COFFIN, Senior Circuit Judge.
In July of 1977, the Picillos agreed to allow part of their pig farm in Coventry, Rhode Island to be used as a disposal site for drummed and bulk waste. That decision proved to be disastrous. Thousands of barrels of hazardous waste were dumped on the farm, culminating later that year in a monstrous fire ripping through the site. In 1979, the state and the Environmental Protection Agency (EPA) jointly undertook to clean up the area. What they found, in the words of the district court, were 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." O'Neil v. Picillo, 682 F. Supp. 706, 709, 725 (D.R.I. 1988).
This case involves the State of Rhode Island's attempt to recover the clean-up costs it incurred between 1979 and 1982 and to hold responsible parties liable for all future costs associated with the site.1 The state's complaint originally named thirty-five defendants, all but five of whom eventually entered into settlements totalling $5.8 million, the money to be shared by the state and EPA. After a month-long bench trial, the district court, in a thorough and well reasoned opinion, found three of the remaining five companies jointly and severally liable under section 107 of the Comprehensive Environmental Response, Compensation, and Liability Act of 1980, 42 U.S.C. § 9601 et seq. ("CERCLA") for all of the State's past clean-up costs not covered by settlement agreements, as well as for all costs that may become necessary in the future. The other two defendants obtained judgments in their favor, the court concluding that the state had failed to prove that the waste attributed to those companies was "hazardous," as that term is defined under the Act.
Two of the three companies held liable at trial, American Cyanamid and Rohm and Haas, have taken this appeal. Both are so called "generators" of waste, as opposed to transporters or site owners. See § 107(a)(3), 42 U.S.C. § 9607. Neither takes issue with the district court's finding that some of their waste made its way to the Picillo site. Rather, they contend that their contribution to the disaster was insubstantial and that it was, therefore, unfair to hold them jointly and severally liable for all of the state's past expenses not covered by settlements. They further contend that it was error to hold them liable for all future remedial work because the state has not demonstrated that such work ever will be necessary. With far less vigor, they also raise a series of equitable defenses, claiming that their liability should be reduced, either in whole or part, because (1) much of the damage to the site resulted from the government's sloppy handling of barrels; (2) the government's clean-up procedures were not cost-efficient as required by the Act, see § 107(a); and (3) their waste ended up at the site through the acts of wholly unrelated third parties, see § 107(b)(3). Finally, they argue that the Act should not be applied retroactively and that it was inappropriate for the district court to award the government prejudgment interest in this case. After a careful review of the record, we conclude that none of these arguments suffices to warrant reversal of the judgment below.
We need not spend time outlining the general factual and statutory background of this case. As we noted above, the district court's opinion is thorough and we see no reason to repeat what was said there. See 682 F. Supp. 706. Nor do we think it necessary to respond in detail to all of appellants' claims, as we could add little to the district court's reasoning or existing case law. We therefore confine our discussion to appellants' arguments concerning the unfairness of holding them jointly and severally liable for the government's past and future clean-up costs.2
Joint and Several Liability
It is by now well settled that Congress intended that the federal courts develop a uniform approach governing the use of joint and several liability in CERCLA actions. The rule adopted by the majority of courts, and the one we adopt, is based on the Restatement (Second) of Torts: damages should be apportioned only if the defendant can demonstrate that the harm is divisible. see, e.g., United States v. Chem-Dyne Corp., 572 F. Supp. 802, 809-11 (S.D.Ohio 1983); United States v. Monsanto Co., 858 F.2d 160, 171-73 (4th Cir. 1988); United States v. Bliss, 667 F. Supp. 1298, 1312-13 (E.D.Mo. 1987).
