Stonewall Insurance v. Asbestos Claims Management Corp.

Contents

Glossary.1186

Background.1187

I. Asbestos Claims Against NGC and NGC’s Insurance Policies.1187

A. Bodily Injury and Property Damage Claims.1187

B. NGC’s Insurance Program.1187

C. The ACF and CCR Claims-Handling Facilities .1188

D. NGC’s Chapter 11 Bankruptcy Reorganization.1189

II. Procedural Background and Rulings.1189

A. Pretrial Proceedings. oo 1 — 1 T — I

B. Jury Trial and Bench Trial Proceedings a 1 — I T — I

C. Judgments and Permission to Appeal.. a T — i T-t

Discussion.1191

I.Summary of Holdings.1191

II. Issues Related to Coverage for Bodily Injury Claims.1192

A.Bodily Injury Trigger of Coverage.1192

1. Policy Language .1192

2. NGC’s and Insurers’ Positions .1192

3. Jury and Bench Trial Decisions .1193

4. New York and Texas Case Law.1194

5. Evidence on Etiology of Asbestos Diseases .1197

(a) Asbestosis and Pleural Plaques t-tH t-H

(b) Asbestos-Induced Cancers — 00 tH tH

6. Coverage for Injury Resulting from Exposure to Other Manufacturers’ Products.1200

7. Estoppel From Applying Standard Trigger of Coverage to Certain 1983-85 Excess Policies.1201

B. Scope of Responsibility for Claims Triggering Multiple Policies.1201

C. “Expected or Intended” Injuries.1204

D. The Wellington and CCR Issues.1206

III. Issues Related to Coverage for Property Damage Claims.1208

A. Existence of Property Damage. tO o CO

B. Property Damage Trigger of Coverage. tO o CD

C. “Business Risk” Property Damage Exclusions tO f-* O

D. Number of Occurrences. tO ^ tO

IV. Issues Related to Coverage for All Claims.1214

A. “Known Loss” Defense.1214

B. Aggregate Policy Limits of Short-Term Policies .1216

C. Excess Policy Defense Obligations .1218

D. Insurers’ Peremptory Jury Challenges.1219

Conclusion 1219

Glossary

ACF.Asbestos Claims Facility

ACIC.American Centennial Insurance Co.

ACMC .Asbestos Claims Management Corp.

ACMs.Asbestos-Containing Materials

AHP.American Home Products Corp. v. Liberty Mutual Insurance Co., 748 F.2d 760 (2d Cir.1984)

AMICO.American Motorists Insurance Co.

BI.Bodily Injury

CCC.Continental Casualty Co.

CCR.Center for Claims Resolution

CGL.Comprehensive General Liability

CSIS.The Constitution State Insurance Co.

CU .Commercial Union Insurance Co.

NGC.National Gypsum Co.

OR.Old Republic Insurance Co.

PD.Property Damage

Before: NEWMAN, Chief Judge, WINTER and MAHONEY, Circuit Judges.

JON 0. NEWMAN, Chief Judge:

These consolidated appeals and cross-appeals present numerous issues concerning liability insurance coverage in the context of claims for personal injury and property damage arising from exposure to asbestos. The principal issue is to determine the relevant time period or periods for which liability insurance coverage is available to a former asbestos product manufacturer confronted with thousands of asbestos-related claims, where the policies at issue are triggered not by the assertion of a claim against the insured but by the occurrence of bodily injury or property damage during the policy period.

National Gypsum Company (“NGC”), now Asbestos Claims Management Corporation (“ACMC”),¹ and a number of its liability insurers (collectively referred to as “the Insurers”) sought declaratory relief clarifying the extent to which NGC is entitled to indemnification for claims arising from NGC’s manufacture of asbestos products. On appeal, the Insurers (appellants/cross-appellees) and NGC (appellee/eross-appellant) challenge several partial declaratory judgments entered on March 30, 1993, by the District Court for the Southern District of New York (John S. Martin, Jr., Judge), relating to the triggering of policies, the apportionment of coverage, the types of claims covered, the defenses and exclusions available to the Insurers, the number of deductibles NGC must pay on property damage claims, the defense obligations of the Insurers, and the amount of aggregate limits available to NGC on excess policies.²

As discussed below, the standard form language of the insurance policies appears to have been drafted in the expectation that it would usually be applied to the ordinary injury where accident and resulting harm take place almost simultaneously, a circumstance normally presenting no difficulty in determining when an injury has occurred. As this case demonstrates, however, substantial issues of interpretation arise where the policies are sought to be applied to injuries of a progressive nature, which may not fully develop or become manifest until years after exposure to the injury-causing substance.³ We must determine, based on the language of the policies, the case law of New York and Texas, and the medical evidence, what constitutes personal injury or property damage in the asbestos context sufficient to trigger coverage.

In addition, we must consider various other issues raised by the District Court’s rulings, including whether coverage, where available, should be allocated between the Insurers and NGC with respect to periods during which NGC was uninsured; whether the costs of removing or replacing asbestos products installed in buildings constitute “property damage” under the policies; and how many deductibles NGC must absorb, pursuant to “per occurrence” deductible provisions in the policies, before being indemnified for property damage claims.

Background

I. Asbestos Claims Against NGC and NGC’s Insurance Policies

A. Bodily Injury and Property Damage Claims

NGC was founded in Buffalo, New York, in 1925 and became a leading manufacturer of gypsum wallboard and other building materials. From 1930 until 1981, NGC manufactured construction products that contained asbestos, including acoustical plasters, joint compounds, textures, ceiling tiles, asbestos-cement siding, and asbestos-cement corrugated and flat-sheét products. NGC discontinued the sale of asbestos-containing products over the period from 1970-81.

Since 1972, NGC has been sued by approximately 100,000 claimants seeking damages for bodily injury allegedly resulting from exposure to and inhalation of asbestos fibers contained in products manufactured, sold, installed, or distributed by NGC at some time in the past. These claimants typically allege that they did not become aware of their injuries until shortly before they filed suit. They contend that they have suffered from a wide range of injuries and diseases, including mesothelioma and asbestosis.

In addition, since 1980, the owners of several thousand buildings have asserted asbestos-related claims against NGC. These suits are founded upon allegations that the incorporation and continued presence of asbestos-containing materials (“ACMs”) in buildings has caused physical damage to the buildings and tangible property therein as a result of contamination by asbestos fibers. Such contamination results from the continued breakdown of the ACMs, which releases fibers into the air, and the re-entrainment of these fibers, which creates an unreasonable health hazard. Typically, the complainants seek monetary damages measured by the cost of testing and evaluating ACMs in buildings; the cost of repairing, removing, enclosing, encapsulating, or abating the ACMs; the cost of operations and maintenance programs; consequential damages for loss of use of the properties; and diminution of value.

B. NGC’s Insurance Program

This action involves all the liability insurance policies, both primary and excess, issued through 1985 to provide coverage to NGC for asbestos-induced bodily injury and property damage claims, except policies issued by those insurers with whom NGC has settled its coverage disputes. The amounts of NGC’s excess insurance typically increased with most insurers over time until 1985, when NGC’s insurers placed asbestos exclusions into NGC’s policies.

Under the terms of the policies at issue, the Insurers agreed to indemnify for “all sums” that NGC became legally obligated to pay as damages as a result of bodily injury or property damage caused by an “occurrence.” Although the definitions vary slightly, a common definition of “occurrence” in the policies is “an accident, or a continuous or repeated exposure to conditions which results, during the policy period, in personal injury, property damage ... neither expected nor intended from the standpoint of the insured.” “Personal injury” is typically defined to include bodily injury, sickness, or disease, while “property damage” is defined to include physical injury to or destruction of tangible property, including the loss of use.

Because the policies are triggered by injury or damage that occurs during the policy period, the trials focused extensively on when asbestos-related bodily injury and property damage occurs for purposes of these policies. In addition, because many claimants alleging the presence of ACMs in their buildings seek the costs of inspection, containment, removal, and replacement, the trial court had to determine whether such claims could be considered claims for “property damage.” Moreover, a number of NGC’s primary liability policies include a “per occurrence” deductible for property damage claims, and consequently another question was how many deductibles should be applied in the context of asbestos-in-building claims against NGC, involving thousands of buildings constructed around the country.

C. The ACF and CCR Claims-Handling Facilities

On June 19, 1985, NGC, along with 33 other former asbestos products producers and 16 insurers, entered into “The Agreement Concerning Asbestos-Related Claims,” generally known as “the Wellington Agreement.” Under the terms of the Wellington Agreement, the Asbestos Claims Facility (“ACF”) was established to evaluate, defend, and settle all asbestos-related bodily injury claims presented to it by its subscribing producers, and to pay settlements, judgments, and legal expenses incurred in the handling of claims against subscribing producers.

Costs incurred by the ACF were allocated to participating producers pursuant to a formula incorporated into the Wellington Agreement. NGC’s share of such costs was borne by those of its insurers that were also subscribers to the Wellington Agreement, to the extent of the insurance coverage afforded to NGC under its policies and pursuant to the terms of the Wellington Agreement. The Wellington Agreement resolved insurance coverage disputes with respect to asbestos-related bodily injury claims between NGC and those of its insurers that had signed the agreement. The ACF handled all those bodily injury claims against NGC from October 1,1985, until October 1,1988. The ACF was dissolved on October 3,1988.

Simultaneous with the dissolution of the ACF, NGC entered into an agreement (“the CCR Producer Agreement”) with 21 former asbestos products producers, all of which had been parties to the Wellington Agreement, to form an organization to replace the ACF. The CCR Producer Agreement created the Center for Claims Resolution (“CCR”), which has the same responsibilities for its subscribing producers as the ACF had under the earlier Wellington Agreement.

No insurer is a signatory to the CCR Producer Agreement, but several insurers who were signatories to the Wellington Agreement have entered into separate agreements concerning the CCR. Pursuant to those separate agreements, each such insurer agreed to support and cooperate with the CCR in the defense of asbestos-related bodily injury claims, to abide by the allocations of liability assigned by the CCR to each subscribing producer, and to pay to the CCR its allocated portion of liability payments and expenses.

Pursuant to the CCR Producer Agreement, the CCR pays settlements, judgments, and legal expenses incurred in the defense of all asbestos-related bodily injury claims filed against a subscribing producer. Prior to December 1, 1991, the cost of each claim was divided among all subscribing producers, regardless of which producer was named as a party-defendant. Each producer’s payments were based upon a formula, incorporated into the CCR Producer Agreement, that assigned a specific share to each producer based upon specific occupational categories of claims, the number of claims in which a producer was named, and each producer’s historical liability for claims in each category.

Effective December 1, 1991, the CCR revised its sharing formula so that, for most types of claims, the cost of a settlement or judgment is shared among only those subscribing producers that are named in that claim, rather than among all subscribing producers. Under the new sharing formula, the percentage allocated to each subscribing producer varies from claim to claim depending upon the occupational category of the claim, the number of subscribing producers named in the claim, and their applicable occupational averages. The insurance coverage provisions of the Wellington Agreement survived the dissolution of the ACF, so that NGC and its Wellington-signatory insurers continue to abide by the coverage agreements incorporated into the Wellington Agreement.

