Koppers Co., Inc. v. Aetna Casualty, 98 F.3d 1440 (3d Cir. 1996)

OPINION OF THE COURT

STAPLETON, Circuit Judge:

Koppers Company, Inc. (“Koppers”),1 asserts breach of contract and declaratory judgment claims against its liability insurers, based on their denial of coverage for various environmental property damage claims. Koppers entered into settlement agreements with several of its insurers prior to trial, leaving only certain excess liability insurers as litigating defendants. Following the jury’s determinations that the oecurrence-based policies had been triggered and that Koppers had incurred a total of about $70 million in property damage liability, the district court entered judgment for Koppers, holding its excess insurers liable for the full amount of the claim without reducing the verdict to account for Koppers’ settlements with the other insurers.

On appeal, the excess insurers allege four errors under Pennsylvania law: (1) the court erroneously instructed the jury that the occurrence-based policies would be triggered if any property damage occurred during the policy period, even if the initial cause of that damage was an event (such as a chemical spill) that occurred prior to the policy period; (2) the court erroneously instructed the jury that the insurer has the burden of proving that the specific property damage that occurred was not fortuitous but was expected or intended by the insured; (3) the court abused its discretion by excluding proffered evidence of Koppers’ failure to mitigate the damage; and (4) the court erroneously failed to reduce the judgment to account for the plaintiffs settlements with other insurers.2 We believe the district court committed reversible error with respect only to the last claim, and we will accordingly reverse and remand with instructions to reduce the judg

ment to account for the settling insurers’ apportioned shares.

I.

Koppers is a large, diverse manufacturing company based in Pittsburgh, Pennsylvania that has been conducting manufacturing operations in locations throughout the United States since the early part of this century. In the 1980s, federal and state agencies brought claims against Koppers demanding remediation for environmental contamination at some 150 plant sites and disposal sites. This environmental contamination included property damage to third-party soil, subsoil and groundwater. Koppers then sought a defense and indemnification from its various liability insurers, all of whom initially denied coverage.

Appellants, defendants below, are a group of certain underwriters for Lloyd’s of London and certain London market insurance companies (hereinafter the “London Insurers”). Over the years, the London Insurers have issued a number of excess liability insurance policies to Koppers.

Koppers commenced this action in 1985 against two of its primary insurers, and has since amended its complaint several times to add other primary and excess insurers, including the appellant insurers. All of the defendant insurers except for the London Insurers settled with Koppers before trial. Koppers, alleging that the London Insurers breached their contracts of insurance, sought damages and a declaratory judgment regarding Koppers’ right to indemnification under those policies.

The district court limited the scope of the trial to twelve specific policies, which provided multiple layers of occurrence-based, excess liability coverage for third-party property damage. Five of these policies were in effect between late 1953 and January 1957, and the remaining seven policies were in effect between January 1957 and January 1960. Thus, although Koppers had insurance from at least the 1940s through at least the 1970s, the trial was limited to a roughly six-year period. The district court further limited the scope of the trial to only eighteen of the contaminated sites. Finally, the trial was limited to determining liability and damages for cleanup and response costs incurred at these sites through the end of 1993, and to determining whether Koppers was entitled to a declaratory judgment for the period thereafter.

After a three-week trial, the jury found by special verdict that, during the applicable policy periods, (1) an “occurrence” had triggered coverage under all of the policies at issue at each site, and (2) Koppers had neither expected nor intended to cause damage to third-party property at any site. The jury awarded some $70 million in damages, and the court entered judgment in May 1995. Pursuant to the court’s instructions, the jury’s damages figure represents the total costs Koppers had incurred because of third-party property damage at the eighteen focus sites through the end of 1993, without regard to the sums of money Koppers received from the settling insurers. The court explained to the jury that the court would adjust the damages award according to rules of law after the jury determined the total damages figure.

Both parties moved to alter or amend the judgment under Fed.R.Civ.P. 59(e). The district court, in its July 1995 order, granted Koppers’ motion and granted the London Insurers’ motion in part but denied it in part. Specifically, as relevant here, the court: (1) granted Koppers’ motion to reduce the judgment to about $66 million,3 but denied the London Insurers’ motion to reduce the judgment further to account for Koppers’ settlements with the other insurers; (2) granted a declaratory judgment (limited to the twelve policies and eighteen focus sites at issue) that Koppers is entitled to indemnification under the London Insurers’ policies for the period following 1993; and (3) certified the July 1995 order as final under Fed.R.Civ.P. 54(b). This timely appeal followed.

II.

The district court had jurisdiction over this diversity action pursuant to 28 U.S.C. § 1332. We have appellate jurisdiction over the final order pursuant to 28 U.S.C. § 1291.