The practical effect of placing the burden on defendants has been that responsible parties rarely escape joint and several liability, courts regularly finding that where wastes of varying (and unknown) degrees of toxicity and migratory potential commingle, it simply is impossible to determine the amount of environmental harm caused by each party. See, e.g., United States v. Chem-Dyne, 572 F. Supp. at 811; Monsanto, 858 F.2d at 172-73. It has not gone unnoticed that holding defendants jointly and severally liable in such situations may often result in defendants paying for more than their share of the harm. Cf. United States v. Monsanto, 858 F.2d. at 173. Nevertheless, courts have continued to impose joint and several liability on a regular basis, reasoning that where all of the contributing causes cannot fairly be traced, Congress intended for those proven at least partially culpable to bear the cost of the uncertainty. see, e.g., United States v. Chem-Dyne, 572 F. Supp. at 809-810.
In enacting the Superfund Amendments and Reauthorization Act of 1986 ("SARA"), Congress had occasion to examine this case law. Rather than add a provision dealing explicitly with joint and several liability, it chose to leave the issue with the courts, to be resolved as it had been -- on a case by case basis according to the predominant "divisibility" rule first enunciated by the Chem - Dyne court. See e.g., United States v. Monsanto, 858 F.2d. at 171 n.23 (Chem-Dyne decision endorsed by Congress); Cf. Garber, Federal Common Law of Contribution Under the 1986 CERCLA Amendments, 14 Eco.L.Q. 365, 374-75 (1987). Congress did, however, add two important provisions designed to mitigate the harshness of joint and several liability. First, the 1986 Amendments direct the EPA to offer early settlements to defendants who the Agency believes are responsible for only a small portion of the harm, so-called de minimis settlements. See § 122(g).3 Second, the Amendments provide for a statutory cause of action in Contribution, codifying what most courts had concluded was implicit in the 1980 Act. See § 113(f)(1). Under this section, courts "may allocate response costs among liable parties using such equitable factors as the court determines are appropriate." We note that appellants already have initiated a contribution action against seven parties before the same district court judge who heard this case.
While a right of contribution undoubtedly softens the blow where parties cannot prove that the harm is divisible, it is not a complete panacea since it frequently will be difficult for defendants to locate a sufficient number of additional, solvent parties. Moreover, there are significant transaction costs involved in bringing other responsible parties to court. If it were possible to locate all responsible parties and to do so with little cost, the issue of joint and several liability obviously would be of only marginal significance. We, therefore, must examine carefully appellants' claim that they have met their burden of showing that the harm in this case is divisible.4
The district court issued two rulings on joint and several liability. First, the court held appellants jointly and severally liable for all of the state's past costs not covered by settlements, roughly $1.4 million including prejudgment interest. According to appellants, this money was spent exclusively on removal" costs or "surface cleanup" (e.g., sampling the waste, contacting responsible parties, and ultimately, removing the barrels and contaminated soil), and not on remedying the alleged damage to groundwater and other natural resources ("remedial" costs).5 Second, the district court held appellants jointly and severally liable for all future removal costs to be incurred by the state, as well as for all cost-efficient remedial action the state (and EPA) may deem necessary after conducting further tests. The parties discuss the two holdings separately and we shall do likewise.
I. Past Costs
Appellants begin by stressing that the state's past costs involved only surface cleanup. They then argue that because it was possible to determine how many barrels of waste they contributed to the site, it is also possible to determine what proportion of the state's removal expenses are attributable to each of them simply by estimating the cost of excavating a single barrel. The EPA advances two reasons why this approach is incorrect. First, it claims that it was not possible to determine how many barrels were traceable to appellants, nor was it possible to determine how much of the contaminated soil removed by the state was attributable to each appellant, and therefore, that it is impossible to apportion the state's removal costs. Second, it argues that even if it were possible to determine what proportion of the state's removal costs are attributable to appellants, joint and several liability still would have been proper because the "harm to be apportioned is not the cost but the environmental contamination that prompts the response action." We shall discuss the EPA's two arguments in reverse order.6
We state at the outset that we have some trouble with the EPA's second argument. Assuming the government ultimately undertakes remedial action to clean the groundwater in the area and then seeks to recover the costs of doing so, it will have in effect submitted two separate bills, one for the cost of removing the barrels and soil, and one for cleaning the water. We think it likely that the harm to the water will be indivisible, and therefore, that appellants could properly be held jointly and severally liable for the cost of this remedial action. But simply because the costs associated with cleaning the groundwater cannot be apportioned does not mean that we should decline to apportion the costs of removing the barrels and soil if those costs are in fact divisible. This would seem to follow from the basic common law principle that defendants not be held responsible for those costs traceable to others. We think that the EPA would have to accept as much. Nonetheless, the Agency adheres to the position that it is irrelevant whether or not the costs of removal can be apportioned.