Because the Insurers in this action were not signatories to the Wellington and CCE Agreements, the District Court had to determine whether NGC’s decision to resolve the asbestos-related bodily injury claims against it by joining these claims-handling facilities was reasonable, and whether all amounts paid by NGC pursuant to these agreements, including amounts paid on claims in which NGC was not named as a defendant, could be recovered from these non-signatory Insurers under the terms of their policies.

D. NGC’s Chapter 11 Bankruptcy Reorganization

On October 28, 1990, NGC filed a voluntary petition for relief under Chapter 11 of the Bankruptcy Code. The NGC Plan of Reorganization was confirmed on March 9, 1993, and consummated on July 1, 1993. As of July 1, 1993, the NGC Settlement Trust was formed to resolve bodily injury and property damage claims against NGC arising from the exposure to or installation of NGC’s asbestos-containing products. The ownership of all NGC liability insurance policies that are potentially available to pay claims remains with ACMC. Accordingly, ACMC has been substituted for NGC in the coverage actions and on appeal.

II. Procedural Background and Rulings

A. Pretrial Proceedings

In December 1986, Stonewall Insurance Company (“Stonewall”) commenced an action for declaratory relief against NGC and certain of NGC’s other liability insurers. Stonewall sought a determination of the parties’ rights and obligations with respect to claims by property owners against NGC to recover the costs of removing and replacing NGC’s asbestos-containing products installed in their buildings. NGC filed a counterclaim and cross-claims, asserting that it was entitled to coverage for asbestos-in-building claims under the liability insurance policies.

In July 1991, while summary judgment motions on the property damage coverage issues were pending, NGC amended its counterclaim and cross-claims to assert an additional claim for coverage for the asbestos-related bodily injury claims that had been and would be asserted against NGC. NGC sought to require the Insurers to make payments under their policies in accordance with the claims-handling arrangements in the Wellington and CCR Agreements.

On March 10, 1992, the District Court denied certain insurers’ motion to obtain partial summary judgment based on the “known loss” doctrine. The District Court held that the doctrine precluded coverage only for claims that NGC actually was aware of prior to the inception of a particular insurance policy.

On May 26, 1992, the District Court issued a Memorandum Opinion and Order that considered a number of issues. First, the District Court ruled that the incorporation of ACMs into a physical structure constitutes “property damage” within the meaning of the insurance policies. Second, the District Court concluded that the appropriate “occurrence” for purposes of any applicable deductible is NGC’s decision to manufacture and sell asbestos-containing products, rather than each separate installation of ACMs. Third, the District Court ruled that damage to property, sufficient to invoke or “trigger” coverage under the policies, occurs only when the ACMs are installed in buildings. Fourth, after determining that New York law applies to the substantive interpretation of NGC’s insurance policies issued prior to the spring of 1976 and that Texas law applies to the substantive interpretation of NGC’s insurance policies issued after the spring of 1976 (a ruling not challenged on appeal), the District Court ruled that, as a matter of New York and Texas law, certain “business risk” policy exclusions do not bar coverage for asbestos claims against NGC.

At a hearing on June 15,1992, the District Court issued two rulings concerning the defense obligations of certain excess insurers. Specifically, the District Court found that: (1) certain excess policies create an obligation to defend and pay defense costs only if those insurers consent to participate in NGC’s defense, and (2) excess policies containing the so-called “New York Amendatory Endorsement” lack any defense obligation whatsoever.

On June 22,1992, the District Court issued two other significant rulings. First, the District Court denied certain insurers’ motion for partial summary judgment, concluding that those insurers that did not sign the Wellington Agreement are nevertheless obligated to indemnify NGC for its reasonable settlement of bodily injury claims through the claims-handling facilities. Second, the District Court granted certain insurers’ motion for partial summary judgment and denied NGC’s cross-motion, finding that when a claim triggers multiple policies, the liability must be prorated among the policies, with the policy-holder also sharing pro rata for uninsured periods.

On July 22,1992, the District Court issued another Memorandum Opinion and Order concerning certain insurance policies that were issued to NGC for partial years — policies that were either canceled prior to the expiration of the policy period or extended for a period of several months. The District Court generally concluded that the policies’ full aggregate limits of liability apply for any period less than one year. However, the District Court made an exception for two excess insurers on the ground that, in canceling and then renewing the policy, the parties had intended merely to increase the aggregate limit available under the old policy, rather than to create an additional aggregate policy limit.

On September 28, 1992, the District Court ruled that certain policies issued by Transit Casualty Company (“Transit”) and American Centennial Insurance Company (“ACIC”) impose a duty to defend NGC; that the Texas Amendatory Endorsement to those policies excuses only the duty to defend claims arising in Texas; and that several excess policies contain no defense obligation whatsoever.

B. Jury Trial and Bench Trial Proceedings

In the course of its final pretrial proceedings, the District Court severed the cases of those insurers that had demanded a jury trial from those insurers that had not. Several insurers subsequently waived their right to a jury trial, and the jury trial proceeded against only Commercial Union (“CU”).

On October 22, 1992, the jury returned a special verdict in favor of NGC on three separate issues. The jury found that: (1) NGC did not “expect or intend” to cause asbestos-related bodily injury, (2) for persons who contracted asbestosis and asbestos-induced cancers, their bodily injury occurred from the time of first exposure and continuously and progressively thereafter, and (3) NGC’s decisions to participate in the ACF and CCR claims-handling facilities were reasonable methods of resolving the bodily injury claims asserted against NGC. The District Court denied CU’s motion to set aside the verdict and order a new trial.

NGC and the remaining insurers then litigated their disputes in a non-jury trial. In addition to considering all of the evidence that was presented to the jury, the District Court heard several days of supplemental testimony and received additional exhibits. On December 22, 1992, the District Court rendered its findings of fact and conclusions of law, including finding that: (1) for persons who contracted asbestosis and fibrosis, their bodily injury occurred from the time of first exposure and continuously and progressively thereafter, (2) for persons who contracted asbestos-induced cancer, bodily injury occurred only at the time of exposure and did not continue progressively thereafter, (3) NGC did not “expect or intend” to cause asbestos-related bodily injury, (4) NGC’s decisions to participate in the ACF and CCR claims-handling facilities were reasonable methods of settling the bodily injury claims asserted against NGC, (5) NGC is bound by an exposure-only trigger of coverage with respect to policies that it purchased between 1983 and 1985, and (6) the asbestos-related bodily injuries were not a “known loss” to NGC.

C. Judgments and Permission to Appeal

On March 30, 1993, partial declaratory judgments were entered incorporating all prior rulings. These judgments: (1) determined coverage for asbestos-related bodily injury claims under insurance policies involved in the jury trial (Jury Trial BI Judgment), (2) determined coverage for asbestos-related bodily injury claims under insurance policies involved in the bench trial (Bench Trial BI Judgment), (3) determined coverage for asbestos-related property damage claims (PD Judgment), and (4) resolved NGC’s bad faith claims against American Motorists Insurance Company (“AMICO”). The bad faith claims are not at issue on this appeal. The judgments were certified for entry under Fed.R.Civ.P. 54(b) and for interlocutory appeal under 28 U.S.C. § 1292(b).

NGC filed notices of appeal pursuant to Rule 54(b) and a petition for leave to appeal under section 1292(b). The Insurers filed petitions for leave to appeal, conditional petitions for leave to appeal, and protective notices of appeal. On January 11, 1994, we accepted interlocutory review of the Jury Trial BI Judgment, the Bench Trial BI Judgment, and the PD Judgment.

Discussion

I. Summary of Holdings

With respect to the bodily injury claims asserted against NGC, we conclude:

(1) The District Court was correct that under New York and Texas law, occurrence-based policies could be triggered throughout a gradual disease process where injury can be shown by a preponderance of the evidence to be occurring at each point in that process.

(2) The factual findings of the jury against CU that injury was occurring continuously from exposure through claim or death for both non-cancer and cancer diseases, and the factual findings of the District Court with respect to the other insurers that injury was occurring continuously from exposure through claim or death for non-cancer diseases, but not for cancer diseases, were all adequately supported by the evidence. However, in view of the ambiguity in the District Court’s application of the injury-in-faet approach to cancer claims, the bench trial finding as to that issue is remanded for further consideration.

(3) The District Court did not err in allocating to NGC the pro rata share of liability attributable to those periods during which NGC was uninsured, but this “proration-to-the-insured” approach is modified so as not to apply to injuries occurring after 1985, when asbestos insurance was unavailable.

(4) The District Court correctly ruled that NGC did not expect or intend the bodily injuries caused by its products.

(5) The District Court properly concluded that NGC’s participation in the ACF and CCR asbestos claims-handling facilities was reasonable and that payments made by NGC through these facilities are covered by the policies.

With respect to the asbestos-in-building claims against NGC, we conclude:

(6) The costs of removing and replacing asbestos products from buildings were properly considered “property damage.”

(7) The District Court was correct that NGC could not establish by a preponderance of the evidence that property damage was occurring continuously or at any given point following installation of NGC’s building materials.

(8) The District Court erred in holding that, for purposes of “per occurrence” deductible provisions in the policies, all the asbestos-in-building claims arose out of a single occurrence — namely, NGC’s decision to manufacture and sell asbestos-containing building materials. Rather, each installation of NGC’s products constituted a separate occurrence, requiring the application of another deductible.

With respect to all claims, we conclude:

(9) The District Court properly ruled that neither policy exclusions nor the “known loss” defense barred NGC from indemnification.

(10) The District Court correctly ruled: (a) as to the aggregate policy limits available to NGC under policies that were either canceled before the policy had expired or were extended for a period beyond the expiration date; (b) as to almost all rulings with respect to the scope of excess insurers’ defense obligations; and (c) in treating all the Insurers as a single defendant in allocating peremptory challenges.

We turn first to those issues relating to coverage of asbestos-induced bodily injury claims against NGC, and then consider those issues relating to coverage of asbestos-in-building claims. Finally, we consider questions pertaining to all claims — specifically, the “known loss” defense, the amount of aggregate policy limits available to NGC, the defense obligations of certain excess insurers, and the District Court’s allocation of peremptory challenges to the Insurers.

II. Issues Related to Coverage for Bodily Injury Claims

A. Bodily Injury Trigger of Coverage

1. Policy Language. Unlike “claims-made” policies in wide use today, which require the assertion of a claim against the insured during the policy period, NGC’s “occurrence-based” policies respond to its liabilities arising out of “bodily injury” or “property damage” that took place during the time that such policies were in effect, even if the claim is not made until years after the termination of the policy.⁴ See generally Hartford Fire Insurance Co. v. California, — U.S. -, -, 113 S.Ct. 2891, 2896, 125 L.Ed.2d 612 (1993).