The parties agree, as do we, that Pennsylvania law governs this case. We review the district court’s interpretation and prediction of state law de novo. Wiley v. State Farm Fire & Cas. Co., 995 F.2d 457, 459 (3d Cir.1993). In adjudicating a case under state law, we are not free to impose our own view of what state law should be; rather, we are to apply existing state law as interpreted by the state’s highest court in an effort to predict how that court would decide the precise legal issues before us. Kowalsky v. Long Beach Township, 72 F.3d 385, 388 (3d Cir.1995). In the absence of guidance from the state’s highest court, we must look to decisions of state intermediate appellate courts, of federal courts interpreting that state’s law, and of other state supreme courts that have addressed the issue. Wiley, 995 F.2d at 459-60. We must also consider “analogous decisions, considered dicta, scholarly works, and any other reliable data tending convincingly to show how the highest court in the state would decide the issue at hand.” McGowan v. University of Scranton, 759 F.2d 287, 291 (3d Cir.1985) (quoting McKenna v. Ortho Pharmaceutical Corp., 622 F.2d 657, 663 (3d Cir.1980)) (internal quotation marks omitted).

A timely appeal from the denial of a Rule 59 motion to alter or amend the judgment brings up the underlying judgment for review, so that our standard of review varies with the underlying judicial decision. Federal Kemper Ins. Co. v. Rauscher, 807 F.2d 345, 348 (3d Cir.1986).

We exercise plenary review in determining whether a jury instruction misstates a legal standard. Savarese v. Agriss, 883 F.2d 1194, 1202 (3d Cir.1989). We consider the jury instructions as a whole to determine whether they fairly and adequately contain the law applicable to the case. See Douglas v. Owens, 50 F.3d 1226, 1233 (3d Cir.1995); Savarese, 883 F.2d at 1202.

“We review pre-trial and trial court rulings concerning the admission of evidence for abuse of discretion,” but “error may not be predicated upon a ruling which admits or excludes evidence unless a substantial right of the party is affected.” Glass v. Philadelphia Elec. Co., 34 F.3d 188, 191 (3d Cir.1994) (internal quotation marks omitted). “[I]f we find non-constitutional error in a civil suit, such error is harmless only if it is highly probable that the error did not affect the outcome of the case.” Id. (internal quotation marks omitted).

III.

A.

We will address first the London Insurers’ argument that the district court gave erroneous jury instructions regarding the trigger of coverage under their policies. The district court instructed the jury that the policies could be triggered in either of two ways:

App. at 2046. The London Insurers argue, however, that their policies could be triggered only by a causative event taking place during the policy period, not by the resulting property damage alone if the causative event occurred pre-policy.

We need not predict how the Supreme Court of Pennsylvania would interpret these particular insurance contracts, however. Koppers introduced uncontroverted evidence that the property damage (mostly groundwater contamination through leaching) was continuous, progressive, and indivisible throughout the relevant policy periods. It also introduced uncontroverted evidence that the causes of the contamination (e.g., leaks, drips, spills, or disposals) existed at each

site during each policy period.4 Based on this evidence, the jury found that each of the policies had been triggered. While such a finding could theoretically have been made under the court’s instruction based solely on the jury’s acceptance of Koppers’ uncontra-dicted evidence regarding the occurrence of property damage, we perceive no basis on which the jury might have accepted that evidence yet rejected Koppers’ uncontrovert-ed evidence regarding the occurrence of the causes of that damage. Accordingly, we find it more than highly probable that the district court’s charge on this point did not affect the outcome of the case. In short, the charge, even if erroneous, was harmless error.

B.

1.

The London Insurers next argue that the court erred in instructing the jury that the insurers had the burden of proving, as an affirmative defense, that the losses were not fortuitous — i.e., that Koppers expected or intended the third-party property damage. The London Insurers argue that, because a plaintiff insured must prove that it is entitled to coverage, and because only fortuitous losses are covered, Koppers had the burden of proving that it neither expected nor intended the harm.

The question of which party bears the burden of proof in a diversity case is a matter of state substantive law. Blair v. Manhattan Life Ins. Co., 692 F.2d 296, 299 (3d Cir.1982). In Pennsylvania, the insured bears the burden of proving facts that bring its claim within the policy's affirmative grant of coverage. See, e.g., Riehl v. Travelers Ins. Co., 772 F.2d 19, 23 (3d Cir.1985). By contrast, the insurer bears the burden of proving the applicability of any exclusions or limitations on coverage, since disclaiming coverage on the basis of an exclusion is an affirmative defense. See, e.g., Aetna Life & Cas. Co. v. Barthelemy, 33 F.3d 189, 194 (3d Cir.1994); Compagnie des Bauxites de Guinee v. Insurance Co. of N. Am., 551 F.Supp. 1239, 1243 (W.D.Pa.1982). Our research has revealed no Pennsylvania case allocating the burden of proof on the fortuity requirement. We nonetheless predict that the Pennsylvania Supreme Court would place the burden on the insurer in this ease.