The reason the Agency takes this position is not, then, because the environmental harm that actually occurred was indivisible, but because the additional environmental harm that the government averted would have been indivisible had it occurred. This argument gives us pause because it appears to contravene the basic tort law principle that one pays only for the harm that was, and not for the harm that might have been.
Assume that it costs the government $1 million to remove all of the barrels from a site, but of this million, only $300,000 were spent removing the defendant's barrels. Also assume that had the barrels not been removed, the additional damage to the environment would have been $5 million and that this five million would not have been divisible. The government certainly would not take the position that it could recover $5 million in such a situation. Instead, it would ask only for the $1 million that it actually spent. Yet when it comes to apportioning that million, the Agency argues that we should look to whether the $5 million of averted harm would be divisible.
If we were to accept the EPA's "averted harm" argument, it appears that apportionment would be appropriate only in the highly unlikely event that (1) all of the barrels were empty and no further environmental harm was possible; (2) the individual barrels were sufficiently far apart that even if further spillage occurred, there would be no commingling of wastes and thus no difficulty determining whose waste caused what damage; or (3) every barrel contained precisely the same type of waste so that even if there was further spillage and commingling, the environmental harm could be apportioned according to the volumetric contribution of each defendant. As the EPA undoubtedly recognizes, it rarely, if ever, will be the case that one of these three conditions is present. As a practical matter, then, joint and several liability will be imposed in every case.
Because we believe Congress did not intend for joint and several liability to be imposed without exception, we are troubled by the practical implications of the Agency's arguments, the more so because it seems to find no support in common law tort principles, which were to be one of our benchmarks in developing a uniform approach to govern the imposition of joint and several liability. At oral argument, the Agency did not claim, however, that its theory fit within the common law framework of joint and several liability, but instead, took the position that these CERCLA cases are not standard tort suits. Although we recognize that Congress deviated from certain tort principles, see New York v. Shore Realty Corp., 759 F.2d 1032, 1044 (2nd Cir. 1985), we had thought that on the issue of joint and several liability we were to take our lead from evolving principles of common law. It would seem incumbent upon the Agency, then, to demonstrate that on this particular question of joint and several liability, Congress intended for us to abandon the common law.
Having said all that, we choose not to resolve the issue in this case. Had appellants met their burden of showing that the costs actually incurred by the state were capable of apportionment, we would have had no choice but to address the EPA's theory. But because we do not believe appellants have done so, we can, and do, choose to leave the question for another day. We turn now to the EPA's first contention that the state's removal costs are not capable of apportionment.
The state's removal efforts proceeded in four phases (0-3), each phase corresponding roughly to the cleanup of a different trench. The trenches were located in different areas of the site, but neither party has told us the distance between trenches. Appellants contend that it is possible to apportion the state's removal costs because there was evidence detailing (1) the total number of barrels excavated in each phase, (2) the number of barrels in each phase attributable to them, and (3) the total cost associated with each phase. In support of their argument, they point us to a few portions of the record, but for the most part are content to rest on statements in the district court's opinion. Specifically, appellants point to the following two sentences in the opinion: (1) "I find that [American Cyanamid] is responsible for ten drums of toxic hazardous material found at the site;" and (2) as to Rohm and Haas, "I accept the state's estimate [of 49 drums and 303 five-gallon pails]." Appellants then add, without opposition from the government, that the ten barrels of American Cyanamid waste discussed by the district court were found exclusively in Phase II, and that the 303 pails and 49 drums of Rohm and Haas waste mentioned by the court were found exclusively in Phase III. They conclude, therefore, that American Cyanamid should bear only a minute percentage of the $995,697.30 expended by the state during Phase II in excavating approximately 4,500 barrels and no share of the other phases, and that Rohm and Haas should be accountable for only a small portion of the $58,237 spent during Phase III in removing roughly 3,300 barrels and no share of the other phases. We disagree.