These policies typically provide that the insurer will indemnify the insured for “all sums” that must be paid as damages because of personal injury “caused by or arising out of an occurrence.” The term “occurrence” is in turn defined as “an accident, or a continuous or repeated exposure to conditions which results, during the policy period, in personal injury ... neither expected nor intended from the standpoint of the insured.”⁵

2. NGC’s and Insurers’ Positions. In this case, we must resolve the following question: at which point or points in time between the initial inhalation of asbestos fibers and the date of claim is there “bodily injury” or “personal injury” within the meaning of the policies so as to “trigger” coverage? As already noted, the bodily injury claimants suing NGC were directly exposed to asbestos fibers at various points in time beginning as early as the 1930s. Years after first exposure, the claimants are diagnosed as having an asbestos-related disease, and then they bring suit against NGC.

The insurance industry has been and remains unable to agree on a consistent interpretation of the form wording in NGC’s policies as applied to asbestos claims and other progressive injury claims.⁶ For example, on this appeal, CU takes the position that “bodily injury” within the meaning of its policies occurs only at the point in time when the asbestos-related diseases were either manifested or became fully developed, while AMI-CO, joined by a second group of insurers,⁷ contends that only those policies in effect during the period of a claimant’s exposure to asbestos must respond to the asbestos-induced bodily injury claims.

The CU position would result in coverage limited to the time period when the diseases were, or should have been, diagnosed. For the overwhelming number of cases now being brought against NGC, that date of diagnosis or diagnosability will be in the mid to late 1980s or 1990s. By that time, however, NGC’s insurers had inserted asbestos exclusions into most or all of NGC’s policies. CU’s position would leave NGC largely uninsured for current claims.

On the other hand, if, as AMICO contends, bodily injury within the meaning of the policies is confined to those injuries occurring at the time of inhalation of asbestos fibers, then the policies in effect after direct exposure to asbestos ceases (but while the progressive asbestos injury process is developing) will not be required to respond. In many instances, NGC had relatively little insurance available to it during the time periods of a claimant’s direct exposure to asbestos fibers, because from the 1930s through the 1960s, the policy limits sold by insurers typically were low and other claims have depleted or will deplete these limits. By contrast, manufacturers such as NGC had available far more substantial policy limits during the 1970s and 1980s, the time periods after direct exposure ceased in a typical asbestos case.

Unsurprisingly, NGC’s position is that all insurance policies “on the risk” at any point during the asbestos-induced bodily injury process are triggered, from the time of exposure through claim or death, whichever is earlier, regardless of whether during a particular policy period there was “exposure” to asbestos and regardless of whether during a policy period the injuries were fully developed, discoverable, or compensable.

3. Jury and Bench Trial Decisions. The jury and bench trials had to determine, among other things, when, under the language of the policies, as interpreted under New York and Texas law and in light of the medical evidence presented, injury took place with respect to the asbestos-induced bodily injury claims asserted against NGC.

Based on the jury’s verdict, the Jury Trial BI Judgment provides that bodily injury within the meaning of CU’s policies occurs from the date of first injurious exposure through the date of claim or death, whichever is earlier, and that all policies in effect during any portion of that period are triggered. This ruling applies to both the cancer and non-cancer bodily injury claims against NGC. The District Court also found, and the Bench Trial BI Judgment provides, that with respect to the non-cancer asbestos-induced bodily injury claims, bodily injury within the meaning of the policies occurs from the date of first exposure to the date of claim or death, whichever is earlier, and that all policies in effect during any portion of that time period are triggered. Finally, the District Court found, and the Bench Trial BI Judgment provides, that, with respect to asbestos-induced cancer claims, bodily injury occurs only at or shortly after inhalation of asbestos fibers and only those policies in effect during that limited time period are triggered.

NGC contends that the trigger of coverage rulings set forth in the Jury Trial BI Judgment and the Bench Trial BI Judgment should be affirmed insofar as they relate to non-cancer claims, but that the District Court’s trigger ruling in the Bench Trial BI Judgment with respect to cancer claims should be reversed. All the Insurers challenge the Jury Trial BI Judgment as well as that part of the Bench Trial BI Judgment that applies to non-cancer bodily injury claims. CU also challenges the District Court’s ruling with respect to cancer claims, on the ground that such claims should be triggered, not at or shortly after inhalation of asbestos fibers, but only when the diseases are manifested or fully developed.

4. New York and Texas Case Law. Our inquiry as to how New York would answer the triggering inquiry begins with American Home Products Corp. v. Liberty Mutual Insurance Co., 748 F.2d 760 (2d Cir.1984) (“AHP ”), where we noted that under New York law, coverage is based upon the occurrence of an injury-in-fact during the policy period. See id. at 764. Thus, “ ‘a real but undiscovered injury, proved in retrospect to have existed at the relevant time, would establish coverage, irrespective of the time the injury became [diagnosable].’ ” Id. at 766 (emphasis added) (quoting AHP, 565 F.Supp. 1485, 1497 (S.D.N.Y.1983)).

Our decision in AHP, interpreting a variant of the standard language of the Comprehensive General Liability (“CGL”) Policy,⁸ rejected interpretations of the standard language urged by both parties in that case. Specifically, we rejected the insured’s contention that the policy was triggered if exposure, or injury, or manifestation occurred during a policy period, and the insured’s alternate contention that exposure alone was the triggering event. We also rejected the insurer’s contention that manifestation was the triggering event. In our view, the plain meaning of the terms of the CGL policy provided that a policy was triggered by an injury-in-fact during the policy period. We upheld the District Court’s interpretation and analysis, except for the District Court’s requirement that the injury be compensable or diagnosable in order to trigger a policy.

Our decision in AHP did not have occasion expressly to consider the precise issue posed by the pending case — whether under the injury-in-fact approach a progressive disease process like asbestosis or cancer can be found, on sufficient evidence, to precipitate successive injuries that trigger all of the policies in force during each period in which such successive injuries occur. Though our opinion said it was upholding the District Court’s rejection of “AHP’s continuous trigger theory,” AHP, 748 F.2d at 763, our reference was to AHP’s contention that, even without evidence of injury-in-fact, all policies should be triggered that were in effect during either exposure, injury, or manifestation. However, the District Court, whose judgment we substantially affirmed, id. at 763-64 n. 1, had specifically stated:

In sum, the insured would be permitted under a plain reading of the policy to establish that occurrences have taken place at any or all points from exposure to manifestation, on the ground that identifiable injuries have been proved to have occurred at each point to a reasonable degree of medical certainty.

American Home Products Corp. v. Liberty Mutual Insurance Co., 565 F.Supp. 1485, 1498 (S.D.N.Y.1983) (emphasis added). In light of this explicit statement, our approval of the District Court’s rejection of “AHP’s continuous trigger theory” is properly understood to be a rejection only of the theory that in every progressive disease case all policies are automatically triggered that were in effect at exposure, injury, or manifestation; by substantially affirming the District Court’s judgment, we impliedly upheld then-Judge Sofaer’s statement that triggering can be shown by competent evidence to result at successive points where “identifiable injuries” have occurred. By deleting the word “diagnosable” from the judgment, we made clear that the requisite injury could be found based on probative medical opinion evidence, without clinical findings. But cf. Eagle-Picher Industries, Inc. v. Liberty Mutual Insurance Co., 682 F.2d 12, 19 (1st Cir.1982) (“[I]t is difficult to consider sub-clinical insults to the lung to constitute an ‘injury’ when these results do not cause ‘loss, pain, distress, or impairment’ until, if ever, they accumulate to become clinically evident or manifest.”). See generally Abex Corp. v. Maryland Casualty Co., 790 F.2d 119, 127 n. 36 (D.C.Cir.1986) (noting that different views of etiology of asbestos-related diseases make summary judgment inappropriate).

Our Court’s subsequent decision in Maryland Casualty Co. v. W.R. Grace & Co., 23 F.3d 617 (2d Cir.1993) (opinion amended 1994) (“Grace ”), is instructive on the pending issue. Considering a coverage dispute concerning property damage, we ruled that property damage from asbestos occurs at installation and does not continue to occur after installation. Id. at 628. We then observed, in a passage pertinent to the issue of personal injury triggering, that “[s]ome types of property damage — such as the gradual contamination of earth and groundwater by leaking landfills — may be analogous to the slow progression of diseases such as asbestosis and cancer,” id. at 627. By that we meant that there can be triggering at more than one point in time when a claimant asserts injury-in-faet due to asbestosis or cancer. That meaning is evident from the citation to New Castle County v. Continental Casualty Co., 725 F.Supp. 800, 809-10 (D.Del.1989), aff'd in part and rev’d in part, 933 F.2d 1162 (3d Cir.1991), which discussed the possibility of continuous triggering when asbestos causes progressive diseases and relied on the Third Circuit’s continuous trigger decision in ACandS, Inc. v. Aetna Casualty and Surety Co., 764 F.2d 968 (3d Cir.1985), a personal injury, asbestosis continuous trigger holding.

Thus, this Court’s interpretation of the CGL policy provisions applies the “injury-in-fact” approach and, with respect to progressive diseases, permits triggering at various points when evidence shows injury to have occurred. See Montrose Chemical Corp. of California v. Admiral Insurance Co., 10 Cal.4th 645, 694, 42 Cal.Rptr.2d 324, 353, 897 P.2d 1, 30 (1995) (“Whether the damages and injuries alleged were in fact ‘continuous’ is itself a matter for final determination by the trier of fact.”). What we must further consider is whether our view of New York law, as expressed in AHP, must be modified, as some of the Insurers contend, by the decision of the New York Court of Appeals in Continental Casualty Co. v. Rapid-American Corp., 80 N.Y.2d 640, 593 N.Y.S.2d 966, 609 N.E.2d 506 (1983) (“Rapid-American”).

In Rapid-American, a personal injury coverage case, both the insured and the insurer agreed that the “injury-in-fact” approach was applicable under New York law. However, the insurer contended that the insured’s “practical construction” of the policy had resulted in triggering only upon manifestation. Id. at 651, 593 N.Y.S.2d at 971, 609 N.E.2d at 511. The Court of Appeals rejected that contention, and the holding of the case is not pertinent to the pending appeal; there was no dispute between the parties as to how the “injury-in-fact” test is generally to be applied under New York law.

However, language in the opinion of the Court of Appeals requires further consideration. The opinion posits four possible approaches to triggering: “exposure to asbestos,” “manifestation of disease,” “onset of disease,” or “all of the above — in other words ‘a continuous trigger.’ ” Id. In identifying the “onset” possibility, the Court of Appeals labeled it in a parenthetical as “injury-in-faet,” cited the District Court’s decision in AHP as an example, and said that the “injury-in-fact” test “rests on when the injury, sickness, disease or disability actually began.” Id. It is arguable that the Court of Appeals viewed the “injury-in-fact” approach as triggering only those policies in effect at the time when claimants’ diseases began.