The terms of the policies do not mention fortuity. The portions of the policies defining the scope of coverage state that the London Insurers will “indemnify the Assured for all sums which the Assured shall be obligated to pay by reason of the liability imposed upon the Assured ... for damages ... on account of ... property damage.” App. at 141.5 There is no requirement in the coverage provisions that the loss be fortuitous or “unexpected and unintended.” Nor do the exclusion provisions of any of the policies expressly exclude losses that are non-fortuitous or “expected and intended.”

In Intermetal Mexicana v. Insurance Co. of N. Am., 866 F.2d 71, 72 (3d Cir.1989), the insurance policy at issue covered “ ‘all risks of direct physical loss or damage from any external cause except hereinafter excluded.’ ” The insurer acknowledged that the loss claimed was within the scope of coverage as so stated but denied coverage because that loss was not fortuitous. The policy exclusions contained no reference to the fortuity of the loss. This court there predicted that the Supreme Court of Pennsylvania would recognize a “judicially created ‘fortuity’ exclusion from coverage.” Id. at 76. Our prediction was based on the generally accepted principle that “every ‘all risk’ contract of insurance contains an unnamed exclusion— the loss must be fortuitous in nature.” Id. at 75. We recognized that this generally recog

nized principle had a “public policy” basis. Id. at 77.

The rationale supporting the generally accepted rule against indemnity for non-fortuitous losses is succinctly explained in Robert E. Keeton & Alan I. Widiss, Insurance Law § 5.3(a), at 476-77 (student ed. 1988):

Consistent with Intermetal Mexicana, we predict that it is the public policy of Pennsylvania not to enforce an insurance coverage contract providing coverage for a non-fortuitous loss. As with exclusions stated in an insurance policy itself, when an insurer relies on public policy to deny coverage of a claim, the insurer must bear the burden. See, e.g., Butterfield v. Giuntoli, 448 Pa.Super. 1, 670 A.2d 646, 654 (1995) (finding no exclusion for punitive damages in insurance policy, burden shifted to insurer to prove that punitive damages are excluded as against public policy in Pennsylvania); Princeton Ins. Co. v. Chunmuang, 292 N.J.Super. 349, 678 A.2d 1143, 1147 (1996) (“Whether the exclusion is based on an expressed provision or on the public policy prohibition of insurance against criminal conduct, the insurer bears a substantial burden of demonstrating that the loss falls outside the scope of coverage.”); Continental Cas. Co. v. Fibreboard Corp., 762 F.Supp. 1368, 1373 (N.D.Cal.1991), appeal dismissed and remanded, 4 F.3d 777 (9th Cir.1993) (burden on insurer to prove damages at issue fall within California public policy exclusion). In particular, if an insurer has issued a policy that on its face covers the loss at issue and seeks to deny coverage on the basis that enforcing the policy as written would offend the public policy against indemnification of non-fortuitous losses, we predict that the Pennsylvania Supreme Court would place on the insurer the burden of proving that the circumstances of the loss were such that coverage would be inconsistent with that public policy. We therefore conclude that the district court did not erroneously allocate the burden of proof as to whether Koppers’ losses were expected and intended.6

2.

In addition to challenging the district court’s fortuity instruction for misallocating the burden of proof, the London Insurers argue that the instruction improperly limited this public policy affirmative defense to situations in which the insured “expected or intended” the specific harms that occurred. They insist that, under Pennsylvania law, coverage is defeated if the insured expected or intended a harm of the same general type as the harm that occurred. Koppers does not contest this proposition; it does dispute, however, that the court’s instructions were inconsistent with this proposition.

We agree that the district court’s instructions on this issue did not limit the fortuity defense to those harms specifically expected or intended by the insured. As the London Insurers point out, the district court instructed the jury: “If you find that Koppers intentionally or knowingly caused property damage for which it seeks coverage, then Koppers may not recover for that specific damage.” App. at 1013. However, the court concluded this instruction by stating: “To lose coverage, Koppers must have intended the same general type of property damage that occurred.” Id. We think that the instructions on this point, when read as a whole, fairly conveyed to the jury what the parties agree is the correct legal standard. See Douglas v. Owens, 50 F.3d 1226, 1233 (3d Cir.1995); Savarese v. Agriss, 883 F.2d 1194, 1202 (3d Cir.1989).

C.

The London Insurers’ third claim is that the district court abused its discretion by excluding their proffered evidence of Koppers’ failure to mitigate the property damage. The insurers argue that, under Pennsylvania law, an insured has an ongoing duty to mitigate its losses, and that Koppers’ recovery must be reduced by the amount of loss which could have been prevented if Koppers had undertaken reasonable efforts to mitigate the property damage.

The district court ruled, as a general matter, that evidence of Koppers’ failure to mitigate damages would be admissible. Although the precise basis for the district court’s decision to exclude the London Insurers’ proffered evidence is not clear from the record, we find no reversible error.

