co., 858 F.2d 160 (4th Cir.1988). That case involved a site at which there were "thousands of corroded, leaking drums ... not segregated by source or waste type. Unknown, incompatible materials commingled to cause fires, fumes, and explosions." 653 F. Supp. at 994. The district court concluded that the harm was indivisible, because all of the substances at the site contributed synergistically, and it was impossible to ascertain the degree or relative contribution of each substance. Id. The court rejected volume as a basis for apportionment, finding that it "is not an accurate predictor of the risk associated with the waste because the toxicity or migratory potential of a particular hazardous substance generally varies independently of the volume." Id. at 995 (quoting Chem-Dyne, 572 F. Supp. at 811). The court concluded that "[s]uch arbitrary or theoretical means of cost apportionment do not diminish the indivisibility of the underlying harm, and are matters more appropriately considered in an action for contribution between responsible parties after plaintiff has been made whole." Id.

On appeal, the Fourth Circuit affirmed the imposition of joint and several liability. 858 F.2d at 173. It noted that the generator defendants had presented no evidence of a relationship between the volume of waste, the release of hazardous substances, and the harm at the site. Because the substances had commingled, apportionment was impossible "without some evidence disclosing the individual and interactive qualities of the substances deposited there." Id. at 172. Because "[c]ommon sense counsels that a million gallons of certain substances could be mixed together without significant consequences, whereas a few pints of others improperly mixed could result in disastrous consequences," the court concluded that evidence of the relative toxicity, migratory potential, and synergistic capacity of the various substances was both relevant and necessary. Id. at 172 & n. 26. The court noted, however, that under other circumstances, volume could be a reasonable basis for apportioning liability, in a situation in which independent factors had no substantial effect on the harm to the environment. Id. at 172 & n. 27.

The Fourth Circuit apparently agreed with the district court's rejection of the A & F moderate approach, stating that, while equitable factors are relevant in an action for contribution, "[t]hey are not pertinent to the question of joint and several liability, which focuses principally on the divisibility among responsible parties of the harm to the environment." 858 F.2d at 171 n. 22. Other courts have similarly concluded that equitable factors, such as those listed in the Gore amendment, have no place in making the decision whether to impose joint and several liability, but are appropriate in an action for contribution among jointly and severally liable defendants. See Alcan-Butler, 964 F.2d at 270 n. 29 ("the contribution proceeding is an equitable one in which a court is permitted to allocate response costs based on factors it deems appropriate, whereas the court is not vested with such discretion in the divisibility determination"); United States v. Western Processing Co., Inc., 734 F. Supp. 930, 938 (W.D.Wash.1990) ("defendants may ... bring contribution actions for ultimate allocation of damages among the responsible parties where it is entirely appropriate to utilize the Gore Factors to determine the burden each party must bear"); United States v. Stringfellow, 661 F. Supp. at 1060 ("the Court's discretion in apportioning damages among the defendants during the contribution phase does not [a]ffect the defendants' liability").12

To summarize, our review of the jurisprudence leads us to conclude that there are three distinct, although closely-related, approaches to the issue of joint and several liability. The first is the "Chem-Dyne approach," which relies almost exclusively on the principles of the Restatement (Second) of Torts. Under that approach, a defendant who seeks to avoid the imposition of joint and several liability is required to prove the amount of harm it caused.

The second approach, the "Alcan approach," is adopted by the Second and Third Circuits. Although that approach also relies on the Restatement, it recognizes that, under the unique statutory liability scheme of CERCLA, the plaintiff's common law burden of proving causation has been eliminated. Under the Restatement, the plaintiff must first prove that the defendant's conduct was a substantial factor in causing the harm; the defendant may limit its liability by proving its contribution to the harm. In contrast, the Alcan approach suggests that a defendant may escape liability altogether if it can prove that its waste, even when mixed with other wastes at the site, did not cause the incurrence of response costs.

The third approach is the "moderate" approach taken in A & F. Under that approach, the court applies the principles of the Restatement in determining whether there is a reasonable basis for apportionment. If there is not, the court may impose joint and several liability; the court, however, retains the discretion to refuse to impose joint and several liability where such a result would be inequitable.

Although these approaches are not entirely uniform, certain basic principles emerge. First, joint and several liability is not mandated under CERCLA; Congress intended that the federal courts impose joint and several liability only in appropriate cases, applying common-law principles. Second, all of the cases rely on the Restatement in resolving the issues of joint and several liability. The major differences among the cases concern the timing of the resolution of the divisibility question, whether equitable factors should be considered, and whether a defendant can avoid liability for all, or only some portion, of the damages. Third, even where commingled wastes of unknown toxicity, migratory potential, and synergistic effect are present, defendants are allowed an opportunity to attempt to prove that there is a reasonable basis for apportionment (although they rarely succeed); where such factors are not present, volume may be a reasonable means of apportioning liability.

With respect to the timing of the "divisibility" inquiry, we believe that an early resolution is preferable. We agree with the Second Circuit, however, that this is a matter best left to the sound discretion of the district court. We also agree with the majority view that equitable factors, such as those listed in the Gore amendment, are more appropriately considered in actions for contribution among jointly and severally liable parties, than in making the initial determination of whether to impose joint and several liability.13 We therefore conclude that the ChemDyne approach is an appropriate framework for resolving issues of joint and several liability in CERCLA cases. Although we express no opinion with respect to the Alcan approach, because it is not necessary with respect to the issues we are faced with in this case, we nevertheless recognize that the Restatement principles must be adapted, where necessary, to implement congressional intent with respect to liability under the unique statutory scheme of CERCLA.14

We now turn to consider the application of these traditional and evolving common law principles of joint and several liability to the facts of this case.

First, we conclude that the district court erred in determining that there is no reasonable basis for apportionment. We reject the EPA's assertion that the clearly erroneous standard of review applies to these findings of the district court. According to the Restatement, "the question whether the harm to the plaintiff is capable of apportionment among two or more causes is a question of law." Restatement (Second) of Torts, Sec. 434.

In the district court, the EPA contended that there was no reasonable basis for apportionment, because the harm to the Trinity Aquifer was a single harm, and a that single harm is the equivalent of an indivisible harm, thus mandating the imposition of joint and several liability. Apparently now recognizing the lack of support for that position,15 the EPA on appeal acknowledges that apportionment is available, at least theoretically, when there is a reasonable basis for determining the contribution of each cause to a single harm. It asserts, however, that Sequa failed to meet its burden of proof on that issue. Sequa responds that the district court was misled by the EPA's incorrect view of the law, and erroneously required it to prove a certain--as opposed to reasonable--basis for apportionment.

Essentially, the question whether there is a reasonable basis for apportionment depends on whether there is sufficient evidence from which the court can determine the amount of harm caused by each defendant. If the expert testimony and other evidence establishes a factual basis for making a reasonable estimate that will fairly apportion liability, joint and several liability should not be imposed in the absence of exceptional circumstances. The fact that apportionment may be difficult, because each defendant's exact contribution to the harm cannot be proved to an absolute certainty, or the fact that it will require weighing the evidence and making credibility determinations, are inadequate grounds upon which to impose joint and several liability.16

Our review of the record convinces us that Sequa met its burden of proving that, as a matter of law, there is a reasonable basis for apportionment. This case is closely analogous to the Restatement 's illustrations in which apportionment of liability is appropriate. For example, where cattle owned by two or more defendants destroy the plaintiff's crops, the damages are apportioned according to the number of cattle owned by each defendant, based on the reasonable assumption that the respective harm done is proportionate to that number. Thus, the Restatement suggests that apportionment is appropriate even though the evidence does not establish with certainty the specific amount of harm caused by each defendant's cattle, and even though there is a possibility that only one of the defendant's cattle caused all of the harm, while the other defendant's cattle idly stood by. Likewise, pollution of a stream by two or more factories may be treated as divisible in terms of degree, and apportioned among the defendants on the basis of evidence of the respective quantities of pollution discharged by each.