Though the beginning of a disease is clearly one point at which an injury-in-fact has been sustained, we do not believe that the New York Court of Appeals intended its phrasing in Rapid-American to preclude triggering of policies in force in periods after the beginning of a disease, so long as competent evidence persuades the trier that subsequent injuries are occurring. The “continuous trigger” that the Court contrasted with the “injury-in-fact” approach would trigger policies in effect at the time of exposure, even if no injury then occurred, and would also trigger policies in effect at the time of manifestation, even if no subsequent injury then occurred. We agree that such triggering is not consistent with the “injury-in-fact” approach. But triggering by successive injuries, proven to have occurred, is another matter.

Prior to Rapid-American, the New York Court of Appeals had considered the triggering issue in the property damage context with respect to policies issued in two consecutive years. See McGroarty v. Great American Insurance Co., 36 N.Y.2d 358, 368 N.Y.S.2d 485, 329 N.E.2d 172 (1975). The insured’s property was damaged by lateral pressure of boulders, earth, and surface water against an adjacent garage. First, the Court noted that “application of the term accident in such contexts as that before us provides a question of fact.” Id. at 363, 368 N.Y.S.2d at 489, 329 N.E.2d at 174. More significantly, the Court held that the plaintiff’s claim was enforceable under all of the policies issued by the defendant to its insured. This indicates that New York recognizes that whenever the facts show injury during a relevant policy period, the policy applies, even though injury was also shown to have occurred in an earlier period covered by a prior policy. That is the understanding of the Appellate Division. See Cortland Pump & Equipment Inc. v. Firemen’s Insurance Co., 194 A.D.2d 117, 604 N.Y.S.2d 633 (3d Dep’t 1993) (stating, after Rapid-American, which was cited, that “the injury-in-fact standard should be applied to the property damage, a continuous occurrence resulting from plaintiffs alleged negligence which permits a finding that the property damage occurred during the policy period”); id. at 118, 604 N.Y.S.2d at 636. We do not believe the phrasing in Rapid-American, which was not critical to any disputed issue between the parties in that case, was intended to limit the approach previously expressed in McGroarty or to preclude triggering of policies in effect in any period in which successive injuries are shown to have occurred.

We also note that several states, including California, see Montrose Chemical, 10 Cal.4th 645, 42 Cal.Rptr.2d 324, 897 P.2d 1, New Jersey, see Owens-Illinois, Inc. v. United Insurance Co., 138 N.J. 437, 650 A.2d 974 (1994), and Pennsylvania, see J.H. France Refractories Co. v. Allstate Insurance Co., 534 Pa. 29, 626 A.2d 502 (1993), have interpreted the CGL policy to provide for a continuous trigger. As the California Supreme Court noted, “most courts that have analyzed the issue have found the continuous injury trigger of coverage applicable to the standard occurrence-based CGL policy.” Montrose Chemical, 10 Cal.4th at 676, 42 Cal.Rptr.2d at 342, 897 P.2d at 19. The New Jersey Supreme Court does not even require the medical evidence and fact-finding that we require. Owens-Illinois, 138 N.J. at 453-54, 650 A.2d at 982-83.⁹ Though New York rejects such an expansive version of continuous triggering, we do not believe it limits the injury-in-fact approach to trigger only the policy in effect when the disease began.

Texas appears likely to adopt a similar position with respect to asbestos-related personal injury claims. In American Physicians Insurance Exchange v. Garcia, 876 S.W.2d 842, 853 n. 20 (Tex.1994), the Supreme Court of Texas identified various approaches that courts have taken to the trigger of coverage issue and declined to adopt any particular approach. In the pending case, the District Court noted that there was no controlling Texas authority on trigger of coverage but predicted that the Texas Supreme Court would embrace the injury-in-fact principles of AHP, a ruling that no party challenges on appeal.

Indeed, in National Standard Insurance Co. v. Continental Insurance Co., No. CA-3-81-1015-D (N.D.Tex. Oct. 4, 1983), the District Court construed Texas law, in the context of injuries caused by exposure to carcinogenic chemicals, to permit triggering of all policies in effect between the initial exposure and time of manifestation, without the need for specific evidence to prove that successive injuries were recurring. Similarly, in Dayton Independent School District v. National Gypsum Co., 682 F.Supp. 1403 (E.D.Tex.1988), the District Court observed, citing National Standard, that all policies are triggered where injury is continuous, id. at 1410, and although the Fifth Circuit reversed on jurisdictional grounds, it did not question the District Court’s statement of the law as to triggering of coverage for bodily injury claims. See W.R. Grace & Co. v. Continental Casualty Co., 896 F.2d 865 (5th Cir.1990).

We do not read Texas law as expansively as did the District Court in National Standard to allow continuous triggering whenever claimants of an insured have been exposed to any carcinogenic substances. But we can confidently predict that Texas and New York will permit triggering throughout the period between exposure and date of claim or death in all eases in which the evidence persuades the trier of fact that successive injuries are recurring. Continuous triggering, established simply by proof of exposure to carcinogenic substances, is an approach adopted by some courts, see, e.g., Owens-Illinois, 138 N.J. at 474-79, 650 A.2d at 993-95, and rejected by others, see, e.g., Armstrong World Industries, Inc. v. Aetna Casualty & Surety Co., 30 Cal.App.4th 1117, 1169-75, 26 Cal.Rptr.2d 35, 50-54 (1993) (affirming trial court’s continuous triggering decision in view of factual findings that injury-in-faet occurred during each policy period), remanded for reconsideration in light of Montrose Chemical, 10 Cal.4th 645, 42 Cal.Rptr.2d 324, 897 P.2d 1, In re Asbestos Insurance Coverage Cases, — Cal.4th -, 46 Cal.Rptr.2d 174, 904 P.2d 370 (1995). In the pending ease, we decide only that both states would rule that, at least where the evidence establishes a progressive bodily disease, with injury-in-fact recurring throughout the disease process, all policies in effect at any time during that process are triggered.¹⁰ We therefore reject both AMICO’s position, which limits injury-in-faet to the initial injury occurring at or shortly after exposure to asbestos, and CU’s position, which limits injury-in-fact to the time when the injury becomes either diagnosable or fully developed.¹¹ We now turn to an examination of the medical evidence presented to determine whether the findings of the jury and the District Court on this issue are supportable.

5. Evidence on Etiology of Asbestos Diseases. Both the District Court and the jury concluded that NGC had made a sufficient showing of injury-in-fact at each point in the development of asbestosis and pleural plaques, beginning shortly after exposure and ending with claim or death. In addition, the jury found that NGC had made a similar showing with respect to asbestos-related cancers. However, the District Court concluded that there was insufficient evidence to establish that injury was occurring at any given point in the spectrum between exposure and manifestation of the cancer, and that at most NGC had shown that injury, in the form of cell mutation, was occurring intermittently at some points in the process. Consequently, the District Court decided that injury-in-faet had been shown to occur by a preponderance of the evidence only at the time of exposure.

We conclude that there was sufficient evidence to support all the findings of both the jury and the District Court. Though it conflicts with the jury’s finding of continuous injury, the District Court’s conclusion as to asbestos-induced cancers was not clearly erroneous. Moreover, in view of the different etiologies of asbestosis and cancer, it is possible to hold, as the District Court did, that these two diseases trigger policies differently. However, in view of the ambiguity in the District Court’s application of the injury-in-fact approach to cancer claims, we remand this issue for further consideration.

(a) Asbestosis and Pleural Plaques. Both the jury and the District Court independently found that coverage is triggered for nonmalignant asbestos-related diseases (diseases other than cancer) commencing with the claimant’s first exposure to asbestos and progressing continuously thereafter. The Insurers argue that these findings in the jury and bench trials as to asbestosis and pleural plaques were not supported by sufficient evidence.

NGC presented evidence to the effect that there is a period of time prior to the point at which asbestosis and pleural plaques are clinically diagnosable when the condition is evolving cumulatively or incrementally. This period when the asbestosis or pleural plaques develop undetected is the latency period, which is normally 15 years and may be as long as 40 years. The latency period begins at the time of initial exposure to asbestos and ends with the onset of symptoms and diagnosis. From the time of exposure, asbestos fibers are readily inhaled into the lungs and can immediately injure the cells by causing inflammation and scarring. The disease process for a person who develops asbestosis begins within days of initial exposure to asbestos and continues for decades. How far this disease process progresses depends upon the quantity of asbestos inhaled and deposited in the lungs.

Even if an individual is no longer exposed to asbestos, asbestos fibers within the lung continue to split or divide over time, which is the main cause of the disease continuing and progressing. “Injury” continues as long as there is asbestos in the lung. In short, NGC’s evidence adequately supported the findings of both the jury and the District Court that injury due to the presence of asbestos fibers in the lung continues throughout the development of asbestosis and pleural plaques.

AMICO argues that, even if asbestosis progresses, the progression does not constitute a “new” or “different” injury that would trigger additional coverage, and CU contends that sub-clinical changes to a claimant’s lung are not “real injuries.” However, continual progression of a disease process into successive policy years is not excluded from the definition of “bodily injury” contained in these policies, nor from the concept of “injury-in-fact,” and the jury and the District Court were entitled to credit the testimony of NGC’s experts regarding the timing and nature of the injuries involved in asbestosis and pleural plaques.¹²

(b) Asbestos-Induced Cancers. With respect to asbestos-induced cancers, the jury’s determination that injury-in-fact commences with the time of first exposure and progresses continuously thereafter was sufficiently supported by the evidence regarding the etiology of such cancers. This evidence demonstrated the ability of asbestos to cause cancer, and permitted a finding that asbestos potentially plays a role in each and every mutation of cells. The jury could reasonably have concluded that such mutations, and thus injury-in-fact, had been established by NGC to occur at every moment in the disease process, thereby triggering all policies in effect at any point during that process.

Specifically, upon exposure to a carcinogenic substance, a cell “mutates,” after which the cell containing that mutation replicates itself until millions of cells exist with the same mutation. One of those mutated cells may then undergo a second mutation, and so on until enough mutations have taken place to result in a cell that scientists call a “cancer cell.” Asbestos can contribute to the process in four ways: it can (1) cause the mutation directly, (2) increase the likelihood that another mutation will occur, (3) introduce other carcinogenic substances into the cells, and (4) cause the release of other substances capable of causing certain mutations that asbestos may not be able to effect on its own. These processes may involve millions of cells, each at various stages of the process.

In short, the medical evidence regarding the carcinogenic mechanisms of asbestos sufficiently establishes an insidious disease process involving billions of cellular mutations that constitute injuries. The jury could reasonably have concluded that these mutations occurred in a continuous sequence from initial exposure to the manifestation of the clinical diseases known as lung cancer and meso-thelioma. Based on this evidence, the jury found that, for asbestos-induced cancer claims, CU’s policies (the only ones at issue in the jury trial) are triggered from the date of first exposure to the date of death or claim, whichever is earlier. We will not disturb this finding.