As a matter of general contract law, the Pennsylvania Supreme Court has held that a plaintiffs duty to mitigate its damages arises upon the defendant’s breach of the contract. E.g., Bafile v. Borough of Muncy, 527 Pa. 25, 588 A.2d 462, 464 (1991). The superior court has applied this rule in the context of an insurance contract, holding that, upon the insurer’s breach by refusing to indemnify the insured, the insured has a duty to mitigate its damages. See Forest City Grant Liberty Assocs. v. Genro II, Inc., 438 Pa.Super. 553, 652 A.2d 948, 952 (1995) (holding that insurer, once found to be hable, need not reimburse insured for unnecessary roof repairs). Here, the defendant insurers breached by refusing to indemnify Koppers in the 1980s, but all of the proffered “mitigation” evidence concerned the prior two decades. Even assuming, however, that the Pennsylvania Supreme Court would require an insured to mitigate its damages prior to the insurer’s breach of contract, the district court’s exclusion of the proffered evidence was proper because the evidence was legally insufficient to make out a claim of failure to mitigate damages.

Mitigation is an affirmative defense, so the burden of proving a failure to mitigate is on the defendant. See Williams v. Masters, Mates and Pilots of Am., 384 Pa. 413, 120 A.2d 896, 901 (1956); Ecksel v. Orleans Constr. Co., 360 Pa.Super. 119, 519 A.2d 1021, 1028 (1987). To prove a failure to mitigate, a defendant must show: (1) what reasonable actions the plaintiff ought to have taken, (2) that those actions would have reduced the damages, and (3) the amount by which the damages would have been reduced. See, e.g., Ecksel, 519 A.2d at 1021 (finding defendant’s failure to mitigate defense unproven because defendant did not show how plaintiff could have mitigated damages or how damages were made worse by alleged inaction); State Pub. Sch. Bldg. Auth. v. W.M. Anderson Co., 49 Pa.Cmwlth. 420, 410 A.2d 1329 (1980) (affirming judgment for contractor where breaching school district failed to show that contractor’s work was unreasonable and avoidable); see also New Castle County v. Hartford Accident & Indem. Co., 685 F.Supp. 1321, 1332 (D.Del. 1988) (holding that, where insurance policy contained express mitigation provision, insurer must specify what further injury occurred as result of the insured’s lack of mitigation).

The London Insurers’ evidence would not have satisfied any of these three elements, as it merely purports to show that Koppers was aware of property damage in the 1960s and 1970s but failed to correct the problem until

the 1980s.7 The London Insurers offered no evidence of what reasonable actions Koppers might or ought to have taken,8 no evidence tending to show that any actions would have measurably reduced the harm, and, thus, no evidence from which the jury could have determined how much damages could have been reduced by mitigation. For this reason, the court’s exclusion of the London Insurers’ evidence was not reversible error.

D.

The London Insurers’ final claim is that the district court erroneously saddled them with liability for the entire loss by improperly failing to reduce the judgment to account for Koppers’ settlements with the other defendant insurers. We agree.

1.

In J.H. France Refractories Co. v. Allstate Insurance Co., 534 Pa. 29, 626 A2d 502 (1993), the Pennsylvania Supreme Court was called upon to allocate coverage responsibility among six primary insurers, each of which was obligated to cover an indivisible loss. J.H. France had brought a declaratory judgment action to determine its insurers’ duty to defend and indemnify against lawsuits involving asbestos-related bodily injury claims. The insurance policies at issue were all pre-1986 comprehensive general liability (“CGL”) policies which would be triggered by the occurrence during the policy period of “bodily injury.” Affirming the lower court’s application of the “multiple-trigger” or “continuous-trigger” theory of determining liability of the insurers, under which all phases of the disease — exposure, progression, and manifestation — independently constitute “bodily injury” triggering coverage, the supreme court held that “every insurer which was on the risk at any time during the development of a claimant’s asbestos-related disease has an obligation to indemnify J.H. France.” Id. 626 A.2d at 507.

The next question, naturally, was “how to allocate the liability of each insurer when, as is commonly the case, more than one insurer was on the risk at one time or another during the development of a claimant’s disease.” Id. The superior court had held that the several insurers must share the obligation to indemnify on a pro rata basis, apportioned upon the amount of time each policy was in effect. Id. Declining to adopt that approach, the supreme court held instead that the insurers whose coverage had been triggered were jointly and severally liable for the full amount of the claim up to policy limits, and that the insured was entitled to select the policy or policies under which it would be indemnified. Id. at 508. As the supreme court explained:

Id. at 509. The supreme court added, however, that its holding “does not alter the rules of contribution or the provisions of ‘other insurance’ clauses in the applicable policies,” so that an insurer who is saddled with more than its fair share of liability may seek to obtain “a share of indemnification or defense costs from other insurers.” Id.9

In the instant case, the district court predicted that the Pennsylvania Supreme Court would extend its holding on the allocation issue in J.H. France to cases involving environmental property damage claims, such that all insurers whose pohcies were triggered to cover an indivisible loss would be jointly and severally Hable, up to pohcy limits, for the full amount of that loss. Although we conclude that J.H. France does not completely control the disposition of this case, we agree as a general matter with this prediction.