As is evident from our previous discussion of the jurisprudence, most CERCLA cost-recovery actions involve numerous, commingled hazardous substances with synergistic effects and unknown toxicity. In contrast, this case involves only one hazardous substance--chromium--and no synergistic effects. The chromium entered the groundwater as the result of similar operations by three parties who operated at mutually exclusive times. Here, it is reasonable to assume that the respective harm done by each of the defendants is proportionate to the volume of chromium-contaminated water each discharged into the environment.

Even though it is not possible to determine with absolute certainty the exact amount of chromium each defendant introduced into the groundwater, there is sufficient evidence from which a reasonable and rational approximation of each defendant's individual contribution to the contamination can be made. The evidence demonstrates that Leigh owned the real property at the site from 1967 through 1981, and conducted chrome-plating activities there in 1971 and 1972. In 1972, Bell purchased the assets of the shop and leased the property from Leigh. It continued to conduct similar, but more extensive, chrome-plating activities there until mid-1976. In August 1976, Sequa purchased the assets from Bell, leased the property from Leigh, and conducted similar chrome-plating activities at the site until late 1977. In response to the EPA's motion for summary judgment, Sequa introduced evidence regarding chrome flake purchases during each operator's tenure. It also introduced evidence with respect to the value of the chrome-plating done by each, as well as summaries of sales. Given the number of years that had passed since the activities were conducted, the records of these activities were not complete.17 However, there was testimony from various witnesses regarding the rinsing and wastewater disposal practices of each defendant, and the amount of chrome-plating activity conducted by each.18

During the Phase III hearing, Sequa introduced expert testimony regarding a volumetric approach to apportionment. The first expert, Henderson, calculated the total amount of chromium that had been introduced into the environment by Leigh, Bell, and Sequa, collectively and individually. The second expert, Mooney, calculated the amount of chromium that would have been introduced into the environment by each operator on the basis of electrical usage records.

In addition to rejecting apportionment because of competing theories, the district court also rejected volume as a basis for apportionment, because there was no method of dividing the liability among the defendants which would rise to any level of fairness above mere speculation. It stated that each of the proposed apportionment methods involved significant assumption factors, because records had been lost, and because the theories differed significantly.

The existence of competing theories of apportionment is an insufficient reason to reject all of those theories. It is true, as the district court noted, that the records of chrome-plating activity were incomplete. However, under the facts and circumstances of this case, and in the light of the other evidence that is available, that factor may be taken into account in apportioning Sequa's share of the liability. Finally, the fact that Sequa's experts relied on certain assumptions in forming their opinions is not fatal to Sequa's ability to prove that there is a reasonable basis for apportionment. Expert opinions frequently include assumptions. If those assumptions are well-founded and reasonable, and not inconsistent with the facts as established by other competent evidence, they may be sufficiently reliable to support a conclusion that a reasonable basis for apportionment exists.19

In sum, we conclude that the district court erred in imposing joint and several liability, because Sequa met its burden of proving that there is a reasonable basis for apportioning liability among the defendants on a volumetric basis. We therefore remand the case to the district court for apportionment.

Sequa also challenges the EPA's decision to provide an alternate water supply (AWS) as an interim measure pending the completion of final remedial action. The scope of our review of the EPA's selection of the AWS is governed by the 1986 amendments to CERCLA, which provide that such review is "limited to the administrative record." 42 U.S.C. Sec. 9613(j)(1). We are to uphold the EPA's decision "unless the objecting party can demonstrate, on the administrative record, that the decision was arbitrary and capricious or otherwise not in accordance with law." 42 U.S.C. Sec. 9613(j)(2).

The scope of review under the "arbitrary and capricious" standard is narrow and a court is not to substitute its judgment for that of the agency. Nevertheless, the agency must examine the relevant data and articulate a satisfactory explanation for its action including a rational connection between the facts found and the choice made.... In reviewing that explanation, we must consider whether the decision was based on a consideration of the relevant factors and whether there has been a clear error of judgment.

Motor Vehicle Manufacturers Ass'n v. State Farm Mutual Automobile Ins. Co., 463 U.S. 29, 43, 103 S. Ct. 2856, 2867, 77 L. Ed. 2d 443 (1983) (citations and internal quotation marks omitted).

Sequa challenges the EPA's decision to provide the AWS on a number of grounds, including that: (1) the administrative record demonstrates that the EPA failed to recognize that "substantial danger to public health or the environment," as specified in the National Contingency Plan, is the standard against which the decision to implement an alternative water supply system must be measured; (2) there is no analysis of why the EPA believed the public health was at risk and required protection at the subject site; (3) the Safe Drinking Water Act's maximum contaminant level for chromium is based on a lifetime (70-year) exposure, but the alternate water supply system was merely a short-term (10-15 year) response; further, the administrative record contains no discussion of whether chromium presents a danger to humans on the basis of short-term exposure; and (4) the EPA failed to analyze the likelihood that the contaminated water would be ingested.

The EPA's defense of its decision to implement the alternate water supply system is, we think, singularly weak. The EPA contends primarily that we should defer to its technical expertise. It argues that the existence of chromium at levels exceeding the maximum contaminant level allowed under the SDWA presumptively establishes that its response was appropriate. We cannot agree.

Although the arbitrary and capricious standard of review is very lenient on the agency, and we will not substitute our judgment for that of the agency, "[j]udicial review 'must be based on something more than trust and faith in EPA's experience.' " American Petroleum Institute v. E.P.A., 661 F.2d 340, 349 (5th Cir.1981) (quoting Appalachian Power Co. v. Train, 545 F.2d 1351, 1365 (4th Cir.1976)). Our determination of whether the EPA's decision was arbitrary and capricious must be made on the basis of the rationale relied on by the EPA as contained in the administrative record. We will not accept the EPA's post-hoc rationalizations in justification of its decision, nor will we attempt to supply a basis for its decision that is not supported by the administrative record. See State Farm, 463 U.S. at 50, 103 S.Ct. at 2870.20

After thoroughly reviewing the administrative record, we conclude that the EPA's decision to furnish the AWS was arbitrary and capricious. In vain we have searched the over 5,000 pages of administrative record, and found not one shred of evidence that anyone in the area was actually drinking chromium-contaminated water. Amazingly, the EPA made no attempt to learn whether anyone was drinking the water, or whether anyone intended to utilize the AWS, until after it had made its decision to construct the AWS. One would think that surely such information was essential in order to reach an informed, rational decision as to whether an AWS was necessary, and whether it would reduce any significant threat to public health. The administrative record reveals that the chromium-contaminated wells in the area all served commercial establishments, which the EPA prohibited from connecting to the AWS. Moreover, the EPA did not require residents to connect to the system, and did not prohibit them from using contaminated water from their wells. Thus, on the basis of the administrative record, it appears that the AWS did not even reduce, much less eliminate, any public health threat. No technical expertise is necessary to discern that the EPA's implementation of the AWS was arbitrary and capricious, as well as a waste of money.21

All Costs?

Having determined that the EPA's decision to implement the AWS was arbitrary and capricious, we must now decide whether the EPA nevertheless is entitled to recover its costs for designing and constructing the AWS.

CERCLA Sec. 107 provides for the recovery of the following costs:

(A) all costs of removal or remedial action incurred by the United States Government or a State or an Indian tribe not inconsistent with the national contingency plan;

(B) any other necessary costs incurred by any other person consistent with the national contingency plan;

(C) damages for injury to, destruction of, or loss of natural resources, including the reasonable costs of assessing such injury, destruction, or loss resulting from such a release; and

(D) the costs of any health assessment or health effects study carried out under section 9604(i) of this title.