In addition to the Insurers’ challenges to the sufficiency of the evidence, CU contends that the jury’s findings concerning asbestosis and cancer were against the weight of the evidence. That contention requires us to decide whether a trial judge’s denial of a motion to set aside a verdict as against the weight of the evidence is reviewable on appeal. We recently ruled that a trial judge’s grant of such a motion was reviewable, Binder v. Long Island Lighting Co., 57 F.3d 193 (2d Cir.1995), but explicitly declined to consider the reviewability of a denial of such a motion, id. at 202 n. 3. Our recent decision in Blissett v. Coughlin, 66 F.3d 531, 533 (2d Cir.1995), stated that a denial was not reviewable, but did not purport to discuss the issue that had pointedly been left open in Binder. Blissett relied on Dunlap-McCuller v. Riese Organization, 980 F.2d 153, 157 (2d Cir.1992), cert. denied, — U.S. -, 114 S.Ct. 290, 126 L.Ed.2d 239 (1993), which had denied review of an order granting such a motion, a result subsequently rejected by Binder. Whether a denial ruling is reviewable must therefore be considered anew.

Review of a trial court’s ruling assessing the weight of the evidence imposes on an appellate court far more of a burden than arises from review of a ruling rejecting a challenge to the sufficiency of the evidence. The latter ruling can be readily affirmed as soon as the reviewing court identifies adequate evidence in the record that permits the disputed issue to go to the jury, despite the existence of significant opposing evidence. Such review does not require an assessment of all the evidence. Reviewing a ruling on a “weight of the evidence” challenge, however, obliges a reviewing court to examine in some detail all of the evidence. That burdensome review is warranted in the rare case where a trial judge rejects a jury’s verdict as against the weight of the evidence, as Binder held, but is not warranted in the far more frequent circumstance where a trial judge denies a “weight of the evidence” challenge and leaves in place a jury verdict supported by legally sufficient evidence. In the latter circumstance, the loser’s only appellate recourse is to challenge the legal sufficiency of the evidence. The loser is also entitled to argue to the trial judge that the verdict is against the weight of the evidence and to obtain a new trial if the judge can be persuaded, but the denial of that challenge is one of those few rulings that is simply unavailable for appellate review. See Portman v. American Home Products Corp., 201 F.2d 847, 848 (2d Cir.1953) (L. Hand, J.).¹³

In the bench trial, the District Court ruled, based on the evidence recounted above and additional evidence regarding the etiology of asbestos-induced cancers, some of which was conflicting, that only those policies in effect at the date of direct exposure must respond to these cancer claims. NGC argues that this ruling is inconsistent with the evidence and the District Court’s own findings regarding non-cancer asbestos claims.

The District Court appeared to accept as true the fact that even after direct exposure to asbestos fibers ceases, cells that become aneuploidic as a result of exposure to asbestos multiply and proliferate throughout the body. As a result of these mutations, the claimant may eventually contract cancer. Nevertheless, the District Court reasoned, “[I]f it were clear from the medical evidence that exposure to asbestos caused cells to become cancerous at the time of exposure and that the only thing that happened over the latency period was that the number of those cells proliferated throughout the system until they became detectable, then [AHP ] would say that the cancer, the injury, sickness or disease, occurred in the year of exposure, not thereafter.” “[N]o one could tell you with any degree of medical certainty when that cancer began,” and “if under [AHP ] we are trying to look to an event that causes cancer, it seems to me we are back to the injury that occurred at the time the person was exposed to asbestos.”

In assessing the District Court’s reasoning, we encounter some uncertainty as to Judge Martin’s meaning. It is possible that he somewhat overread AHP and analyzed the evidence before him to determine only the point at which NGC had proven that the cancers began, rather than to determine whether NGC had proven that injury-in-fact was occurring continuously throughout a claimant’s disease process. On the other hand, some aspects of Judge Martin’s analysis suggest that he correctly understood the injury-in-fact approach and concluded that, though NGC had proven sporadic injuries in the form of clonal expansion, it had not persuaded him that such injuries were occurring continuously.

As we have explained, we understand AHP to have used an injury-in-fact approach that ordinarily leaves to the fact-finder the task of determining whether an insured, contending for a continuous trigger of coverage, has proven by a preponderance of the evidence that injuries were in fact occurring continuously during the disease process. Despite the jury’s finding of continuous injuries, the Judge in the bench trial was free to assess the record before him differently. Under all the circumstances, we conclude that the appropriate course is to afford the District Judge, on remand, an opportunity to reassess the record in light of this opinion. If, upon such reconsideration, Judge Martin is unpersuaded that the record shows continuous injuries, he should reconfirm this aspect of the judgments; however, if he finds that continuous injuries have been proven to be occurring, he should modify the judgments accordingly.

6. Coverage for Injury Resulting from Exposure to Other Manufacturers’ Products. The underlying asbestos claimants seek to hold NGC and others jointly and severally hable for all damages they have suffered, as long as NGC’s products had some causal relationship to the development of the disease. In keeping with joint and several liability, the District Court required coverage by all policies in effect from time of first exposure to asbestos — whether to the insured’s or some other company’s asbestos — and continuing to date of claim or death. The Insurers object that this results in coverage even where the claimant was not exposed to NGC’s product until after the policy had expired, and that NGC should be required to show exposure to its products within the policy period.

The District Court declined to limit coverage to those claims in which exposure to NGC’s products occurred during a policy period, on the ground that NGC acted reasonably in settling claims through the ACF and CCR facilities, even though some of those claims might be based in part on exposure to other manufacturers’ products. As discussed below, see Part 11(D), infra, we conclude that the District Court was correct in upholding all aspects of NGC’s payments through these two claims-handling facilities. Indeed, the Insurers would likely have contended that NGC was unreasonable if it had litigated a class action just to reject those few claimants who had been exposed solely to other manufacturers’ products. As long as injury-in-fact occurred during the policy period, the Insurers could not withhold coverage until NGC demonstrated that each member of the class had been exposed to NGC’s products while the policy was in effect.

The Insurers contest the District Court’s ruling by arguing that their policies provide liability coverage to NGC for liabilities, arising from “occurrences” for which NGC is responsible, that involve injuries-in-fact that NGC causes. However, these liability policies are far from unambiguous with respect to whether the insured must be the cause of the injury that occurs during the policy period. The policies may reasonably be interpreted to provide coverage for any bodily injury during the policy period for which the policy-holder is held liable, even if the policyholder is later shown not to have caused that bodily injury.

Had the Insurers wished to limit indemnification to occurrences caused by the insured, and thus to require NGC to prove that the relevant injury-in-fact was caused by NGC’s own products, they could have expressly provided for this limitation in their policies, just as they required a showing of injury during the policy period. In view of the reasonableness of NGC’s payments through the claims-handling facilities and the ambiguity in the policy language, we hold that the relevant exposure for coverage purposes in this case is exposure to asbestos generally, which causes injury-in-fact, and not merely exposure to NGC’s products.

7. Estoppel From Applying Standard Trigger of Coverage to Certain 1983-85 Excess Policies. Notwithstanding its ruling with respect to the trigger for non-cancerous asbestos diseases, the District Court ruled that those insurers that had issued excess policies for 1983-85 (“Excepted Insurers”)¹⁴ were not hable for coverage unless the claimant was actually exposed to the product during the respective time periods of these policies. The decision was based on promissory estoppel arising from statements allegedly made by NGC prior to the formation of the insurance contracts. NGC contends that the parol evidence should not have been admitted, and that the principle of promissory estoppel was improperly applied.

We reject NGC’s contentions. In interpreting the meaning of “injury” under the policies for purposes of asbestos-related claims, the District Court relied on testimony and documents showing that NGC had represented to several excess underwriters that it had stopped manufacturing asbestos-containing products by 1981, and that those insurers therefore faced a minimal risk of paying asbestos claims because “most of the exposure to these products will not fall within the current policy year.”

Though NGC disputes that its statements to these insurers constituted representations that it would seek coverage only for exposure claims, the District Court reasonably concluded that “National Gypsum was spreading the word that it had settled its actions and was proceeding on an exposure pro rata basis,” and that “the insurers relied on that fact and that was a basis for them to agree to provide the coverage that they provided.” In view of the ambiguity in the definition of “injury” for purposes of latent diseases, it was appropriate to admit extrinsic evidence, not to vary the terms of the policies, but to aid the District Court’s inquiry into the parties’ understanding of how the term “injury” was to be applied to asbestos-induced bodily injury claims.

However, we agree with NGC that, with respect to American Centennial Insurance Company (“ACIC”), no evidence was presented that NGC made any representation of the type allegedly made to the other insurers, or that ACIC relied upon any such representation.¹⁵ The District Court might have surmised that NGC was making the same representations to all its excess insurers with respect to the 1983-85 policies, and that ACIC should therefore be included with the other insurers. However, in the absence of any evidence specific to ACIC, we must reverse the ruling as to that insurer on this issue.

B. Scope of Responsibility for Claims Triggering Multiple Policies

As described above, the Jury Trial BI Judgment and the Bench Trial BI Judgment provide that the Insurers’ policies respond to the underlying asbestos-related bodily injury claims on an “actual injury” or injury-in-fact basis. Under this trigger of coverage, an asbestos-related bodily injury claim typically will implicate multiple policies in effect during the multi-year period of the injury process. Each of these triggered polir cies promises to pay “all sums” that NGC becomes liable to pay to the underlying claimants as a result of bodily injury occurring during the policy period. Thus, for any single asbestos-related bodily injury claim, there may be several policies each independently responsible under their explicit terms for paying “all sums” that NGC becomes liable to pay to the claimant.

Confronted with multiple insurance policies covering the same claim, the District Court decided, on motion for summary judgment, that the triggered policies’ obligations were to be prorated based upon the policies’ respective triggered time periods. Specifically, the District Court ruled that each triggered policy was responsible for only a pro rata share of NGC’s liability as to a particular claimant. The share was determined by multiplying the judgment or settlement by a fraction that has as its denominator the entire number of years of the claimant’s injury, and as its numerator the number of years within that period when the policy was in effect.

Having adopted this apportionment formula, the District Court then ruled that, to the extent that NGC had no insurance (because it had been uninsured or its insurance had been consumed by prior payments) during any portion of the time period of a particular claimant’s injury, NGC itself would be responsible for the pro rata share attributable to such period (“proration-to-the-insured approach”). That is, for “uninsured” periods, NGC would be treated as if it had issued itself an insurance policy and would be required to “contribute” accordingly.

Neither NGC nor the Insurers object to the formula used by the District Court to apportion among the Insurers shares of a total claim that NGC is obligated to pay.¹⁶ However, NGC objects to the District Court’s proration-to-the-insured approach. NGC observes that the policies available to it in the 1950s had low aggregate limits, and that no coverage was available for asbestos claims after 1985. Because the injuries suffered by the claimant population have progressed well into these uninsured policy years, prorating NGC’s liability evenly to each triggered policy year and to each uninsured year enables the Insurers to reallocate an increasingly large percentage of asbestos liabilities back to NGC.