The district court's statements concerning the waste attributable to each appellant were based on the testimony of John Leo, an engineer hired by the state to oversee the cleanup. We have reviewed Mr. Leo's testimony carefully. Having done so, we think it inescapably clear that the district court did not mean to suggest that appellants had contributed only 49 and 10 barrels respectively, but rather, that those amounts were all that could be positively attributed to appellants.
Mr. Leo testified that out of the approximately 10,000 barrels that were excavated during the four phases, only "three to four hundred of the drums contained markings which could potentially be traced." This is not surprising considering that there had been an enormous fire at the site, that the barrels had been exposed to the elements for a number of years, and that a substantial amount of liquid waste had leaked and eaten away at the outsides of the barrels. Mr. Leo also testified that it was not simply the absence of legible markings that prevented the state from identifying the overwhelming majority of barrels, but also the danger involved in handling the barrels. Ironically, it was appellants themselves who, in an effort to induce Mr. Leo to lower his estimate of the number of barrels attributable to each defendant, elicited much of the testimony concerning the impossibility of accurately identifying all of the waste.7
In light of the fact that most of the waste could not be identified, and that the appellants, and not the government, had the burden to account for all of this uncertainty, we think it plain that the district court did not err in holding them jointly and severally liable for the state's past removal costs. Perhaps in this situation the only way appellants could have demonstrated that they were limited contributors would have been to present specific evidence documenting the whereabouts of their waste at all times after it left their facilities.8 But far from doing so, appellants deny all knowledge of how their waste made its way to the site.9 Moreover, the government presented evidence that much of Rohm and Haas' waste found at the site came from its laboratory in Spring House, Pennsylvania and that during the relevant years, this lab generated over two thousand drums of waste, all of which were consigned to a single transporter.10 Under these circumstances, where Rohm and Haas was entrusting substantial amounts of waste to a single transporter who ultimately proved unreliable, we simply cannot conclude, absent evidence to the contrary, that only a handful of the 2,000 or more barrels reached the site.11
II. Future liability
The district court held appellants jointly and severally liable for all further removal costs taken by the state, as well as for all necessary remedial actions. Appellants have two principal objections. First, they claim that it was error to hold them responsible for the removal of certain piles of soil because the settling parties had agreed to undertake this cost. The state represents to us that these parties have now taken care of the piles and we therefore find the issue moot.
Second, appellants contend that it was improper to hold them liable for future remedial action because the state has not shown that such work will ever be needed. They do not claim, however, that if remedial action is shown to be necessary, it would be a mistake to assume that their waste contributed to the damage. We see no problem with the court giving the state (and EPA) time to conduct further tests. If after conducting the necessary tests, the government concludes that there was in fact no harm to the area's groundwater, then appellants will have nothing to worry about. Moreover, the district court ruled that under section 107 of the Act, the state may take only such measures as are cost-efficient. Appellants, therefore, will have an opportunity to challenge the state's chosen remedial measures at the appropriate time.
Appellants have argued ably that they should not have been held jointly and severally liable. In the end, however, we think they have not satisfied the stringent burden placed on them by Congress. As to all other issues, we affirm substantially for the reasons set out by the district court.12 Appellants should now move on to their contribution action where their burden will be reduced and the district court will be free to allocate responsibility according to any combination of equitable factors it deems appropriate. Indeed, there might be no reason for the district court to place any burden on appellants. If the defendants in that action also cannot demonstrate that they were limited contributors, it is not apparent why all of the parties could not be held jointly and severally liable. However, we leave this judgment to the district court. See, e.g., Developments, Toxic Waste Litigation, 99 Harv.L.Rev. 1458, 1535-43 (1986).
Costs to appellees.
* Of the District of Puerto Rico, sitting by designation.
1. The EPA is not a party to this action, but has filed an amicus brief. For the sake of simplicity, we will refer collectively to the State and the EPA as the "government." Where it is necessary to distinguish between the two, we will do so.