42 U.S.C. Sec. 9607(a)(4) (emphasis added).

Sequa contends that the EPA is authorized to recover only reasonable and necessary costs, relying on the statutory language as well as federal procurement laws and regulations. The EPA takes the position that it is entitled to recover all costs--even if unreasonable or unnecessary22--unless Sequa proves that such costs are inconsistent with the National Contingency Plan.23 The district court held that the EPA could recover all of its response costs, so long as they were not the product of "gross misconduct" by the agency.

Although we approve of the district court's attempt to impose some restraints on the EPA's ability to recover costs from private parties, we find no statutory basis for its "gross misconduct" limitation. Nevertheless, we are troubled by the implications of the EPA's position on this issue. Sequa contends that, under the EPA's interpretation, defendants will be liable even if the EPA allows a contractor to pay its officers and other employees unjustified millions and allows each of them a Rolls-Royce for transportation. Interestingly, the EPA did not attempt to refute Sequa's assertion, either in its appellate brief or at oral argument. Instead, the EPA asserts a policy reason to support its interpretation:

By refusing to permit defendants to defend against cost recovery actions by engaging in detailed attacks on the "reasonableness" of individual government cost items, Congress provided an incentive to those defendants to conduct the necessary response actions themselves. Where defendants refuse to conduct the appropriate response actions, CERCLA allows the Government to undertake the response actions it deems necessary and appropriate without being constrained by the possibility that each line item of the costs of these actions will be challenged in cost recovery.

In addition, the EPA asks us to take comfort in the fact that, through internal agency audits and other forms of self-policing, costs will be controlled.

Acceptance of the EPA's position would effectively prohibit judicial review of the EPA's expenditures. In short, we would give the EPA a blank check in conducting response actions.24 We seriously doubt that Congress intended to give the EPA such unrestrained spending discretion.25 Moreover, such unbridled discretion removes any restraint upon the conduct of the EPA in exercising its awesome powers; if the EPA knows there are no economic consequences to it, its decisions and conduct are likely to be less responsible.

We do not have to decide the question in this case, however, because the only costs Sequa challenges as unreasonable and unnecessary are those associated with implementation of the alternate water supply system, a decision that we have already concluded was arbitrary and capricious. The Tenth Circuit recently held that, "[t]o show that the government's response action is inconsistent with the NCP, a defendant must demonstrate that the EPA acted arbitrarily and capriciously in choosing a particular response action to respond to a hazardous waste site." Hardage, 982 F.2d at 1442. We find this reasoning persuasive, as well as adequate for resolving the issue before us, and therefore adopt it.26 Because the decision to implement an AWS was arbitrary and capricious, it is inconsistent with the NCP. Accordingly, the EPA is not entitled to recover the costs of designing and constructing the AWS.

We realize that, as a result of our decision disallowing the EPA's costs for the AWS, those costs will have to be borne by the Superfund. Although regrettable, this is the inevitable result of arbitrary and capricious EPA decisionmaking. Without knowing, or even attempting to learn, whether the AWS would serve to protect the safety and health of anyone, the EPA officiously ignored the comments of Bell and Sequa, and the results of its own remedial investigation, and stubbornly proceeded to spend over $300,000 to furnish a water supply system that was not needed, was not allowed to be used by the commercial establishments whose wells (according to the administrative record) were the only ones with chromium contamination in excess of the SDWA standards, and did very little--indeed, if anything--to reduce any perceived public health threat posed by the chromium-contaminated groundwater. We can only assume that the EPA was not concerned about the cost of the AWS, because it believed that it could recover whatever was spent from Sequa. Although the EPA's powers under CERCLA are indeed broad, Congress has not provided that private parties must pay for the consequences of arbitrary and capricious agency action.

CERCLA Sec. 113(f)(2), 42 U.S.C. Sec. 9612(f)(2), provides that a settlement by one defendant "reduces the potential liability to the others by the amount of the settlement." Bell and Leigh settled with the EPA for a combined total of $1.1 million. Sequa contends that the district court (1) improperly refused to credit that amount against the total recovery obtained by the government; and (2) erred in allocating the Bell settlement proceeds, $1,000,000, first to amounts for which Bell was severally liable (litigation costs incurred by the EPA in the Bell bankruptcy adversary proceeding before suit was filed against Sequa), and then toward costs for which Bell and Sequa were jointly and severally liable.

Because Sec. 113(f)(2) logically can be applied only to reduce a defendant's joint and several liability, which we have decided is inappropriate in this case, we need not address this issue.

CERCLA Sec. 107(a)(4) provides for the recovery of prejudgment interest, which "shall accrue from the later of (i) the date payment of a specified amount is demanded in writing, or (ii) the date of the expenditure concerned." 42 U.S.C. Sec. 9607(a)(4). The district court awarded prejudgment interest calculated from the date of expenditures.

Sequa contends that the statute requires a written demand for a specified amount of response costs before any prejudgment interest may be awarded. The EPA does not contest Sequa's assertion that a written demand is required, but contends that the notices sent to Sequa, advising it generally that the United States considered it to be potentially liable for response costs, satisfied that requirement. The EPA further contends that the 1986 ROD to install the alternate water supply system put Sequa on notice of the potential cost of that decision. Finally, the EPA asserts that the complaint constitutes the necessary written demand.

The statute plainly requires a written demand for specified response costs as a prerequisite to an award of prejudgment interest. Neither the notices informing Sequa that generally the EPA would look to it for potential reimbursement "at some future time," nor the ROD satisfy that requirement. Although the complaint does not specify an exact amount, we conclude that it constitutes a sufficient written demand for payment. We therefore hold that, with respect to costs incurred before the complaint was filed, prejudgment interest should be assessed from the date the complaint was filed. With respect to costs, if any, incurred after the complaint was filed, prejudgment interest should be assessed on those costs from the date of the expenditures.

The judgment of the district court is REVERSED insofar as it imposes joint and several liability and allows recovery of the costs of designing and constructing the AWS; the portion of the judgment awarding prejudgment interest is VACATED; and the case is REMANDED for further proceedings consistent with this opinion.

REVERSED in part, VACATED in part, and REMANDED.

PARKER, District Judge,* concurring in part and dissenting in part:

I concur in the majority's thorough and reasoned approach to the difficult questions addressed in Parts IV(A) and IV(B) of its opinion. I concur also in the majority's holding regarding the prejudgment interest issue. However, I must dissent in substantial part from the majority opinion, for the following reasons.

* Joint and Several Liability

I cannot agree with the majority's holding on the joint and several liability/quantitative apportionment issue in this case. I do agree that the determination of whether the type harm involved in this case is capable of quantitative apportionment is a question of law. And the majority is correct that the single chromium harm suffered by the Trinity Aquifer is the sort theoretically capable of apportionment. However, while Sequa met its legal burden of establishing that the type harm involved is capable of apportionment, it failed to meet its factual burden relative to apportionment. If proof exists by which the fact-finder could determine, on a reasonable basis, the extent of environmental injury attributable to a party, then certainly that party is entitled to escape the heavy hand of joint and several liability and to have its liability restricted to its actual, quantitative contribution to the single harm. The majority correctly places the burden of proof on the party seeking such a finding, to produce credible evidence to meet its burden. But the majority confuses the distinction between the legal burden that the single harm at issue caused is of a type capable of apportionment, and the factual burden of proving the amount of harm attributable to a particular party. See majority opinion at 903. ("Our review of the record convinces us that Sequa met its burden of proving that, as a matter of law, there is a reasonable basis for apportionment." This case is closely analogous to the Restatement's illustrations in which apportionment of liability is appropriate.").

The gist of the majority opinion is this legal fallacy: because the evidence is clear that Sequa did not cause 100% of the harm to the aquifer, Sequa must be entitled to a finding by the district court apportioning the amount of harm attributable to it under the Restatement (Second) of Torts, Sec. 433. We are not to approach our analytical task from that end. The majority's "rule of thumb" miscasts the role of the district court and eviscerates the very concept of joint and several liability.