NGC argues that the District Court’s approach is contrary to the policy language, because although the policies contain “other insurance” clauses that allocate liability among multiple insurers whose policies cover the same insured against the same hazards at the same time, the policies contain no provision permitting the Insurers to shift a portion of a covered loss back to the insured for uninsured periods. NGC emphasizes that the policies promise to pay “all sums” for which NGC becomes hable as a result of bodily injury sustained during the policy period, not some pro rata portion of damages based upon the percentage of time on the risk.

The Insurers argue that unless the pro-ration-to-the-insured approach is adopted, NGC would obtain a windfall that would treat it and other manufacturers that fail to purchase insurance in the same manner as manufacturers that purchase appropriate amounts of insurance. Moreover, the Insurers assert, their policies cover liability only for injury that occurs during the policy period, and not for injury before or after. The Insurers further argue that if NGC is permitted to choose the policy it prefers for indemnification of the entire injury, this would be unfair to an insurer who was on the risk for a short period but then is burdened with the entire loss incurred over several years. NGC responds that its approach does not result in a windfall because, although two insureds might receive the same amount of coverage for a particular claim, an insured with one year of coverage will exhaust its policy faster when confronted by 50 claims than an insured with 20 years of coverage confronted by the same number of claims. NGC further responds that obliging an insurer on the risk for a short period to be burdened with the entire loss is not unfair but simply the required consequence of a policy that promises to indemnify for “all sums” the insured is obligated to pay as a result of an injury.

Neither New York nor Texas courts have as yet decided whether to adopt the pro-ration-to-the-insured approach.¹⁷ However, we have the benefit of persuasive opinions by thoughtful judges, concluding that, in the context of multiple policies triggered for continuous injuries, proration-to-the-insured is a sensible way to interpret insurance policies that do not squarely resolve the allocation issue. Perhaps the leading opinion in the field, and surely one of the best reasoned, is Justice O’Hem’s opinion for the New Jersey Supreme Court in Owens-Illinois, 138 N.J. 437, 650 A.2d 974. He begins by rejecting both sides’ contentions based on the wording of the CGL policy. The “all sums” language relied on by the insured, he points out, “was never intended to cover apportionment when continuous injury occurs over multiple years.” Id. at 465-66, 650 A.2d at 988-89. And the insurers’ reliance on language limiting their coverage to injury “which occurs during the policy period” ignores their obligation to indemnify for subsequent damages attributable to an injury occurring during the relevant policy period. Id. He then relates the allocation issue to the continuous trigger approach and argues for an allocation formula that provides incentives to increasing available resources and internalizing costs. He concludes that

a fair method of allocation appears to be one that is related both to time on the risk and the degree of risk assumed. When periods of no insurance reflect a decision by an actor to assume or retain a risk, as opposed to periods when coverage for a risk is not available, to expect the risk-bearer to share in the allocation is reasonable.

Id. at 479, 650 A.2d at 995.

We agree with the analysis of the New Jersey Supreme Court and think it likely that New York and Texas will also agree. We also note that proration-to-the-insured, in the context of continuous triggering, has been approved by the Sixth Circuit, see Insurance Co. of North America v. Forty-Eight Insulations, Inc., 633 F.2d 1212, 1224-25 (6th Cir.1980), clarified, 657 F.2d 814, 816 (6th Cir.), cert. denied, 454 U.S. 1109, 102 S.Ct. 686, 70 L.Ed.2d 650 (1981), and by Judge Wald in the District of Columbia Circuit, see Keene Corp. v. Insurance Co. of North America, 667 F.2d 1034, 1057-58 (D.C.Cir.1981) (Wald, J., concurring); see also Uniroyal, Inc. v. Home Insurance Co., 707 F.Supp. 1368, 1392 (E.D.N.Y.1988) (“Self-insurance is called going bare for a reason.”).

We agree with Judge Martin that pro-ration-to-the-insured is a sensible way to adjust the competing contentions of the parties in the context of continuous triggering of multiple policies over an extended span of years. We agree that such proration is appropriate as to years in which NGC elected not to purchase insurance or purchased insufficient insurance, as demonstrated by the exhaustion .of its policy limits. However, we do not agree with the District Judge’s subsidiary ruling that proration-to-the-insured should be applied to years after 1985 when asbestos liability insurance was no longer available. Judge Martin applied proration-to-the-insured even after 1985. His rationale was that NGC had “bargained away coverage by accepting asbestos exclusion clauses.” We think that is not a realistic view of the situation. There is no reason to believe that any bargaining occurred with respect to the asbestos exclusion clauses.

Moreover, we note that judges who have endorsed proration-to-the-insured have done so only to oblige a manufacturer to accept a proportionate share of a risk that it elected to assume, either by declining to purchase available insurance or by purchasing what turned out to be an insufficient amount of insurance. Thus, Justice O’Hern’s opinion in Owens-Illinois explicitly contrasts its pro-ration approach to “periods when coverage for a risk is not available,” 138 N.J. at 479, 650 A.2d at 995. Similarly, Judge Wald endorsed proration-to-the-insured only “prior to the time when such coverage could no longer be obtained.” Keene, 667 F.2d at 1058 (emphasis in original).¹⁸ Judge Martin’s opinion appears to be the only one applying proration-to-the-insured to years when asbestos liability insurance was no longer available.

We therefore modify the judgments so as not to apply the proration-to-the-insured approach to years after 1985, the point at which asbestos liability insurance ceased to be available. There are two ways to apply this modification, however. These may be illustrated by the following example. Consider a disease process spanning 25 years, which first inflicted injury in 1981 and will result in a claim or death 24 years later in 2005. Assume that Insurer A issued a policy for 1981, Insurer B issued policies for 1982, 1983, and 1984, NGC was uninsured in 1985 (either because it elected not to buy insurance or because the limits of its coverage were exhausted), and no policies were issued for the 20 years from 1986-2005 (because of the unavailability of asbestos insurance).

Under Judge Martin’s method, Insurer A would have paid l&th of the claim, Insurer B would have paid feths, and NGC would have paid 2IMhs. One way of precluding pro-ration-to-the-insured for the post-1985 years would be to require Insurer A to pay jéth of the claim, Insurer B to pay %ths, and NGC to pay jéth; this method applies a fraction that has as its denominator the number of years in which both injury-in-fact was occurring and insurance was available, and as its numerator the number of years within that period when the insurance was in effect (treating NGC as a self-insurer). A second way would be to require Insurer A to pay %r,ths, Insurer B to pay x%6ths, and NGC to pay j&th; this method allocates to NGC one year’s share of the 25-year injury and allocates the remaining 24 years to Insurers A and B in proportion to their years on the risk. The second alternative obviously is the most beneficial to NGC.

We conclude that the first of these two alternatives should be followed, since it better comports with the principle of treating NGC as a self-insurer for the one year when insurance was available. Had NGC, in- this example, purchased a policy from Insurer C in 1985, that insurer would have been liable for jéth of the claim (since no proration-to-the-insured is being allowed for the years after 1985). Our solution imposes that same jéth share on NGC for the one year when it was uninsured due to its own choice.

We therefore implement proration-to-the-insured by obliging NGC to pay a share of each claim represented by a fraction that has as its denominator the number of years of the injury up to 1985, and as its numerator the number of those years in which NGC was uninsured (either because it purchased no insurance or its policy limits were exhausted).¹⁹

C. “Expected or Intended” Injury Defense

Through various formulations, NGC’s policies exclude coverage for bodily injuries that are expected or intended by NGC. For example, the 1974-77 CU policy provides coverage for NGC’s liability arising out of “personal injuries” caused by an “occurrence.” An occurrence, in turn, is defined as “an accident or a continuous or repeated exposure to conditions which results during the policy period in personal injury, property damage or advertising liability neither expected nor intended from the standpoint of the insured.” In both the jury and bench trials, the Insurers contended that coverage of the underlying asbestos-related bodily injury claims should be precluded because NGC “expected” or “intended” the injuries within the meaning of the “occurrence” definition. Both fact-finders rejected this contention, finding that NGC did not “expect” or “intend” the bodily injuries.

On appeal, Continental Casualty Company (“CCC”), joined by various other insurers,²⁰ points to four purported errors of law on this issue that it contends require reversal of the Jury Trial BI Judgment and the Bench Trial BI Judgment. First, these insurers argue that the trial court should not have placed on them the burden of proving that NGC expected or intended the underlying injuries. Allocation of the burden of proof is determined pursuant to New York law, even with respect to those policies whose substantive interpretation is governed by Texas law. See Woodling v. Garrett Corp., 818 F.2d 543, 552 (2d Cir.1987) (“The question of burden of proof ... is regarded by New York law as a question of procedure to which the law of the forum applies.”).

Because the exclusionary language of NGC’s policies is located in a section of the policy denominated “definitions” or “insuring agreement” rather than “exclusions,” the Insurers contend that NGC must bear the burden of proving that it did not expect or intend the injuries for which it seeks coverage. However, under New York law, the exclusionary effect of policy language, not its placement, controls allocation of the burden of proof. See Utica Mutual Insurance Co. v. Prudential Property and Casualty Insurance Co., 103 A.D.2d 60, 64, 477 N.Y.S.2d 657, 660 (2d Dep’t 1984) (“[I]t is the insurer which has the burden of proof to establish that a claim is encompassed by an exclusion in a policy ... and any limitation in coverage must be described in clear and explicit language.”), aff'd, 64 N.Y.2d 1049, 489 N.Y.S.2d 704, 478 N.E.2d 1305 (1985). Several decisions treat the “expected” or “intended” policy language as an exclusion. See, e.g., City of Johnstown v. Bankers Standard Insurance Co., 877 F.2d 1146 (2d Cir.1989); Town of Moreau v. Orkin Exterminating Co., 165 A.D.2d 415, 568 N.Y.S.2d 466 (3d Dep’t 1991).

Second, the Insurers argue that the trial court should not have applied a subjective standard with respect to whether NGC expected or intended injury. They contend that even if NGC did not actually expect or intend the underlying harm, NGC should be deprived of coverage if the evidence established that NGC “should have” expected such harm. However, the policies on their face preclude coverage only if NGC expected or intended the injuries, and not if NGC merely should have expected injury. See City of Johnstown, 877 F.2d at 1151 n. 1 (adopting subjective standard). Moreover, as we noted in City of Johnstown, ordinary negligence does not constitute an intention to cause damage, and a calculated risk does not amount to an expectation of damage. Id. at 1150. A number of courts applying the “expected” or “intended” policy language under Texas law have similarly inquired into the insured’s actual expectations or intent without consideration of whether the insured “should have” expected or intended the underlying harm. See, e.g., American Home Assurance Co. v. Safway Steel Products Co., 743 S.W.2d 693, 701 (Tex.Ct.App.1987) (“punitive damages arising out of the insured’s gross negligence are not ... excluded from coverage” by “expected or intended” clause).