The Pennsylvania Supreme Court’s reasons for adopting the joint and several approach in J.H. France are fully apphcable to the ease before us. “First, and most compelling, is the language of the pohcies themselves.” Id. at 507. In J.H. France, “[e]ach insurer obhgated itself to ‘pay on behalf of the Insured all sums which the Insured shah become legally obhgated to pay as damages because of bodily injury to which this insurance applies.’” Id. (emphasis in original). Thus, as the supreme court explained, “each insurer contracted to pay all sums which the insured becomes legally obhgated to pay, not merely some pro rata portion thereof.” Id. (emphasis in original).

Similarly, the pohcy at issue here obhgated the London Insurers “to indemnify the Assured for all sums which the Assured shall be obhgated to pay by reason of the liability imposed upon the Assured ... for damages ... on account of ... property damage.” App. at 141 (emphasis added).10 We accordingly beheve the Pennsylvania Supreme Court would interpret the London Insurers’ pohcies here, as it did the pohcies at issue in J.H. France, to “cover [the insured’s] entire liability once they are triggered.” 626 A.2d at 508 (quoting Keene Corp. v. Insurance Co. of N. Am., 667 F.2d 1034, 1048 (D.C.Cir.981), cert. denied, 455 U.S. 1007, 102 S.Ct. 1644, 1645, 71 L.Ed.2d 875 (1982)) (alteration in original omitted); see ACandS, Inc. v. Aetna Cas. & Sur. Co., 764 F.2d 968, 974 (3d Cir.1985) (predicting Pennsylvania Supreme Court would interpret “the phrase ‘ah sums’ to provide for full coverage once coverage under a pohcy was triggered”).

The second reason the supreme court gave for adopting the joint and several approach in J.H. France is the indivisibility of asbestos-related bodily injury. No medical evidence was presented to substantiate the assumption, implicit in the superior court’s allocation method of pro rata by time on the risk, “that the progression of asbestos-related disease is linear in character.” 626 A.2d at 508. The supreme court elaborated: “To apportion liability among the insurers on a strictly temporal basis in direct proportion to the length of time each insurer was on the risk, ... notwithstanding its surface attractiveness, assumes a linearity of disease progression which this record does not support.” Id.

As with asbestos-related bodily injury, environmental property damage is a progressive harm that, as a practical matter, is indivisible. See, e.g., New Castle County v. Continental Cos. Co., 725 F.Supp. 800, 811-12 (D.Del.1989) (concluding “it would be impossible in this case to determine when the

first molecule of contaminant damaged neighboring property, or at what rate the contamination spread”), aff'd and rev’d in part on other grounds, 933 F.2d 1162 (3d Cir.1991); Abraham, supra, at 120 (“Given the progressive nature of the environmental harms in question, finding the facts necessary [to hold each policy liable only for those harms that occurred during the policy period] usually would be administratively difficult, scientifically impossible, or both.”); John H. Mathias, Jr. et al., Allocation: J.H. France and the Insureds’ Right to Select from Multiple-Triggered Policies, Coverage (A.B.A.Sec.Litig.), Mar.-Apr.1994, at 19-20 (“Like the progression of asbestosis or other insidious diseases, the migrations of contaminants over years or decades may result not from a predictable, linear process, but rather from sporadic or periodic events or conditions that vary in magnitude and frequency over the years.”). Accordingly, it is appropriate to hold the triggered policies jointly and severally liable here, given that “because each has been triggered to provide coverage against liability for a single indivisible injury, there is no basis for apportioning responsibility among them.” Abraham, supra, at 120-21.

Finally, the supreme court in J.H. France relied on the fact that, according to the policies’ terms, each of the several policies triggered to cover a specific claim was potentially liable for the entire claim. The policies provided insurance against “occurrences” and, as defined in the policies, an “occurrence” included “ ‘continuous or repeated exposure to conditions which result in bodily injury.’ ” 626 A.2d at 508. Thus, the entire “process of exposure [would] constitute one occurrence.” Id. The supreme court explained:

Id.; see also ACandS, 764 F.2d at 974 (noting that, “if a plaintiffs damages are caused in part during an insured period, it is irrelevant to [the insured’s] legal obligations and, therefore, to the insurer’s liability that they were also caused, in part, during another period”).