I agree with the majority that certainty is not required. What is required is proof by a preponderance of the evidence. The majority properly embraces the applicability of the Restatement (Second) of Torts to this case, but then seeks to divorce itself from the applicable preponderance of the evidence standard of proof so as to mandate that the district court "pick a number" apportioning liability.

Civil cases are decided by a preponderance of the evidence because such proof affords a reasonable basis for decision. In other words, while certainty of proof is not required in civil cases, probability is. Evidence by "fifty-one percent," or to the extent of "more likely than not," is deemed sufficiently reliable for resolution of civil disputes. But proof by less than this amount is unacceptably speculative; and amounts to mere possibility, not probability. Dean William Prosser said it well in his influential treatise:

On the issue of the fact of causation, ... [the one bearing the burden of proof by a preponderance of the evidence] must introduce evidence which affords a reasonable basis for the conclusion that it is more likely than not that [the causation exists]. A mere possibility of such causation is not enough;[ ] and when the matter remains one of pure speculation or conjecture,[ ] or the probabilities are at best evenly balanced, [ ] it becomes the duty of the court to direct a verdict for the defendant. Where the conclusion is not one within the common knowledge of laymen, expert testimony may provide a sufficient basis for it, [ ] but in the absence of such testimony it may not be drawn.1

If proof by a preponderance of the evidence is to be abandoned in CERCLA apportionment cases, the district court is at least entitled to guidance regarding the level of possibilities that is acceptable. Will 10% do? 20%? 30%?2

The majority quotes the Restatement's Sec. 433:

(1) Damages for harm are to be apportioned among two or more causes where

(a) there are distinct harms, or

(b) there is a reasonable basis for determining the contribution of each cause to a single harm.

(2) Damages for any other harm cannot be apportioned among two or more causes.

The majority proceeds to quote comment d on subsection (1) of this Section, to the effect that a single harm that is conceptually divisible, "while not so clearly marked out as severable into distinct parts, [is] still capable of division upon a reasonable and rational basis, and of fair apportionment among the causes responsible.... Where such apportionment can be made without injustice to any of the parties, the court may require it to be made." The majority discusses two examples of such harm given in this comment: the first being where cattle owned by two or more persons trespass upon another's land and destroy the other's crops; and the second involving the pollution of a stream by two or more factories. But in both of these examples, as the Restatement's comment explains, a reasonable, factual basis for division must exist in order for the court to actually draw the possible apportionment. In the cattle example, the comment explains that, although "the aggregate harm is a lost crop, ... it may nevertheless be apportioned among the owners of the cattle, on the basis of the number owned by each, and the reasonable assumption that the respective harm done is proportionate to that number." In the stream pollution example, the comment makes it plain that "the interference with the plaintiff's use of the water may be treated as divisible in terms of degree, and may be apportioned among the owners of the factories, on the basis of evidence of the respective quantities of pollution discharged into the stream." If the Restatement (Second) of Tort's term, "reasonable basis," as used in the majority opinion, means something other than preponderance of the evidence, the majority should at least say so, and why.

"As other courts have noted, apportionment itself is an intensely factual determination." United States v. Alcan Aluminum Corp., 990 F.2d 711, 722 (2nd Cir.1993) (citing e.g., United States v. Chem-Dyne Corp., 572 F. Supp. 802, 811 (S.D.Ohio 1983)). The example used by the majority is a good one. If cows belonging to Farmers A and B damage another's crop, that is the type harm that as a matter of law is capable of apportionment. However, to evade joint and several liability, Farmer A or B must meet the burden of proving that apportionment is reasonable on some basis, such as the number of cows in the field belonging to each farmer or the amount of time each farmer's cows were in the field. Proof that each farmer had some unknown number of cows in the field, or that an established number of cows belonging to each farmer were in the field for some unknown period of time, is not enough--because under such circumstances the fact-finder is left to speculate on the question of the amount of harm reasonably attributable to each farmer's cows.

In this case an experienced and careful district judge heard and reviewed the quantitative apportionment testimony and exhibits in this case, and it possessed opportunities to assess their convincingness far superior to those of this (appellate) court. The district court found Sequa failed to meet its quantitative apportionment burden of proof by a preponderance of the credible evidence. That finding is reviewable at the Court of Appeals only on the basis of whether it was clearly erroneous. The district court's finding was not clearly erroneous.

The district court afforded Sequa its full apportionment due--the opportunity to avoid joint and several liability by meeting its burden of proof through the presentation of credible evidence persuading the district court that the amount of harm caused by Sequa can be apportioned to a level of knowledge that is sufficiently reliable (i.e., by a preponderance of the evidence). Sequa simply failed to meet its burden. The majority opinion notwithstanding, there is no reason to believe Sequa's appellate claim that the district court applied the wrong standard for apportionment (i.e., a standard of certainty, as opposed to the appropriate standard, of a reasonable basis). Rather, the district court's analysis, in the record, demonstrates that court's clear understanding of and application of the appropriate reasonable basis standard for apportionment questions like the one it faced. See e.g., District Court Order of May 9, 1990 (emphasis added here) ("this Court is of the opinion the chromium contamination found in the ground waters below the Odessa I Site is not divisible. The evidence at both the Phase I and Phase III hearings clearly demonstrated there is no method of dividing the liability among the Defendants which would rise to any level of fairness above mere speculation. * * * Having heard the evidence adduced at trial of Phase III, this Court is of the opinion none of the [defendants' proffered methods of actual quantitative apportionment] offer viable methods for dividing liability among John Leigh, Bell or Sequa.").

The majority remands the case to the district court for a finding apportioning liability on a volumetric basis. Such was, however, precisely the purpose of Phase III of the trial. In Phase III of the trial, the district court heard approximately 400 pages of testimony from 19 witnesses, 3 of whom were experts. The district court reviewed over 150 exhibits: 80 new exhibits were admitted during Phase III of the trial; and the district court allowed for the more than 70 exhibits from Phase I to also be used during Phase III.

A review of the record reveals that Sequa attempted to climb the preponderance hill by focusing on several potential methods of achieving a reasonable basis for quantitative apportionment of liability on a volumetric basis.

Under one proffered method of such apportionment by a Sequa expert, the expert assumed that Sequa's electrical usage for plating operations was 30% of its total electrical usage, while, in contrast, he attributed to both Bell and Leigh a plating percentage of 50% of their respective total electrical usages. But the bases for this expert's electrical percentage assumptions were effectively refuted by other evidence in the case.

Sales records served as the springboard for another proffered method of apportionment. The sales record approach suffered fatally from Sequa's ability to produce only scattered invoices.

An attempt was then made to compare the defendants' expense records. However, the only expense records for Sequa demonstrated that it purchased 3,500 pounds of chromic acid flake within a three month period in 1977. Sequa's other records were destroyed. Any attempt to extrapolate from the three month period in 1977 would have been at best speculative.

A Sequa expert also assumed that Sequa had no waste disposal after the installation of a catch tank. The credibility of this assumption was fatally eroded by contrary evidence--of substantial overflows, spills of plating solution, leaks in the plating tanks, and plating solution dumped by Sequa.

Indeed, the only evidence the district court could view with any comfort was evidence of relative times of facility ownership and the periods of plating activity by the defendants. Yet, the apportionment import of even this evidence was reduced to mere speculation when attempts were made to prove the actual level, or quantity, of plating activity conducted during the known periods of time. In the language of the majority's cited example of cows in the field: the defendants evidenced what periods of time each farmer had cows in the field, but failed to demonstrate to any degree above speculation how many cows each farmer had in the field.

