Five years ago we held that a settlement with the EPA did not foreclose claims for contribution among firms that sent solvents for reprocessing to Fisher- Calo Chemicals, which handled them poorly. Akzo Coatings, Inc. v. Aigner Corp., 30 F.3d 761 (7th Cir. 1994). Fisher-Calo went out of business, leaving its former customers to bear the burden of cleanup under the Comprehensive Environmental Response, Compensation, and Liability Act of 1980 (CERCLA). After the EPA put the Fisher-Calo site on the National Priorities List, Akzo performed some cleanup work in a portion of the site that it concedes received waste for which it is responsible. Aigner did (and continues to do) the bulk of the work at Fisher-Calo's premises as a whole. ("Akzo" is shorthand for Akzo and O'Brien, the plaintiffs; "Aigner" stands in for approximately 50 additional firms.) Akzo demanded that Aigner contribute to the cost of cleanup in Akzo's portion of the site; Aigner made the same demand of Akzo. Akzo's principal argument is that its responsibility is limited to a small portion of the site, so that it need not bear any cleanup costs for the rest. But the district court rejected this position--initially on summary judgment, ruling that Fisher-Calo's business premises are a single site for CERCLA purposes, see 960 F. Supp. 1354 (N.D. Ind. 1996), then, after a bench trial, finding that it is not possible to trace pollutants from different sources to their destinations within the Fisher-Calo complex. The judge ordered Akzo to reimburse Aigner for 12.56% of the total costs that Aigner will incur in performing cleanup work, plus a cash award of about $1.5 million, equal to that portion of costs incurred by the time of trial less what Akzo had spent already. Akzo asks us to hold that it owes nothing, and that if it must pay anything its liability is limited to 9% of costs, because it sent only 9% of the total volume of solvents that Fisher-Calo processed.

Akzo's principal argument for "nothing" is that wastes from its solvents were deposited at discrete locations within the Fisher-Calo site, that it is therefore possible to ascertain the precise injury inflicted by these wastes. Akzo directs most of its fire to the district court's conclusion on summary judgment that the Fisher- Calo business premises must be deemed a single site (as the EPA has treated it), making a search for distinct harms within that site unnecessary. But that was not the end of the matter; after a trial the judge concluded that it is not possible to identify distinct harms, and this factual conclusion makes it unnecessary (and inappropriate) for us to inquire what the judge should have done at an earlier stage of the case, when the record contained less information. See Watson v. Amedco Steel, Inc., 29 F.3d 274 (7th Cir. 1994). Although Akzo contests the district judge's conclusion that pollutants derived from Akzo's solvents were scattered throughout the Fisher-Calo site, the judge's findings are not clearly erroneous.

Even if none of Akzo's wastes can be traced to one of the many parcels within the Fisher-Calo site (the parcel known as Space Leasing), it does not follow that the judge was obliged to carve that parcel out when calculating cleanup costs and responsibility. Section 113(f)(1) of CERCLA, 42 U.S.C. sec.9613(f)(1), says that "[i]n resolving contribution claims, the court may allocate response costs among liable parties using such equitable factors as the court determines are appropriate." Allocation of shares thus is not a mechanical process. Doubtless liability for distinct or divisible harms should be kept separate when possible, see Restatement (2d) of Torts sec.sec. 433A, 881, but the district judge held that at the Fisher-Calo site a reliable division is just not possible. Fisher-Calo's records are so sparse and uninformative, and the costs of matching inputs to outputs so high, that compiling a comprehensive inventory would not be sensible, the judge concluded. Akzo loses to the extent it must contribute toward the costs of cleaning up the Space Leasing parcel, but the same inability to trace contributions means that Akzo gains to the extent that other firms must contribute toward the costs of cleaning up the parcels that contain waste attributable to Akzo. How the balance of advantage comes out is impossible to say for the Fisher-Calo site as a whole, the district judge found, and this assessment does not represent an abuse of discretion.