Like the CGL policies in J.H. France, the London Insurers’ policies provided occurrence-based coverage11 and defined an “occurrence” to mean either “a series of occurrences arising out of one event,” App. at 158, or “one happening or series of happenings, arising out of or due to one event taking place during the term of this policy.” App. at 149. Under either definition, then, all of the effects resulting from a single causative event are considered a single occurrence. As in J.H. France, because the entire injury is defined as one “occurrence,” a triggered policy must indemnify the insured for all damages resulting from that injury.12

For these reasons, we predict that if the Pennsylvania Supreme Court were faced with this case it would apply the J.H. France allocation approach, holding jointly and severally liable all policies triggered to cover a single, indivisible loss. We note that other courts have similarly taken the joint and

several approach where multiple policies cover an indivisible loss. See, e.g., Keene Corp. v. Insurance Co. of N. Am,., 667 F.2d 1034, 1047-50 (D.C.Cir.1981), cert. denied, 455 U.S. 1007, 102 S.Ct. 1644, 1645, 71 L.Ed.2d 875 (1982); Dayton Indep. Sch. Dist. v. National Gypsum Co., 682 F.Supp. 1403, 1410-11 (E.D.Tex.1988), rev’d on other grounds sub nom. W.R. Grace & Co. v. Continental Cas. Co., 896 F.2d 865 (5th Cir.1990).

2.

Given this prediction, if this case, like J.H. France, involved several triggered primary policies where none of the insurers had settled, J.H. France would dictate our resolution of this appeal. But this is not such a case. We must consider several complicating factors that were not present in J.H. France, namely, the presence of excess insurers, the insured’s settlements with some primary and some excess insurers, and the district court’s decision to limit the trial to a certain period of years at certain sites. We believe that, if presented with the complicating factors in our ease, the Pennsylvania Supreme Court would not simply apply J.H. France and hold that the appellant excess insurers are jointly and severally liable, up to policy limits, for the entire property damage liability at the focus sites. Instead, as explained below, we predict that the supreme court would modify the J.H. France rule to hold the London Insurers jointly and severally liable for the amount of the loss in excess of the settling insurers’ pro rata shares of liability.

We begin with the principle of indemnity, a fundamental principle of insurance law which prohibits insurance contracts from conferring a benefit greater than the insured’s loss (i.e., a “double recovery”). See, e.g., J.H. France, 626 A.2d at 508 (stating principle that insured “cannot collect more than it owes in damages”) (quoting Keene, 667 F.2d at 1050); Keeton & Widiss, supra § 3.1(a), at 135. We must apply this principle of indemnity — barring recoveries greater than losses — in conjunction with the J.H. France rule — holding that where multiple insurance policies are triggered to cover an indivisible loss, each insurer may be called upon to cover the entire loss up to policy limits. J.H. France does not, of course, hold that an insured, having recovered part of its loss from one insurer, can recover an amount equal to its entire loss from another.

Because we cannot permit a double recovery, and because several insurers have already paid money to Koppers in complete settlement of Koppers’ claims against them, we must either (1) reduce the judgment to account for the settling insurers’ apportioned shares of liability, or (2) permit the non-settling insurers to seek contribution from the settling insurers and, in turn, permit the settling insurers to seek reimbursement from Koppers. We predict that the Pennsylvania Supreme Court would choose the former rule: reducing the judgment to account for the settling insurers’ apportioned shares of liability. That is, we predict that the supreme court would adopt the “apportioned share set-off rule.”13

Although the Pennsylvania Supreme Court has not had occasion to decide how to allocate coverage responsibility where some of the defendant insurers have settled, the Pennsylvania Superior Court has reached this issue and has resolved it by applying the apportioned share set-off rule. In Gould, Inc. v. Continental Cas. Co., 401 Pa.Super. 219, 585 A.2d 16 (1991), Gould, the insured, sued several insurers seeking indemnification for sums paid in settlement of its employees’ claims of injury from exposure to toxic metals and chemicals in Gould’s lead smelting facility. One of the insurers, National Union, joined an additional insurer, Travelers, claiming entitlement for contribution. Because Travelers had previously paid money to Gould in settlement of Gould’s claims for indemnification, however, the trial court granted an order dismissing the cross-complaint against Travelers. National Union appealed from that order.

The superior court affirmed. The court held that, where two insurers are obligated to cover the same loss and one insurer settles but one does not, the litigating insurer cannot seek contribution from the settling insurer. Id. 585 A.2d at 19 (“Travelers is not ‘contributorily’ liable to National Union in this action.”). Rather, the litigating insurer “will be entitled to prorate the amount of coverage” based on the settling insurer’s proportionate share of coverage responsibility as determined by the terms of the two policies, especially the “other insurance” provisions. Id. The court stated that the proration of a litigating insurer’s liability will not be impaired by the dismissal of any settling insurers because “the applicable limits and pro rata shares pertaining to the policies issued by [the settling insurers] will be examined and their proportionate share of liability determined, even if they are no longer parties to the action.” Id. (internal quotation marks omitted; alteration in original).

Had the Gould decision been issued by the supreme court, it would be controlling. Absent some reason to believe that the supreme court would reach a different result, the superior court’s holding is entitled to great deference in our endeavor to predict state law. See, e.g., Rolick v. Collins Pine Co., 925 F.2d 661, 664 (3d Cir.1991) (predicting Pennsylvania law, and stating that we cannot disregard decision of intermediate appellate court unless we are convinced that state’s highest court would establish different rule).