This case is a simple one by CERCLA standards. But it is nonetheless quite typical of CERCLA-apportionment cases: years after the pollution at issue, it is very difficult for a defendant to prove by a preponderance of the evidence even its rough share of responsibility for the single harm caused by pollution. This is why the equitable (contribution) phase of CERCLA response cost proceedings is so important--as Congress expressly recognized in the 1986 amendments to CERCLA (SARA). See H.R. No. 99-253(I), 99th Cong., 2d Sess. 79, reprinted in 1986 U.S.C.C.A.N. 2835, 2861 (SARA "confirms" federal right of contribution under CERCLA); see also United States v. Alcan Aluminum Corp., 990 F.2d 711, 724 (2nd Cir.1993) ("In [SARA] courts are granted implicit authority, using appropriate equitable factors, to 'allocate response costs among liable parties.' ") (emphasis added) (quoting O'Neil v. Picillo, 883 F.2d 176, 179 (1st Cir.1989)). Sequa's evidence will not improve upon remand. There is no more apportionment evidence available. Unless the district court's view of the evidence on remand is somehow "enlightened" by the majority's view of the evidence, the district court will have to apply a standard of proof of less than a preponderance of the evidence in order to reach a decision in conformity with the apportionment result mandated by the majority.

Actually, the district court did attempt to "apportion" the defendants' liability on a basis other than a reasonable, amount of contribution basis--by making alternative findings on a purely equitable basis taking into account the facts that: Bell occupied the site for the longest period of time; Sequa gained access to the site with knowledge that chromium contamination was a problem and measures to correct the contamination were necessary; and Leigh accrued the least financial gain from the chrome plating venture, but cooperated with the government in the government's efforts to discover the sources of the contamination. The district court apportioned the defendants' equitable responsibility for costs at 35% each to Bell and Sequa, and 30% to Leigh. See District Court Order of May 9, 1990 ("this Court is of the opinion the liability of the parties for contamination of the Chromium I Site is indivisible other than by equitable means. * * * In the alternative [to accepting the proposed Partial Consent Decree attacked by Sequa on quantitative apportionment and equitable cost allocation grounds], this Court is of the opinion the responsibility for costs should be divided roughly equally among the parties with Bell and Sequa shouldering 35% of the burden each and john [sic ] Leigh shouldering 30%. The reasons for such division are purely equitable, as Bell occupied the Site for the longest period of time and Sequa gained access to the Site with knowledge that chromium contamination was a problem and measures to correct the contamination were necessary." John Leigh accrued the least financial gain from his chromium-plating venture and has cooperated at every juncture with the Government in the government's efforts to discover the sources behind the chrome contamination.").

The adoption of the Chem-Dyne approach in Part IV(B) of the majority opinion precludes such equitable apportionment except as part of a contribution claim proceeding. I agree with the majority's embrace of the Chem-Dyne approach. But I think we should address the impact of the Leigh and Bell consent decrees upon Sequa's SARA-bestowed contribution rights--in light of the alternative, equitable divisibility determinations already rendered by the district court after its "Phase III" hearing on the issue of the "relative contributions of Bell, Sequa and John Leigh to the contamination at the ... Site."3 In my opinion, the following is the appropriate appellate court approach to this case.

First, we should hold that the district court was not clearly erroneous in its finding that Sequa failed to meet its burden of proof on the factual, quantitative apportionment issue--of Sequa's proportionate responsibility for the single chromium harm suffered by the aquifer. Then, we should reject the district court's conclusion that, in this case, it did not need to consider the fairness of the proposed consent decrees relative to Sequa's SARA-bestowed, equitable cost allocation rights. See 42 U.S.C. Sec. 9613(f)(1). I think we must address the impact of the consent decrees on the defendants' statutory equitable cost allocation rights--in light of the alternative, equitable "apportionment" finding reasonably rendered by the district court. This approach is consistent with the caselaw on appropriate contribution analyses.4 And my approach certainly offers a much better prospect for bringing this protracted and expensive litigation to an end than does a remand to the district court for more (essentially redundant) proceedings.

As the majority has noted, after concluding that Sequa had failed to meet its burden of demonstrating a reasonable fact basis for apportionment of the relative responsibilities of the defendants, the district court rendered an alternative, purely equitable "apportionment" determination. In reaching its alternative conclusion, the district court considered the following equitable facts: that Bell occupied the site for the longest period of time; that Sequa gained access to the site with knowledge that chromium contamination was a problem and measures to correct the contamination were necessary; and that Leigh accrued the least financial gain from the chrome plating venture, but cooperated with the government in the government's efforts to discover the sources of the contamination. The district court "apportioned" the defendants' equitable responsibility for costs at 35% each to Bell and Sequa, and 30% to Leigh. While the district court's alternative equitable findings are not articulated as a "contribution claim" adjudication, I would hold that they satisfy the essential requirements of CERCLA Sec. 9613(f)(1), and that they are consistent with this Circuit's decision in Amoco Oil Co. v. Borden, 889 F.2d 664 (5th Cir.1989)--and thus, that they are sufficient to constitute contribution findings based in equity.

In light of the procedural posture of this case and, in particular, in light of the district court's reasonably based, 35%--35%--30%, "purely equitable" "apportionment" findings, the parameters of Sequa's equity rights are plainly such that it would be inequitable and violative of the contribution claims provision of SARA, 42 U.S.C. Sec. 9613(f)(1), for those rights to be destroyed by the Leigh and Bell consent decrees.

Consistent with CERCLA Sec. 9613(f)(1) and Sec. 9613(f)(2), I would hold that when, in a case such as this one, the EPA finds it advantageous to enter into a settlement with jointly and severally liable defendants, thereby shielding the settling defendants from contribution liability (by operation of CERCLA's Sec. 9613(f)(2)), the EPA must bear the risk of its bargain being proved less than satisfying upon district court resolution of a non-settling defendant's, consent decree-attacking, Sec. 9613(f)(1) equitable cost allocation claim. Under the facts of this case, the EPA cannot have it both ways. It cannot enjoy the benefits of joint and several liability and at the same time enter into consent decrees with the otherwise jointly and severally liable defendants to destroy a non-settling defendant's statutory right to an equitable allocation of costs under 42 U.S.C. Sec. 9613(f)(1). The majority's sanctioning of such enjoyment by the government improperly allows the government to smelt what is plainly intended by Congress to be a defendants' rights provision (42 U.S.C. Sec. 9613) into a governmental sword against defendants.

Under the facts of this case, Sequa's equitable cost allocation rights are not limited to 42 U.S.C. Sec. 9613(f)(2), which provision focuses on providing for offset contribution. In short, because Sequa raised its claims for an equitable, proportionate cost allocation ruling in what amounts to a contribution claims proceeding--before the district court embraced the consent decrees shielding Bell and Leigh, under 42 U.S.C. Sec. 9613(f)(2), from contribution liability--Sequa is entitled to invoke the broader equitable response cost allocation remedy contained in 42 U.S.C. Sec. 9613(f)(1). Compare 42 U.S.C. Sec. 9613(f)(1) (emphasis added here) ("Any person may seek contribution from any person who is liable or potentially liable under section 9607(a).... In resolving contribution claims, the court may allocate response costs among liable parties using such equitable factors as the court determines are appropriate."), with 42 U.S.C. Sec. 9613(f)(2) (emphasis added here) ("A person who has resolved its liability to the United States or a State in an administrative or judicially approved settlement shall not be liable for claims for contribution regarding matters addressed in the settlement. Such settlement does not discharge any of the other potentially liable persons unless its terms so provide, but it reduces the potential liability of the others by the amount of the settlement.").5

I further disagree with the majority's treatment of Sequa's challenge to the EPA's decision to provide an alternate water supply system (AWS) to the chromium-affected area as an interim measure pending the completion of final remedial action. We are supposed to uphold the EPA's decision "unless the objecting party can demonstrate, on the administrative record, that the decision was arbitrary and capricious or otherwise not in accordance with law." CERCLA, Sec. 113(f)(2), 42 U.S.C. Sec. 9613(f)(2). I think it is clear that Sequa has again failed to meet its burden of proof.