Third, the Insurers argue that the form of the jury verdict concerning the “expected or intended” issue injected a prejudicial requirement that they prove on a product-by-product basis NGC’s expectations or intentions. This argument is meritless. On the jury verdict form, the jury was requested to evaluate NGC’s expectations or intentions with respect to persons exposed to asbestos (1) as a result of working in the NGC mine, (2) as a result of working in the NGC plant, (3) contained in non-spray plasterers and textures, (4) contained in asbestos cement board products, and (5) contained in joint compounds. The District Court’s instructions and the jury verdict form simply acknowledged that the product and workplace exposures to asbestos at issue in the underlying cases were different. Some products were more hazardous than others, and a finding that NGC knew that one product sold for only a few years was hazardous would not necessarily be sufficient to show that NGC “expected” or “intended” injuries to persons using different products at different times.

Thus, the jury could have determined that NGC’s expectations and intentions were sufficiently proved to justify a finding in favor of the Insurers regarding injuries of NGC’s own employees exposed in the course of manufacturing asbestos products, but not regarding injuries suffered by third parties’ employees as a result of exposure in the course of using NGC’s joint compound products. The Insurers were free, however, to argue that NGC’s knowledge of the hazards of asbestos in connection with one set of products and exposures should be deemed sufficient to support an inference of knowledge or expectation with respect to all sets of exposures or products.

Finally, CCC argues that the trial court erred in admitting the testimony of Russell L. Ward in the bench trial, and asserts that in the absence of Ward’s testimony, the only competent evidence proved beyond question that NGC expected or intended the underlying injuries. Ward was an NGC employee for 36 years. During that period of time, he was employed by NGC in various facets of NGC’s operations from credit and sales to production and distribution. In 1985, he began to serve as NGC’s records custodian, and reviewed most of the documents in NGC’s collection of asbestos-related documents. He also conversed with hundreds of NGC employees over the years. His testimony was introduced by NGC as evidence that NGC did not expect or intend to cause injury.

Though CCC raises various objections to this testimony, we conclude that the trier of fact reasonably could have found from the evidence that Ward possessed sufficient personal knowledge regarding NGC’s historical documents and regarding the lack of any showing therein of an expectation or intention that injuries would occur to persons who used NGC’s products. Consequently, the District Court did not abuse its discretion in determining that Ward could testify. We further conclude that even apart from Ward’s testimony, NGC presented sufficient evidence to support the trial court’s finding that NGC did not expect or intend the asbestos-related injuries.

D. The Wellington and CCR Issues

The Wellington Agreement was intended to efficiently resolve asbestos-related bodily injury claims asserted against the subscribing asbestos producers, any cross-claims among these asbestos producers concerning relative responsibility for the claims, and any disputes between these producers and their subscribing insurers regarding insurance coverage for asbestos-related bodily injury claims. The Wellington Agreement included a complex method for determining each producer’s share of liability for claims asserted against them.

Rather than assign individual claims to particular producers, the Wellington Agreement assessed each producer’s respective share of liability through the “producer allocation formula.” That formula was a mathematical attempt to reconstruct a particular producer’s fair share of liability, on an aggregate claims basis, based upon certain historical data (e.g., the frequency with which a producer was named as a defendant in bodily injury claims and the amounts paid by a producer in settlements or judgments). Although the formula assigned to a given producer a share of liability on all claims, including claims in which the producer was not named, this effect was counterbalanced by the fact that the other asbestos producers paid a share of all claims in which the producer was named.

The CCR Agreement, which established the CCR claims-handling facility following dissolution of the ACF, adopted a more detailed producer allocation formula, but similarly determined on an aggregate basis each participating producer’s share of liability for the bodily injury claims asserted against it. By joining the ACF and the CCR, NGC enjoyed a number of benefits, including a reduction in the number of attorneys retained to defend individual claims, the elimination of cross-claim litigation among producers concerning their respective shares of liability, and the ability to hire shared professional staff and reduce the expense of developing information regarding- each producer’s product. Both the jury and the trial court found, based on the evidence, that NGC’s decisions to participate in the ACF and the CCR claims-handling facilities and to abide by their respective producer allocation formulas were reasonable means of settling the asbestos-related bodily injury claims. As Judge Martin concluded, “the evidence is overwhelming that these facilities provided an economical way for National Gypsum to handle these claims, cut their legal expenses tremendously for their claims, and resulted in an overall cost reduction. And indeed, it would have been very difficult for National Gypsum to have handled these claims on any other basis.”

The Insurers continue to argue that NGC’s decisions to participate in the ACF and the CCR were unreasonable. First, they contend that NGC’s decision to participate in the ACF “transformed” NGC into a target for claimants, thereby causing an increase in claims asserted against NGC. The Insurers point to an increase in claims against NGC following its decision to join the ACF, but, as the District Court found, there is no reasonable causal connection. Second, the Insurers assert that NGC decided to participate in the ACF solely to maximize its insurance coverage. But even if this factor partly motivated NGC to participate in the ACF, there was ample evidence of the benefits to NGC resulting from such participation, apart from the settlement of insurance coverage disputes between producers and some of their liability insurers.

Third, the Insurers contend that they are not required to provide coverage for bodily injury claims in which NGC was not named as a defendant. The Insurers observe that although the Wellington signatory insurers agreed, as a compromise of potential coverage disputes, to reimburse NGC for payments on claims in which it was not named, the non-signatory insurers agreed in their policies to reimburse NGC only for NGC’s own proven liability to an individual claimant. Judge Martin, however, denied the Insurers’ motion for partial summary judgment on this issue, finding that the payments made by NGC pursuant to the ACF and CCR producer allocation formulas were “consistent with NGC’s rights under its policies to enter into reasonable and good faith settlements.”

These formulas require NGC to pay for its fair and proper share of liability for the claims asserted against NGC, and amounts billed to NGC on claims in which it is not named are offset by reductions in amounts billed to NGC on the claims in which it is named. Indeed, when the allocation formula was revised in 1991 so that CCR producers would be billed only on claims in which they were named, there was no appreciable effect on NGC’s payments. The Insurers argue that in those cases where NGC was not even named, there was no potential liability to NGC. However, the Insurers could have required that NGC seek their consent or approval of all proposed settlements. Moreover, had NGC been threatened with a class action suit where some members did not even have a claim against NGC, it might have been reasonable under the policies to settle with the entire class rather than attempt to defeat a handful of the claims asserted by class members. If the policy limits were huge, and NGC had declined to join the ACF and CCR claims-handling facilities and instead had chosen to settle claims on its own, the Insurers would now likely be arguing that NGC’s failure to use these settlement mechanisms was unreasonable.

Finally, as discussed above, the Insurers object to the District Court’s decision to require full indemnification even though some of the payments made by NGC pursuant to the Wellington and GCR agreements were to claimants who had been exposed to NGC’s products only after NGC’s policies had expired. For the reasons stated earlier, we conclude that the District Court’s ruling was proper.

III. Issues Relating to Coverage for Property Damage Claims

A. Existence of Property Damage

Prior to 1973, NGC’s liability policies typically provided coverage for “all sums which the insured shall become legally obligated to pay ... for damages ... on account of ... property damage.” “Property damage” in turn was defined as “loss of or direct damage to or destruction of tangible property.”

Beginning in 1973, these policies were revised to provide: “The company will pay on behalf of the insured all sums which the insured shall become legally obligated to pay as damages because of ... property damage ... caused by an occurrence.” “Property damage” was defined as “(1) physical injury to or destruction of tangible property which occurs during the policy period, including the loss of use thereof at any time resulting therefrom, or (2) loss of use of tangible property which has not been physically injured or destroyed provided such loss of use is caused by an occurrence during the policy period.”

Many asbestos-in-building claimants seek injunctions requiring NGC to test for and remove ACMs in schools or other buildings. Others seek equitable restitution of costs that claimants have already incurred to test for and remove ACMs. Still others seek a declaration that NGC must indemnify the claimants for costs that they will incur in responding to potential future claims by persons who may be exposed to ACMs present in the claimants’ buildings.’ The Insurers contended in the District Court, and ■ now argue on appeal, that the underlying complaints do not allege “property damage” within the meaning of the policies.

The District Court granted summary judgment to NGC on this issue, after determining that the underlying claims against NGC seeking damages in the form of costs of inspection, containment, removal, and replacement of ACMs in buildings “are for property damage and hence are covered claims under the policies.” Based on the evidence, including the allegations in the complaints of the building owners, the District Court found that once installed, the ACMs incorporated into the building deteriorate and release asbestos fibers through the structure into which they were installed, contaminating not only the structure itself but also its contents. Accordingly, building owners must remove or contain the ACMs in the structures to prevent the ACMs from further damaging the structures and their contents.

The Insurers argue that the underlying claims are merely for the costs of prophylactic measures undertaken by the building owners, and that such costs represent only the expenses incurred to comply with government-imposed health and safety standards. See Maryland Casualty Co. v. Armco, Inc., 822 F.2d 1348, 1353 (4th Cir.1987) (costs incurred in taking preventive measures to avert possible future health risks do not represent current “property damage”), cert. denied, 484 U.S. 1008, 108 S.Ct. 703, 98 L.Ed.2d 654 (1988).

Nevertheless, the claimants’ buildings have suffered “physical injury” as a result of installation of the ACMs, and the costs incurred to remedy the asbestos hazard and protect against future releases are measures of the damages resulting from that injury. See Grace, 23 F.3d at 627 (“[A]ctual injury to property — the presence of the asbestos hazard- — occurs upon installation and exists regardless of whether it yet has been discovered by the building owners.”). As the District Court recognized, “[T]he damage is from the release of fibers into the buildings, not from the regulatory requirement that the asbestos be removed, even though the cost of removal may be an appropriate item of damages.”

The Insurers also argue that claims for costs of repairing or replacing defective products are not covered “property damage” under the policies but merely non-covered economic loss. However, the claims asserted against NGC go beyond allegations that the asbestos materials failed to perform as intended, or that there is a risk of future harm; rather, these claims allege that the ACMs have already caused injury to other property, namely the structures into which the ACMs were incorporated and the contents of those structures. Such claims are for “property damage,” not economic loss. See Avondale Industries, Inc. v. Travelers Indemnity Co., 887 F.2d 1200, 1206-07 (2d Cir.1989) (cleanup costs of toxic waste were covered damages under CGL policies), reh’g denied, 894 F.2d 498 (2d Cir.), cert. denied, 496 U.S. 906, 110 S.Ct. 2588, 110 L.Ed.2d 269 (1990); City of New York v. Keene Corp., No. 84-44559 (Sup.Ct.N.Y.County, Nov. 19, 1985) (refitting of city’s buildings contaminated by asbestos, to make them safe, was measure of plaintiffs’ property damage), aff'd, 129 A.D.2d 1019, 513 N.Y.S.2d 1004 (1st Dep’t 1987).