Neither did the district judge err in handling Akzo's contention that the residue of its solvents is less toxic than the residue of Aigner's solvents. Akzo wanted the district judge to apportion costs according to a toxicity index. The volume of solvents each party shipped to Fisher-Calo would be multiplied by a factor reflecting the toxicity of the ensuing wastes and liability apportioned according to the result. For example, if Aigner shipped 100,000 gallons of wastes with a toxicity factor of 2, and Akzo shipped 100,000 gallons with a toxicity factor of 1, then Aigner would pay two-thirds of the total cleanup cost. Because sec.113(f)(1) tells the court to use "equitable factors," it is possible that toxicity could play a role in the allocation if it is connected to cost--whether the financial cost of the particular cleanup or the total social cost of the pollution. Suppose that Aigner's wastes were twice as toxic as Akzo's but equally costly (pound for pound or gallon for gallon) to remove from the ground, and no more dangerous to strangers after the cleanup had been completed. Then there would be no sound reason to measure contribution by toxicity rather than by the expense of doing the work. Even the word "toxic" may mislead; sometimes that word means deadly (cyanide is more toxic than polychlorinated biphenyls in this sense), but sometimes it means "hard to purge from other substances" (PCBS are more toxic than cyanide in this sense). Substances that are very poisonous may be simple to eliminate or dilute to a harmless level; substances that cling tenaciously to dirt, water, and living tissues may be very costly to clean up, not only because it is hard to get rid of them but because their toxicity leads to buildup over the years in exposed humans, so that the environmentally safe level is lower than that for acutely poisonous substances. Akzo has not disentangled these different meanings of toxic or demonstrated that the residuals in its solvents are less costly to clean up.

Actually, even proof that its solvents are easier and cheaper to deal with on their own is not enough. A court must ask how each party's wastes affected the total cost of cleanup. Suppose that Akzo and Aigner dumped 100 gallons of solvents in the same pit, that Aigner's chemicals (alone) would have cost $500 to clean up, that Akzo's (alone) would have cost $400, and that the two together cost $600. The marginal cost of Akzo's solvents then is $100, and the marginal cost of Aigner's $200; each party must pay at least that much. But had Aigner sent no solvents to this pit, Akzo would have had to pay $400 in cleanup costs. All an appellate court could say is that Akzo's share in contribution must be between $100 and $400, while Aigner's must be between $200 and $500; any allocation that adds to $600 and respects these constraints cannot be condemned. Browning-Ferris Industries of Illinois, Inc. v. Ter Maat, 195 F.3d 953, 956-59 (7th Cir. Nov. 1, 1999), explains this at greater length. In particular, for this example, requiring each responsible party to contribute according to the number of gallons it shipped to the site would not violate the constraint that neither party pay less than the increase in marginal cost, nor more than the total cost it would have borne had it been the only polluter. Akzo does not contend that its contribution share exceeds the cost that would have been necessary to clean up the pollutants attributable to its solvents if it had been Fisher-Calo's sole customer. The district court's decision to give every gallon the same weight therefore was within its discretion, even on the assumption that Aigner's wastes were more toxic per gallon than Akzo's.

Having decided that all gallons of solvents shipped to Fisher-Calo count equally, the court then ordered Akzo to pay approximately one third more than equal-weighting implies. Akzo generated approximately 9% by volume of all solvents that Fisher-Calo processed, but the court ordered Akzo to reimburse Aigner for approximately 13% of the costs that Aigner has incurred or will expend in completing the cleanup. Aigner and the other firms in its consortium sent about 71% of the total volume of solvents to the Fisher-Calo site, so Akzo's shipments are approximately 13% of the Akzo + Aigner total; other shipments and shippers were ignored because they are not parties to this suit.

According to the district court, this outcome is required by the Uniform Comparative Fault Act. The district court read sec.2 and sec.6 of the UCFA to provide that the responsibility of non- parties must be disregarded, even if they are financially able to pay (indeed, even if they already have paid) their share of the cleanup. To take a simple example, suppose Firm A is responsible for 40% of the pollutants, Firm B for 10%, and Firm C for 50%. Firm A agrees with the EPA to perform the cleanup and sues B for contribution. On the district court's reading of the UCFA, B must pay 20% of the total cleanup costs, because B sent 20% of the pollutants that A and B generated jointly. That C is able to pay its 50% share--indeed, that C has already paid 50%, and that the outcome of the suit between A and B will leave A bearing only 30% of the total costs--is irrelevant on the district court's (and Aigner's) understanding of the UCFA. A polluter that agreed to clean up a Superfund site could turn a tidy profit if this were so. Suppose that ten firms, A through J, sent 10% each, and that A, having agreed to do the cleanup work, sues B for contribution. Firms A and B are responsible for equal volumes of wastes, so the court would order B to pay 50% of the total cleanup costs. Next A sues C and recovers another 50%. If all of Firms B through J were good for the judgments, then A would recover 450% of its total outlay for pollution control. Even if the court set a cap of 100%, to prevent A from making a profit, the upshot would be that of ten equally responsible polluters, B and C would pay 50% each, and the other eight would pay nothing. That is not a sensible outcome of a process that is supposed to yield an "equitable" allocation of expenses.