2. During the course of the trial, the district court resolved numerous statutory questions concerning CERCLA's scope. We, of course, take no position on any ruling not contested on appeal.
3. Appellants apparently were offered settlements, but chose instead to try this case.
4. Before we can turn to the divisibility issue, however, we must first resolve a threshold question raised by appellants. Citing the Restatement (Second) of Torts § 433B, appellants contend that before joint and several liability may be imposed, the government has the initial burden of showing that the defendants were a "substantial" cause of the harm. If the government cannot prove that the defendants were substantial contributors, then joint and several liability may not be imposed and the defendant's burden of demonstrating that the harm is divisible never arises. We reject this approach. Even assuming that a strict application of the Restatement rule would allow appellants to escape joint and several liability in a situation such as this -- something which is far from clear, appellants having provided us with no common law tort precedents to support their reading of the Restatement formulation -- we would nonetheless decline to place this threshold burden on the government in CERCLA actions. As we noted earlier, Congress intended for the federal courts to develop a uniform approach to govern the use of joint and several liability. The Restatement is one source for us to consult. While courts generally have looked to the Restatement for guidance, they have declined to place the burden of showing that defendants are "substantial" contributors on the government, recognizing Congress' concern that cleanup efforts not be held hostage to the time-consuming and almost impossible task of tracing all of the waste found at a dump site. See, e.g., United States v. Chem-Dyne, 572 F. Supp. at 809-11; United States v. Monsanto, 858 F.2d at 171-73. As we also noted earlier, in passing the 1986 Amendments Congress chose not to dismantle the existing approach, but instead, to add provisions dealing with de minimis settlements and contribution actions. It is at these stages, then, that the question of "substantiality" should be considered, and not at the point of determining liability to the government.
5. Appellant's entire argument rests on this point. We assume, therefore, that if the $1.4 million in past expenses did cover more than "removal" costs, the government would have alerted us to this fact.
6. The state essentially advances the same two arguments.
7. Appellants contend that the state's record keeping was subpar. Mr. Leo testified that because a cleanup effort of this magnitude had never before been undertaken in this country, it was natural that the state made mistakes. He also intimated, however, that there appeared to be little excuse for the state's complete failure to keep records during Phase I. We also are troubled by the state's poor effort during this stage of the cleanup. Absent a finding that the state acted in bad faith, however, we do not believe this is a sufficient reason to reverse the district court. In the context of this case, the state's failure to document its work during Phase I was harmless error since Mr. Leo testified that even when the state made an effort to identify the barrels, it could rarely do so.
8. Appellants also might have proven that they were limited contributors by documenting that they had generated only a small amount of waste and then had ceased operations.
9. CERCLA is a strict liability scheme and it was therefore possible to hold appellants liable without such knowledge.
10. Similar evidence was presented about American Cyanamid.
11. Even if it were possible to determine how many barrels each appellant contributed to the site, we still would have difficulty concluding that the state's removal costs were capable of apportionment. To apportion the cost of removing the over 10,000 barrels, we would have to know the cost of removing a single barrel. Appellants have proceeded on the assumption that the cost of removing barrels did not vary depending on their content. This assumption appears untenable given the fact that the state had to take added precautions in dealing with certain particularly dangerous substances, including those traceable to Rohm and Haas. Moreover, in addition to excavating barrels the state had to remove large amounts of soil. Because there was substantial commingling of wastes, we think that any attempt to apportion the costs incurred by the state in removing the contaminated soil would necessarily be arbitrary.
12. One of these other issues concerns the constitutionality of applying CERCLA retroactively. The district court held that CERCLA may be applied to pre-enactment conduct. Appellants contend, however, that even if the statute may be applied to pre-enactment conduct, it may not be applied constitutionally to pre-enactment costs incurred by the government, in this case the costs associated with Phases 0 and 1. We disagree and find persuasive the Eighth Circuit's reasoning in United States v. Northeastern Pharmaceutical & Chemical Co., 810 F.2d 726, 734-37 (8th Cir. 1986) (citing cases).