The superior court’s decision in Gould, insofar as it requires the amount for which the litigating insurers are liable to be reduced by the settling insurers’ apportioned shares, is in no way inconsistent with J.H. France, which involved no settling insurers. Moreover, the Pennsylvania Supreme Court has had occasion to decide the effect of a plaintiffs settlement with fewer than all jointly and severally liable defendants outside of the environmental liability insurance context, and there the supreme court has adopted the apportioned share set-off rule, rejecting any right of contribution against settling defendants because such an action would defeat the finality of the settlement. See Charles v. Giant Eagle Mkbs., 513 Pa. 474, 522 A.2d 1, 2-3 (1987) (holding, in context of joint and several tort liability, verdict amount against litigating defendants shall be reduced by amount of settling defendants’ apportioned share of liability, regardless of amount received by plaintiff in settlement). The decisions in Charles and Gould are readily applicable in a ease in all respects like J.H. France except for the existence of some settling insurers; there must be a set-off, rather than a subsequent contribution action,14 to avoid a double recovery.15

We accordingly predict that, if presented with our case, the Pennsylvania Supreme Court would hold that each non-settling insurer whose policy was triggered to cover an indivisible loss is jointly and severally liable, up to the limits of its policy, for the full amount of the judgment, less the settling insurers’ apportioned shares.

3.

Having predicted how the supreme court would modify its holding in J.H. France to

accommodate the problem of the settling insurers, we note that there are two additional rules of Pennsylvania insurance law that must guide our resolution of the case at hand. These rules are relevant because, unlike J.H. France, this case involves excess as well as primary policies.

First, a true excess or secondary policy is not “triggered” or required to pay until the underlying primary coverage has been exhausted. See, e.g., Occidental Fire & Cos. Co. v. Brocious, 772 F.2d 47, 53-54 (3d Cir.1985). This remains true even where the primary insurer would have paid to exhaustion but for its bankruptcy: in Pennsylvania, an excess insurer is not required to “drop down” to provide primary coverage if the underlying primary insurer is insolvent. See Donegal Mut. Ins. Co. v. Long, 528 Pa. 295, 597 A.2d 1124, 1127-28 (1991) (deciding to “refuse to transform an excess carrier into a primary carrier”); J. Kinderman & Sons, Inc. v. United Nat’l Ins. Co., 406 Pa.Super. 37, 593 A.2d 857, 860 (1991) (holding that excess policy is responsible only for amounts exceeding underlying policy’s limits), aff'd, 533 Pa. 87, 619 A.2d 1058 (1993).

Second, if the underlying primary insurer is solvent but the policyholder settles its claim against that primary insurer for less than policy limits, we predict the Pennsylvania Supreme Court would adopt the widely-followed rule that the policyholder may recover on the excess policy for a proven loss to the extent it exceeds the primary policy’s limits. See Barry R. Ostrager & Thomas R. Newman, Handbook on Insurance Coverage Disputes § 13.04, at 575-77 (7th ed. 1994) (citing, inter alia, Stargatt v. Fidelity & Cas. Co., 67 F.R.D. 689, 691 (D.Del.1975) (opining that to require insured “actually to collect the full amount of the [primary] policies ... in order to ‘exhaust’ that insurance ... seems unnecessarily stringent”), aff'd, 578 F.2d 1375 (3d Cir.1978)). That is, settlement with the primary insurer functionally “exhausts” primary coverage and therefore triggers the excess policy — though by settling the policyholder loses any right to coverage of the difference between the settlement amount and the primary policy's limits. The excess insurer cannot be made liable for any part of this difference because the excess insurer never agreed to pay for losses below a specified floor (i.e., below the limits of the underlying primary policies). Courts have adopted this rule because it encourages settlement and allows the insured to obtain the benefit of its bargain with the excess insurer, while at the same time preventing the insured from obtaining a double recovery.

Accordingly, because all of the London Insurers’ policies provided layers of excess liability coverage over certain specified, underlying policy limits, no London Insurer policy would be triggered until the underlying coverage has been “exhausted,” either by settlement or by payment. With respect to this exhaustion requirement, the London Insurers argue that all applicable primary coverage must be exhausted before any excess insurer will be obligated to pay. This argument is predicated on the policies’ “other insurance” clauses, which state essentially that all other available insurance must be exhausted first. Under J.H. France, however, a policy which promises to pay “all sums” must provide for fall coverage once triggered, without regard for such “other insurance” clauses. 626 A.2d at 507. The court held that it was irrelevant whether other policies were also triggered, concluding that, “The insurer in question must bear potential liability for the entire claim.” Id. at 508. Here, the London Insurers agreed to pay “all sums”16 in excess of the specified limits of the directly underlying policies. Once the directly underlying coverage has been exhausted, then, each excess policy must indemnify the insured for the full excess loss up to policy limits. Under J.H. France, the insured gets indemnified first (pursuant to the insuring agreements) and then the insurers may seek to redistribute the burden among themselves. It is only at this latter stage that the “other insurance” clauses become relevant, so the London Insurers’ exhaustion argument based on the “other insurance” clauses must be rejected.