The scope of review under the "arbitrary and capricious" standard is narrow and a court is not to substitute its judgment for that of the agency. Nevertheless, the agency must examine the relevant data and articulate a satisfactory explanation for its action including a rational connection between the facts found and the choice made.... In reviewing that explanation, we must consider whether the decision was based on a consideration of the relevant factors and whether there has been a clear error of judgment.

Motor Vehicle Manufacturers Ass'n v. State Farm Mutual Automobile Ins. Co., 463 U.S. 29, 43, 103 S. Ct. 2856, 2866-2867, 77 L. Ed. 2d 443 (1983) (citations omitted; internal quotation marks omitted). A year later, in Chevron U.S.A. v. Natural Resources Defense Council, Inc., 467 U.S. 837, 104 S. Ct. 2778, 81 L. Ed. 2d 694 (1984), the Supreme Court further clarified that when a court has determined the intent of Congress is statutorily vague or ambiguous with respect to the specific issue before the court, the court must defer to the "reasonable policy choice" of the agency charged with administering the statute. 467 U.S. at 843-845, 104 S.Ct. at 2782-2783.

State Farm and Chevron make it plain that the courts are not to second-guess the scientific judgments of the EPA. The EPA Administrator may apply his or her expertise to draw conclusions from suspected, but not completely substantiated, relationships between facts, from trends among facts, from theoretical projections from imperfect data, from probative preliminary data not yet certifiable as "fact," and the like. See Ethyl Corp. v. EPA, 541 F.2d 1, 28 (D.C.Cir.) (en banc), cert. denied, 426 U.S. 941, 96 S. Ct. 2662, 49 L. Ed. 2d 394 (1976). And, while deference to agency decisionmaking does not require us to abdicate our judicial duty to carefully review the record in order to ensure that the agency has made a reasoned decision based on reasonable extrapolations from some reliable evidence, the agency's decision need only be reasonable in light of the facts reflected in the administrative record and under the applicable statute(s) and regulations; it need not be the "best" or "most reasonable" decision. See generally Chevron, supra. The majority opinion notwithstanding: in this case, the agency's AWS decision was both a permissible, reasonable reading of the operative CERCLA provisions--and the NCP--under Chevron, and not otherwise arbitrary or capricious under State Farm.

* The Statutory Regime

It will most often be true that the general aims and policies of a controlling statute will be evident from its text. United States v. Gaubert, 499 U.S. 315, ----, 111 S. Ct. 1267, 1274, 113 L. Ed. 2d 335 (1991). It is evident from CERCLA's text that CERCLA's purpose is to enable the executive branch (i.e., the EPA) to target and clean up hazardous waste sites in an efficient manner; and the Superfund amendments of 1986 [SARA] have undoubtedly clarified and strengthened the executive's CERCLA responsibilities and authority. See J.V. Peters & Co., Inc. v. Administrator, EPA, 767 F.2d 263, 264 (6th Cir.1985). In order to effectuate CERCLA's purposes, Congress delegated very broad powers to the EPA--for the agency to reasonably interpret the CERCLA statutory scheme and respond to hazardous substance scenarios in accordance with such interpretation.

In particular, the EPA is under the broad statutory obligation to expeditiously react to any release or threatened release of hazardous substances that may pose harm to the public health, welfare or to the environment. CERCLA, Sec. 101(23), 42 U.S.C. Sec. 9601(23) (emphasis added). There are two types of CERCLA reactions, or responses, envisioned by the statute: (1) removal actions, or interim measures like the AWS at issue in this case; and (2) remedial, or permanent measures. "Removal actions" are defined as actions designed to effect an interim solution to a contamination problem, but very vaguely:

"remove" or "removal" means the cleanup or removal of released hazardous substances from the environment, such actions as may be necessary taken in the event of the threat of release of hazardous substances into the environment, such actions as may be necessary to monitor, assess, and evaluate the release or threat of release of hazardous substances, the disposal of removed material, or the taking of such other actions as may be necessary to prevent, minimize, or mitigate damage to the public health or welfare or to the environment, which may otherwise result from a release or threat of release. The term includes, in addition, without being limited to, security fencing or other measures to limit access, provision of alternate water supplies, temporary evacuation and housing of threatened individuals not otherwise provided for, action taken under section 9604(b) of this title, and any emergency assistance which may be provided under the Disaster Relief and Assistance Act.

CERCLA, Sec. 101(23), 42 U.S.C. Sec. 9601(23) (emphasis added). See also National Contingency Plan, 40 C.F.R. Sec. 300.6 (Definitions) (1986). "Remedial actions" are defined as actions designed to effect a permanent solution to the contamination problem, but they are defined just as vaguely:

"remedy" or "remedial action" means those actions consistent with permanent remedy taken instead of or in addition to removal actions in the event of a release or threatened release of hazardous substance into the environment, to prevent or minimize the release of hazardous substances so that they do not migrate to cause substantial danger to present or future public health or welfare or the environment. The term includes, but is not limited to, such actions at the location of the release as storage, confinement, perimeter protection using dikes, trenches, or ditches, clay cover, neutralization, cleanup of released hazardous substances or contaminated materials, recycling or reuse, diversion, destruction, segregation of reactive wastes, dredging or excavations, repair or replacement of leaking containers, collection of leachate and runoff, onsite treatment or incineration, provision of alternative water supplies, and any monitoring reasonably required to assure that such actions protect the public health and welfare and the environment.

CERCLA, Sec. 101(24), 42 U.S.C. Sec. 9601(24) (emphasis added). See also National Contingency Plan, 40 C.F.R. Sec. 300.6 (Definitions) (1986).

CERCLA provides some much more specific requirements for the EPA's response actions, as well. For example, where groundwater is contaminated by chromium, the maximum contaminant levels allowed by the Safe Drinking Water Act (SDWA), 42 U.S.C. 300f, are "applicable or relevant and appropriate standards," for agency decisionmaking if the groundwater is a potential drinking water supply. 42 U.S.C. Sec. 9621(d)(2)(A). Chromium is one element for which maximum concentration limits ("MCLs") were set under the SDWA. CERCLA nonetheless defines a "potential drinking water supply" quite broadly--as "any raw or finished water source that is or may be used by a public water system * * * or as drinking water by one or more individuals." 42 U.S.C. Sec. 9601(7) (emphasis added).6 Clearly, the agency's determination that the sole source, Trinity Aquifer falls within the statute's drinking water supply definition was not arbitrary or capricious, and reflects a reasonable construction and implementation of the EPA's broad CERCLA enforcement license.

As the majority has stated, the National Contingency Plan (NCP) guides federal and state response activities by specifically identifying methods for investigating the environmental and health problems resulting from a release or threatened release of hazardous substances, and establishing criteria for determining the appropriate extent of response activities. The 1986 NCP was the operative one in this case.

The purpose of the ... (NCP or Plan) ... is to effectuate the response powers and responsibilities created by ... (CERCLA) and the authorities established by section 311 of the Clean Water Act (CWA), as amended.

* * * * * *

(a) The Plan applies to all Federal agencies and this plan is in effect for:

* * * * * *

(2) Releases or substantial threats of releases of hazardous substances into the environment, and releases or substantial threats of releases or pollutants or contaminants which may present an imminent and substantial danger to public health or welfare.

(b) The Plan provides for efficient, coordinated, and effective response to discharges of oil and releases of hazardous substances, pollutants, and contaminants in accordance with the authorities of CERCLA and the CWA.It provides for:

(1) Division and specification of responsibilities among the Federal, State, and local governments in response actions, and appropriate roles for private entities.