Akzo contends that the UCFA requires the district court to undertake a global assessment of responsibility, so that Akzo cannot be required to pay anything until every shipper's share has been determined. That might take years of trial time. Akzo would be happy to skip the trial and chip in 9%, but that would leave Aigner holding the bag if the other shippers were unable to pay their shares. Aigner, for its part, contends that the UCFA requires courts to ignore non-parties just as the district court concluded. None of these approaches is sound. Akzo's would either complicate an already difficult allocation process or saddle firms such as Aigner with excess costs. The Supreme Court has cautioned against the adoption of any contribution rule that would complicate litigation. McDermott, Inc. v. AmClyde, 511 U.S. 202, 211-14 (1994); see also Texas Industries, Inc. v. Radcliff Materials, Inc., 451 U.S. 630 (1981); Northwest Airlines, Inc. v. Transport Workers, 451 U.S. 77 (1981). Aigner's approach would lead to disproportionate liability, with contribution shares turning not on actual responsibility (or on the actual collections of the party performing the cleanup) but on litigation strategy. It is of course possible that both sides are wrong, and that the UCFA requires the inclusion of either pollution shares of, or the actual recoveries from, the additional parties with which Aigner has reached settlements. But we resist all temptation to give the UCFA a close reading--and this despite the fact that Akzo and Aigner agreed in the district court that it supplies the rule of decision. Section 113(f)(1) provides otherwise: "Such claims [for contribution] shall be brought in accordance with this section and the Federal Rules of Civil Procedure, and shall be governed by Federal law." The UCFA is not a federal law, and we are not bound by the parties' agreement to an inapplicable body of legal rules. Kamen v. Kemper Financial Services, Inc., 500 U.S. 90, 99 (1991); United States National Bank of Oregon v. Independent Insurance Agents of America, Inc., 508 U.S. 439, 445-48 (1993).

Although federal law governs, it is possible and often desirable to borrow a rule from state law, when the alternative is judicial invention. United States v. Kimbell Foods, Inc., 440 U.S. 715 (1979); Atherton v. FDIC, 519 U.S. 213 (1997); but see Burlington Industries, Inc. v. Ellerth, 524 U.S. 742, 754-55 (1998). The reference to "Federal law" in sec.113(f)(1) implies that the law should be nationally uniform, rather than varying according to each state's idea of appropriate contribution. Yet the UCFA would not be an attractive national rule. Unlike the Uniform Commercial Code, to which the Court turned in Kimbell Foods, the UCFA has not been adopted throughout the United States. Only two states (Iowa and Washington) have enacted the UCFA; eleven have adopted the Uniform Contribution Among Tortfeasors Act, which supplies a different approach to contribution; the rest resolve contribution issues through the common law or non-uniform statutes. When one of the litigants has settled with a third party, the UCFA reduces other shares by the percentage of total fault of the person released in the settlement (UCFA sec.2); this is the source of Akzo's contention that the district court must hold a comprehensive trial to determine every shipper's share of liability. The UCATA, by contrast, reduces liability only by the dollar amount of third- party settlements (UCATA sec.4). These competing approaches can produce substantial differences in incentives to settle and in the complexity of litigation. See McDermott, 511 U.S. at 214-17; In re Oil Spill by the Amoco Cadiz, 954 F.2d 1279, 1314-18 (7th Cir. 1992); Steven Shavell, Economic Analysis of Accident Law 164-67 (1987); Lewis A. Kornhauser & Richard L. Revesz, Sharing Damages Among Multiple Tortfeasors, 98 Yale L.J. 831 (1989). To the extent language in sec.113 speaks to the issue, it prefers the approach of the UCATA: A settlement with the United States or a state "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." Section 113(f)(2), 42 U.S.C. sec.9613(f)(2). Adopting the UCFA as a federal rule would undermine that decision.