In sum, taking all of the above rules together, we predict that the Pennsylvania Supreme Court would hold that the non-settling excess insurers are jointly and severally liable for the full amount of the loss in excess of: the sum of (1) the policy limits of the directly underlying, “exhausted” primary policies, and (2) the combined pro rata shares of other settling (primary and excess) insurers. The beneficent consequences of this formula are that the insured bears the risk of settling too low while the non-settling insurers bear the risk of being unable to redistribute equitably among themselves the burden of paying the balance (if, for example, some of their number are insolvent).

4.

During the period litigated at trial in this case, 1953-1960, there were two primary policies directly underlying the twelve excess policies issued by the London Insurers. There were no other policies involved in that period. Prior to trial, Koppers settled with the insurer that had issued both primary policies. Accordingly, coverage under these two primary policies has been “exhausted”— regardless of the amount Koppers received in settlement — and the London Insurers’ excess policies are obligated to pay.

If Koppers had not settled with any other insurers (insurers which had issued policies outside of the 1953-1960 litigation period), that would be the end of the matter: the judgment would be reduced simply by the combined limits of the two underlying primary policies, and the London Insurers would be jointly and severally liable for the balance (although they would be free subsequently to seek contribution from the other insurers, all of whom would be non-settlers), But our ease is not so simple: Koppers set-tied with several other primary and excess insurers that had issued policies in effect both before and after the litigation period. On remand, the district court must therefore apply the apportioned share set-off rule for these settled policies as well.

Determining the apportioned share of a settling insurer requires consideration of the “other insurance” clauses of the policies found to cover the loss and the applicable state law governing the interpretation of those clauses and the resolution of any conflicts among them.17 See Gould, 585 A.2d at 19; see also Ostrager & Newman, supra, §§ 11.01-11.04 (discussing allocation of coverage responsibility and resolution of conflicts involving “other insurance” clauses). The record before us does not contain the “other insurance” clauses from all potentially applicable policies. Moreover, because of the district court’s limitation of the period covered by the trial, all policies that may be found to cover the indivisible loss have not yet been identified. In these circumstances, we cannot determine what the apportioned shares of the settling insurers will be. While we leave that determination for the district court in the first instance, we offer two observations to assist the district court in its task.

Our first observation relates to the primary policies that do not directly underlie the London Insurers’ excess policies (i.e., those that cover a period outside of 1953-1960).18 We have today held that the London Insurers’ liability under their excess policies was triggered as soon as the two directly

underlying primary policies were settled, and that the existence of other primary policies applicable to the indivisible loss was irrelevant for the purpose of resolving the threshold issue of whether the London Insurers’ policies were triggered. This does not necessarily mean, however, that the existence of other primary policies applicable to the indivisible loss is irrelevant for the purpose of determining the extent of the London Insurers’ liability under their excess policies. The “other insurance” clauses of the relevant policies and the Pennsylvania law applicable thereto may require that, as between a primary policy and an excess policy triggered to cover the same loss, the primary policy must pay first and, accordingly, that the apportioned share of a settled primary policy covering the same indivisible loss is its full policy limits. See, e.g., Ostrager & Newman, supra, § 11.03[e][l]. Thus, the district court may be required to deduct from the total loss the combined limits of all settled primary policies.

Our second observation relates to the settled excess policies. Koppers settled with some, but not all, of the excess insurers that issued policies in effect outside of the litigation period. Therefore, there are some non-settled, excess policies for which the threshold triggering question has not been answered.

It may be that, under the applicable law, the apportioned share of a settling excess insurer — and, accordingly, the extent of the London Insurers’ liability' — cannot be determined without identifying all policies that are triggered and cover the indivisible loss, whether they were in force during or outside the litigation period.19 This will be true, for example, if the applicable rule of allocation among excess policies here is found to be a pro rata allocation based on the limits of each policy and the total limits of all triggered policies. See Gould, 585 A.2d at 19; Ostrager & Newman, supra, § 11.04. Under this rule of allocation and the peculiar circumstances of our case, however, the district court would not need to determine whether the non-settling pre-1971 policies were triggered because the London Insurers concede — against their interests — that all of Koppers’ policies up to 1971 (the date from which pollution exclusion clauses have appeared in all the policies) were triggered. Reply Br. at 31 (noting that Koppers “admits” this fact).20 With respect to the post-1970 policies, however, the district court may have to decide whether those policies were triggered before it can properly determine the settling insurers’ apportioned shares.21

IV.

For the foregoing reasons, we will reverse and remand for the sole purpose of allowing the district court to mold the verdict to take account of the settling insurers’ apportioned shares of liability.