NCP, 40 C.F.R. Sec. 300.3 (1986) (emphasis added). In this case, the EPA worked with the Texas Water Commission (TWC), and a private environmental research, or investigatory firm--IT Corporation.

Consistent with CERCLA, the 1986 NCP required that drinking water supplies meet the Safe Drinking Water Act (SDWA) standards for chromium--a statutorily defined, "hazardous substance." 40 C.F.R. Sec. 300.68(i), Appendix V(2). And the NCP defined a "drinking water supply" as "any raw or finished water source that is or may be used by a public water system (as defined by the Safe Drinking Act) or as drinking water by one or more individuals." NCP, 40 C.F.R. Sec. 300.6 (Definitions) (1986) (emphasis added). The 1986 NCP also listed, as an appropriate response "to the threat of direct contact with hazardous substances or pollutants or contaminants," the provision of an alternate water supply "where it will reduce the likelihood of exposure of humans or animals to contaminated water." 40 C.F.R. Sec. 300.65(c)(8) (emphasis added). In short, contrary to the majority opinion, the EPA has not been statutorily or administratively handicapped to act only in an "all or nothing" manner relative to threats of hazardous substance exposure; quite the contrary.

The majority's contentions notwithstanding, the Administrative Record in fact contains substantial evidence that the EPA's provision of an alternate water supply system was not arbitrary or capricious at the time the EPA made its AWS decision.

In accordance with the NCP's fair, established procedures, a study was conducted to examine the alternatives available to accomplish the task of providing safe water to those in the affected area. Based on this study, a determination was made that the best option was to extend the public water supply operated by the adjacent city of Odessa, Texas to the site. Indeed, the determination to provide this alternate water supply to those in the affected area was based on an extensive Administrative Record, including in particular a two-volume Remedial Investigation report and a Record of Decision (which incorporates by reference, among other documents, the Remedial Investigation)--outlining the EPA's reasons for selecting the AWS approach to the threats posed by the chromium-contaminated, sole drinking water source, Trinity Aquifer.

The chromium posed a threat to present and future human life in the area. As already stated, in making its assessments of the situation, the government was compelled by the NCP in effect at the time to follow the standards set in the Safe Drinking Water Act (the "SDWA"), 42 U.S.C. Sec. 300f et seq.; and chromium is one element for which maximum concentration limits ("MCLs") were set under the SDWA. The government found twelve of the fifteen sample wells tested had chromium levels at or above the MCL for chromium. Further, nine of these twelve met or exceeded the higher recommended MCLs for chromium proposed by the EPA in the Federal Register of November 13, 1985. Administrative Record at 3146. And the Remedial Investigation "determined that about thirty (30) people were presently being served by seven (7) wells that produce the groundwater with chromium concentrations above the drinking water standard." Administrative Record at 4015.

The Administrative Record reflects a decision "to provide the residents and businesses in the Superfund Impacted Service Area with an alternate water supply from the City of Odessa (City)." Id. (emphasis added) (also stating that this alternative had "been given conceptual approval by the past City Council and is contingent upon the contract being signed between the City of Odessa and the TWC (Texas Water Commission)"). See also id. ("Concurrently [with the Remedial Investigation and the Feasibility Study for the site], a Focused Feasibility Study (FFS) was conducted and completed in August, 1986, to determine what alternate methods were available to supply these people and surrounding potentially affected areas with a safe drinking water source.").

Sequa complains that businesses using the chromium-contaminated wells ultimately were not "allowed" to participate in the alternate water supply system, and argues that this demonstrates that the decision to implement the AWS was arbitrary and capricious. However, as already noted, and contrary to Sequa's contention, the Administrative Record reveals that businesses were indeed a focus of the AWS decision. The Administrative Record further reflects that businesses could be incorporated into the design and construction of the system if they bore their own administrative costs and burdens.7 Moreover, CERCLA requires the EPA to take measures to minimize threats to public health and the environment, not to ensure elimination of all such threats. Accordingly, the 1986 NCP listed, as an appropriate response "to the threat of direct contact with hazardous substances or pollutants or contaminants," the provision of an alternate water supply--"where it will reduce the likelihood of exposure of humans or animals to contaminated water." 40 C.F.R. Sec. 300.65(c)(8) (1986) (emphasis added). It is obvious that the AWS provided to the area (an area otherwise dependent upon a chromium-contaminated aquifer for its sole source of drinking water) (at least) minimized the likelihood, present and future, of exposure of humans and animals to the contaminated water.8

In sum: the majority has erred in substituting its own, post hoc vision of wise response action judgment for that of the agency. The EPA's determination to install the AWS is the type of technically expert decision to which this Court properly accords "great deference." The agency's interpretations of its broad CERCLA directives were reasonable. The Administrative Record supports the agency's particular AWS determination. And the agency determination is not inconsistent with the NCP. We should uphold the district court's decision to grant summary judgment to the agency on the questions associated with the executive agency decision to provide an alternate water supply to the individuals residing in the chromium-affected area. To so uphold the agency's decisionmaking and action does not amount to "meek deference" to the EPA's scientific expertise, as the majority has asserted. However, the scrutiny to which the majority subjects the agency's AWS decision certainly amounts to much more than the appropriate deferential review of the agency's action called for under the Supreme Court's caselaw concerning agency implementation of federal statutes. See e.g., Chevron, U.S.A., Inc. v. Natural Resources Defense, 467 U.S. 837, 104 S. Ct. 2778, 81 L. Ed. 2d 694 (1984).

All Costs? Yes.

I think we should decide whether the EPA is entitled to recover all of its costs for designing and constructing the AWS. We should decide that it is.

As the majority has stated, CERCLA's Sec. 107 provides for the recovery of the following costs:

(A) all costs of removal or remedial action incurred by the United States Government or a State or an Indian tribe not inconsistent with the national contingency plan;

(B) any other necessary costs incurred by any other person consistent with the national contingency plan;

(C) damages for injury to, destruction of, or loss of natural resources, including the reasonable costs of assessing such injury, destruction, or loss resulting from such a release; and

(D) the costs of any health assessment or health effects study carried out under section 9604(i) of this title.

42 U.S.C. Sec. 9607(a)(4) (emphasis added). I share the majority's serious doubt that Congress intended to give the EPA completely unrestrained spending discretion. But we are bound to pay attention to the fact that, while CERCLA's Sec. 9607(a)(4)(A) provides that the United States is entitled to recover "all costs of removal or remedial action incurred by the United States Government or a State or an Indian tribe not inconsistent with the national contingency plan," other provisions of the statute--dealing with recovery of costs by private parties under CERCLA--state that these private parties are entitled to recover only "reasonable costs" of certain activities. See 42 U.S.C. Sec. 9607(a)(4)(B).

We must presume that when Congress wants to make cost a factor of statutory analysis it knows how to do so. See e.g., Union of Concerned Scientists v. United States Nuclear Regulatory Commission, 824 F.2d 108, 115 (D.C.Cir.1987), and cases cited therein.9 Were we to ascribe no meaning to the distinctions drawn in CERCLA's Sec. 9607, we would be derelict in our duty to pay close heed to both what Congress has said and what Congress has not said in the statute under review. We would be derelict in our duty to interpret the law as plainly written by Congress.

I note also that the congressional intent reflected in CERCLA's Sec. 9607's "all costs" language reasonably reflects a fundamental purpose of CERCLA--to ensure that there be rapid recovery of response costs from polluters, which in turn ensures that the Superfund will be made whole quickly and that the funds recovered can be applied to still other hazardous sites. As the Second Circuit explained recently:

In passing CERCLA Congress faced the unenviable choice of enacting a legislative scheme that would be somewhat unfair to generators of hazardous substances or one that would unfairly burden the taxpaying public. The financial burdens of toxic clean-up had been vastly underestimated--in 1980 when CERCLA was enacted $1.8 billion was thought to be enough. In 1986 when the Superfund Amendments and Reauthorization Act of 1986 (SARA), Pub. L. No. 99-499, 100 Stat. 1613 (1986), was passed, $100 billion was held to be needed. It may well be more today. It is of course the public-at-large that is already bearing the economic brunt of this enormous national problem.