As a proposition of federal law, the district court's approach has nothing to recommend it, for it produces disparities in liability when third parties have settled. We assume, with the district court, that the proportionate share of the parties is a good starting point. Suppose that Akzo and Aigner were the only two financially sound parties responsible for the pollution. Then the final contribution shares properly would reflect only their relative responsibility, and the 13% share for Akzo would stand. But they are not. Aigner has settled with some other firms that sent solvents to Fisher- Calo and with some past owner-operators of portions of the site. It has claims pending against still more potentially responsible parties. It is very unlikely that 13% is an accurate estimate of Akzo's share among the financially sound firms that will eventually chip in.

McDermott considered at length the proper treatment of settling parties under a body of federal law that includes contribution--the law of admiralty. The Court deemed two approaches "closely matched" (511 U.S. at 217): claim reduction (also known as "proportionate share"), see Restatement (2d) of Torts sec.886A Comment m(3), and reduction pro tanto by the actual amounts recovered in settlements, see Restatement sec.886A Comment m(2). The claim-reduction approach requires the court to determine the responsibility of all firms that have settled, as well as those still involved in the litigation, and to ignore any firms that have not settled. To return to the ten-firm hypothetical above, if Firm A had settled with C and D, then their shares of responsibility (and the corresponding cleanup costs) would be excluded from the calculation. In a contribution suit between A and B, Firm B would be ordered to pay 50% of the costs after excluding C and D, or 40% of the total costs of restoring the site. It would not matter how much A actually had recovered from C and D. By contrast, under the pro tanto approach, anything A recovered from C and D in settlement would be deducted from the total cleanup costs, and the court would order A and B to bear the remaining costs equally. In McDermott the Court adopted claim reduction, deeming it most compatible with the way related issues in admiralty have been handled.

If as McDermott explained the choice between the pro tanto approach and claim reduction is a tossup, 511 U.S. at 217, then it is best to match the handling of settlements with the way intersecting principles of law work. For admiralty that meant claim reduction. For CERCLA the most closely related rule of law is sec.113(f)(2), which reduces third-party claims by the actual cash value of settlements reached with governmental bodies. Extending the pro tanto approach of sec.113(f)(2) to claims under sec.113(f)(1) enables the district court to avoid what could be a complex and unproductive inquiry into the responsibility of missing parties. The extended litigation between Akzo and Aigner well illustrates the difficulties of fixing responsibility for wastes sent years (if not decades) ago to a firm that did not keep good records and contaminated a wide area. Excluding only actual collections from third parties enables the court to conserve its resources.

On remand, the district court should determine how much Aigner has collected from third parties in settlement, then require Akzo to pay 12.56% of the costs net of those recoveries, rather than of Aigner's total outlay. The total must be reduced not only by collections Aigner has realized to date, but also by future third-party payments. Phrasing Akzo's liability as "12.56% of the cleanup cost net of third-party collections" or some similar formula will avoid any need to reopen the judgment under Fed. R. Civ. P. 60(b)(5) to account for the outcome of litigation now pending or to be filed in the future.

If some of Aigner's settlements provide for percentage-of-cost payments rather than cash payments, then the district court should exclude that percentage from the pool. (To this extent the pro tanto approach works like claim reduction, but without the need for the court to determine the responsibility of the settling parties.) Even if, as Akzo believes, Aigner settled for too little with any of these third parties, it is not free to bring its own contribution actions against them. McDermott labeled "clearly inferior" the possibility of collecting more from parties who reached private settlements in good faith. 511 U.S. at 211. A potentially responsible party (PRP in CERCLA jargon) that wants to guard against inadequate collections from third parties must either intervene in the suits against them or challenge the bona fides of the settlements immediately after they are reached. Id. at 212-14, 114 S. Ct. 1461.

The judgment is affirmed to the extent it holds Akzo responsible for contribution toward the cleanup costs of the Fisher-Calo site as a whole, and to the extent it adopts an approach treating each gallon of solvents as equally responsible for cleanup costs. The judgment is vacated to the extent it quantifies Akzo's contribution liability, and the case is remanded for further proceedings consistent with this opinion.