United States v. Alcan Aluminum Corp., 990 F.2d 711, 716-717 (2nd Cir.1993).

Sequa has failed to show that the EPA's AWS action in this case was inconsistent with the NCP guiding EPA responses at the time of the agency's AWS decision and action. Thus, in this case at least, Sequa's "reasonable cost" argument must fail. See United States v. Northeastern Pharmaceutical, 810 F.2d 726, 747-748 (8th Cir.1986) (noting that CERCLA's Sec. 9607(a)(4)(A) does not refer to all reasonable costs, but simply to all costs, and concluding therefore that all costs incurred by the government that are not inconsistent with the NCP are conclusively presumed to be reasonable), cert. denied, 484 U.S. 848, 108 S. Ct. 146, 98 L. Ed. 2d 102 (1987); United States v. Hardage, 982 F.2d 1436, 1443 (10th Cir.1992) ("[a]s long as the government's choice of response action is not inconsistent with the NCP, its costs are presumed to be reasonable and therefore recoverable.").

Finally, I cannot join in the majority's commensuration with Sequa over the imagined "horrible" of unbounded liability for response costs assertedly effectuated by the district court's ruling. As the Second Circuit discussed in Alcan Aluminum Corp., Congress and the courts have constructed a framework of fairness to avoid the majority's feared "lack of limits" to the scope of CERCLA liability. See United States v. Alcan Aluminum Corp., 990 F.2d 711, 721-722 (2nd Cir.1993).10 The majority nonetheless finds particularly frightening Sequa's hypothetical in which the EPA enters into an independent contract with someone to investigate and respond to hazardous waste possibilities, and then collects from the defendant polluters, as among the costs of this response, a Rolls Royce to be provided to the independent contractor as a perk for the latter's good labors. Yet, the majority's fears are simply unfounded. The legal reality is that this sort of grossly-attenuated "horrible" cannot come to pass. As the United States Supreme Court put it, in a similar context:

There are obviously discretionary acts performed by a Government agent that are within the scope of his employment but not within the discretionary function exception [to the Federal Tort Claims Act] because these acts cannot be said to be based on the purposes that the regulatory regime seeks to accomplish. If one of the officials involved in this case drove an automobile on a mission connected with his official duties and negligently collided with another car, the exception would not apply. Although driving requires the constant exercise of discretion, the official's decisions in exercising that discretion can hardly be said to be grounded in regulatory policy.

United States v. Gaubert, 499 U.S. 315, ---- n. 7, 111 S. Ct. 1267, 1275 n. 7, 113 L. Ed. 2d 335 (1991) (emphasis added). Thus, there exist well-established standards whereby the majority's imagined Rolls Royce "horrible" would surely be adjudged arbitrary and capricious. In stark contrast to the Rolls Royce "horrible" constructed by Sequa and the majority, the EPA response action at issue in this case--as reflected in the Administrative Record--is well grounded in CERCLA regulatory policy, not to mention the plain language of the statute, and is not arbitrary and capricious.

Finally, I do not think the district court erred in crediting the consent decree proceeds toward reimbursing the government for the costs incurred in pursuing Bell through bankruptcy proceedings, before allowing the leftover proceeds from the Bell settlement to be credited toward the sum Sequa was left owing the government. Sequa's interpretation of 42 U.S.C. Sec. 9613(f)(2) is unduly narrow and unreasonably strained.11

First, reimbursable "response costs" made the subject of the Bell consent decree are specifically defined in the consent decree, as including enforcement expenses--including in particular, attorneys' fees. And such expenses are recoverable under CERCLA--at least in response cost recovery cases brought by the government. See e.g., United States v. Northernaire Plating Co., 685 F. Supp. 1410, 1418 (W.D.Mich.1988) (recoverable costs include "attorney fees and litigation expenses incurred by the staffs of the EPA and the Department of Justice"), aff'd sub nom. United States v. R.W. Meyer Inc., 889 F.2d 1497 (6th Cir.), cert. denied, 494 U.S. 1057, 110 S. Ct. 1527, 108 L. Ed. 2d 767 (1990); United States v. Northeastern Pharmaceutical, 579 F. Supp. 823, 851-852 (W.D.Mo.1984) (to the same effect), aff'd in part, rev'd in part on other grounds, 810 F.2d 726 (8th Cir.1986), cert. denied, 484 U.S. 848, 108 S. Ct. 146, 98 L. Ed. 2d 102 (1987). But cf. Stanton Road Associates v. Lohrey Enterprises, 984 F.2d 1015 (9th Cir.1993) (holding that private parties are not entitled to recover attorneys' fees and expenses as costs incurred in bringing CERCLA cleanup cost recovery action), with General Electric Co. v. Litton Industrial Automation Systems, Inc., 920 F.2d 1415 (8th Cir.1990) (holding that private parties are entitled to recover attorneys' fees and expenses incurred in bringing CERCLA cleanup cost recovery action), cert. denied, 499 U.S. 937, 111 S. Ct. 1390, 113 L. Ed. 2d 446 (1991).

Moreover, the purpose of the Sec. 9613(f)(2) settlement offset provision is to prevent the government from obtaining "double recoveries" in cases in which joint and several liability has been imposed. See United States v. Northeastern Pharmaceutical, 810 F.2d 726, 748-749 (8th Cir.1986) ("Appellants argue that unless the judgment is offset by the amount of the Syntex settlement, the government will improperly receive a double recovery of that amount from Syntax and the appellants."), cert. denied, 484 U.S. 848, 108 S. Ct. 146, 98 L. Ed. 2d 102 (1987). The district court's treatment of the consent decree proceeds is fully consistent with this anti-double recovery purpose. The district court's ruling, that the initial crediting of the Bell settlement funds must go toward making the government whole with respect to the enforcement expenses it incurred in connection with pursuing Bell in bankruptcy proceedings, does not provide the government with a "double recovery." Indeed, an adoption of Sequa's view of the crediting provision would provide Sequa with a "windfall" at the expense of the Superfund. In short, the result advocated by Sequa is contrary to the most fundamental purposes of CERCLA.

In my opinion, Sequa should be provided its statutory right of equitable response cost allocation under 42 U.S.C. Sec. 9613(f)(1). My approach approves the district court's alternate, equitable "appropriation" as a Sec. 9613(f)(1), equitable response cost allocation. Under my approach, Sequa would be responsible for its 35% share of cost responsibility, and no more--notwithstanding the settlement the government has negotiated with Bell. If that amount would, absent a crediting from the settlement proceeds, result in the government reaping a "double recovery," then the crediting provision should be applied to prevent that result. In such circumstances, the money remaining from the Bell settlement after the application of some of it toward the reimbursement of the government for its enforcement expenses incurred against Bell in bankruptcy proceedings may be credited to the joint and several liability of Sequa. If, on the other hand, the combination of Sequa's equitable allocation of response costs payment plus the leftover settlement (crediting) proceeds still fails to make the Superfund whole, it is my opinion that such is simply the proper consequence of the bargain the government struck in this case; the government must live with its bargain.

While I concur with much of the majority opinion, I must also dissent from much of it. Contrary to congressional intent and traditional judicial doctrines--not to mention the bedrock principle of a prudent separation of federal governmental powers--the majority has become much more than an appellate court in order to reach its rulings in this case regarding quantitative apportionment and the executive branch decision to provide the chromium-affected area with an alternate water supply system. The majority has usurped for itself the special powers of the executive agency and the trial court as well.