Though the case law of Texas is ambiguous with respect to the status of asbestos-in-bmlding claims, see W.R. Grace & Co. v. Continental Casualty Co., 896 F.2d 865, 875 (5th Cir.1990) (“It is not clear [under Texas law] that costs associated with such removal [of asbestos products] should be treated as property damage.”), we agree with the District Court’s prediction that Texas, like New York, would regard such claims as claims for property damage. And although under New York law an insurer’s duty to indemnify is determined by the “actual basis” for a third party’s claims against the insured, rather than the allegations in the complaint, see Servidone Construction Corp. v. Security Insurance Co. of Hartford, 64 N.Y.2d 419, 488 N.Y.S.2d 139, 477 N.E.2d 441 (1985), the District Court did not err in concluding that the asbestos-in-building claims and the settlements paid by NGC involved “property damage” within the meaning of the policies.

B. Property Damage Trigger of Coverage

In its summary judgment ruling, the District Court determined that under the law of New York and Texas, property damage within the meaning of the policies occurs at the time of installation of the ACMs into a building and that policies in effect during that time period are triggered. However, the District Court also determined that property damage does not occur within the meaning of the policies in effect at any time after installation of the ACMs and that such policies are not triggered.

CU alone challenges the finding that its policies are triggered at the date of installation. CU argues that under New York law, mere installation of a defective component into other property does not comprise “property damage” so as to trigger coverage. Under CU’s argument, no coverage is triggered until damage — the defective nature of asbestos-containing products — is realized. However, courts routinely have found that installation of ACMs constitutes “property damage” and that policies in effect at that time are triggered. See Grace, 23 F.3d at 627 (“[Installation of asbestos is an occurrence of damage-in-fact and triggers insurance coverage in effect at that time.”).

CU contends that the damage asserted by the underlying claimants — the reduction in the value of their buildings — by definition cannot occur at the date of installation because, so long as the hazards posed by installation of the ACMs remain unknown, the reduction in value of the building and the decrease in its marketability cannot occur. CU thus urges this Court to apply a manifestation-only trigger, relying primarily on Sturges Manufacturing Co. v. Utica Mutual Insurance Co., 37 N.Y.2d 69, 371 N.Y.S.2d 444, 332 N.E.2d 319 (1975). However, Sturg-es did not consider the question of when property damage occurred, and nothing in Sturges suggests that only upon discovery of diminished value is coverage triggered. See id. 371 N.Y.S.2d at 447, 332 N.E.2d at 321-22 (property may be “damaged” if defective product is sufficiently integrated into property and if, as result, value of property is reduced beyond cost of replacing product).

Moreover, the policies respond to “damages” incurred because of “property damage” that takes place during the policy period. Thus, the fact that some of the damages sought by the underlying claimants — reduction in the buildings’ value — may not “occur” until the alleged dangers of the ACMs are perceived by the market does not preclude policies in effect prior to such time from being triggered by the existence of undiscovered property damage. See Grace, 23 F.3d at 627 (“A reduction in marketability serves as a measure of damages, but it does not constitute an injury to property itself.”).

Although the District Court ruled that installation of the ACMs constitutes property damage covered by the policies, the Court also determined on motions for summary judgment that, as a matter of law, NGC’s proof could not establish property damage after installation. The District Court first expressed doubt that “there is actual or potential release of additional asbestos fiber the entire time the asbestos containing material is present in the building.” But even if NGC could demonstrate that fibers are being released continuously, the Court held, the damage “flowing from the installation of ACM[s], i.e., the costs of removing or encapsulating the in-place asbestos[,] ... does not arise incrementally, cumulatively, or continuously, and is in no sense an injurious process during the course of which the property damage somehow ... progresses from an incipient to a fully developed state.” In Grace, we reached a similar result, holding that after installation, “[n]o further property damage occurs because the need to remove or encapsulate the asbestos, which occurred upon the product’s installation, remains unchanged.” 23 F.3d at 628.

NGC argues that the District Court’s summary judgment ruling improperly resolved questions of fact and ignored the damage to property within the structures that continues even after installation as a result of the ongoing release of asbestos fibers. NGC also asserts that Grace is distinguishable because in that case, the insured did not contend, as does NGC, that additional property within the buildings is contaminated even after the date of installation, and that the insured is liable for that damage. Consequently, NGC argues that notwithstanding our holding in Grace, the property damage was continuous, and insurance coverage should apply from the time of installation to the time of abatement.

We reject NGC’s arguments. The District Court properly concluded that, as a matter of law, NGC’s proof could not establish that property damage was more likely than not occurring during a particular policy period after the date of installation. NGC merely offered to present evidence that the release of fibers was continuous and that therefore property must continue to be contaminated. But, as the District Court observed, even if NGC established a continuous release of fibers, this was not a sufficient basis for finding that property damage occurred at a given post-installation point, since NGC could not show that each release of fibers more likely than not caused property damage. That NGC might be liable for property damage occurring after installation was irrelevant to the question of which policies were triggered. In view of the fact that NGC failed to offer evidence that property damage, as distinguished from the mere release of fibers, occurred continuously, the District Court did not err in ruling, on summary judgment, that only the date of installation was relevant for purposes of determining the trigger.

C. “Business Risk” Property Damage Exclusions

The Insurers invoke several “business risk” policy exclusions that they contend operate to bar coverage for some or all of NGC’s liability to the asbestos-in-building claimants. We agree with the District Court that none of these exclusions bars NGC from coverage.

The first of these exclusions is the “own product/work product” exclusion, which excludes coverage for “property damage to the named insured’s products arising out of such products or any part of such products; [or] ... property damage to work performed by or on behalf of the named insured arising out of the work or any portion thereof, or out of materials, parts or equipment furnished in connection thereof.” This exclusion bars coverage for property damage to NGC’s own products. The exclusion has no application here because the asbestos-m-building claimants seek to recover for damage to property other than NGC’s products. Such damaged property includes the building structures in which the ACMs were installed as well as property injured by contact with asbestos fibers. See Dayton Independent School District v. National Gypsum Co., 682 F.Supp. 1403, 1412 (E.D.Tex.1988), rev’d on other grounds sub nom. W.R. Grace & Co. v. Continental Casualty Co., 896 F.2d 865 (5th Cir.1990).

The second exclusion the Insurers invoke is the “product reeall/sistership exclusion,” one typical version of which bars coverage for “damages claimed for the withdrawal, inspection, repair, replacement or loss of use of the named insured’s products or work completed by or for the named insured or for any property of which such products or work form a part, if such products, work or property are withdrawn from the market or from use because of any known or suspected defect or deficiency therein.” Under New York law, the exclusion does not preclude coverage in this case, because NGC has not itself withdrawn or removed its products from the market. See Thomas J. Lipton, Inc. v. Liberty Mutual Insurance Co., 34 N.Y.2d 356, 360, 357 N.Y.S.2d 705, 707, 314 N.E.2d 37, 38 (1974) (exclusion precludes coverage only if insured itself withdrew product from market). Texas law appears to be in accord with that of New York on this point. See Parker Products, Inc. v. Gulf Insurance Co., 486 S.W.2d 610, 612-15 (Tex.Ct.App.1972), aff'd, 498 S.W.2d 676 (Tex.1973).

Moreover, even if damage caused by asbestos precipitated a discontinuance of NGC’s products, the exclusion would not preclude coverage for damage to property other than NGC’s products. See Todd Shipyards Corp. v. Turbine Service, Inc., 674 F.2d 401, 419 (5th Cir.) (exclusion provision does not bar coverage for “damages caused by the very product whose failure to perform properly aroused apprehension about the quality of ‘sister’ products”), cert. denied, 459 U.S. 1036, 103 S.Ct. 447, 448, 74 L.Ed.2d 602, 603 (1982). The District Court recognized that NGC’s liability, as alleged in the complaints, arises out of damage that its products have caused to third-party property.

AMICO points to another version of the sistership exclusion found in its 1973-76 umbrella policy.²¹ That version of the exclusion bars coverage for “such part of any damages or expense which represents the cost of inspecting, repairing, replacing, removing, recovering, withdrawing from use or loss of use of, because of any known or suspected defect or deficiency therein, any (1) goods or products or any part thereof (including any container) manufactured, sold, handled or distributed by the named insured or by others trading under his name; or (2) work completed by or for the named insured; or (3) other property of which such goods, products or work completed are a component part or ingredient.”

AMICO argues that the exclusion bars coverage not only for the cost of inspecting, repairing, removing, or replacing NGC’s allegedly defective product, but also for the same costs with respect to the buddings of which such products are a component part. Thus, AMICO’s position is that there is no coverage regardless of whether the damage is to NGC’s products or to other parts of the buildings in which those products were installed. But AMICO’s policies do not define the term “defect,” and the claimants do not allege that NGC’s products faded to perform their insulating or fire retardant functions. See United States Fidelity & Guaranty Co. v. Wilkin Insulation Co., 144 Ill.2d 64, 161 Ill.Dec. 280, 289, 578 N.E.2d 926, 935 (1991); Dayton Independent School District v. National Gypsum Co., 682 F.Supp. at 1412. In view of the case law narrowly construing the sistership exclusion, and the ambiguity in the relevant policy language, we hold that NGC’s products, though capable of injuring people and property, do not contain a “defect or deficiency” within the meaning of the AMI-CO exclusion, which applies only to costs incurred because the insured’s product fads to perform its intended functions. See American Motorists Insurance Co. v. Trane Co., 544 F.Supp. 669, 693-95 (W.D.Wis.1982), aff'd, 718 F.2d 842 (7th Cir.1983).

The third policy exclusion invoked by the Insurers is the “warranty” or “loss of use” exclusion. That exclusion precludes coverage for “loss of use of tangible property which has not been physicady injured or destroyed resulting from (1) a delay in or lack of performance by or on behalf of the named insured of any contract or agreement, or (2) the failure of the named insured’s products or work performed by or on behalf of the named insured to meet the level of performance, quality, fitness or durability warranted or represented by the named insured; but this exclusion does not apply to loss of use of other tangible property resulting from the sudden and accidental physical injury to or destruction of the named insured’s product or work performed by or on behalf of the named insured after such products or work have been put to use by any person or organization other than an insured.”

Again, however, the exclusion is inapplicable because NGC’s liability does not arise from the failure of its products to perform their fireproofing functions. Rather, the claimants allege that, in serving its intended purpose, the insulation product released asbestos fibers, causing property damage to the buildings and contents therein. In addition, the exclusion does not apply where the property, the use of which is lost, has been physically injured. NGC’s liability arises out of physical injury to property by virtue of the release and re-entrainment of asbestos fibers.

Finally, the Insurers rely on a provision that they term “Home exclusion (b).” This exclusion bars coverage for claims made against the insured “(i) for repairing or replacing any defective product or products manufactured, sold, or supplied by the insured or any defective part or parts thereof nor for the cost of such repair or replacement; (ii) for the loss of use of any such defective product or products or part or parts thereof; (iii) for improper or inadequate performance, design, or specification; but nothing herein contained shall be construed to exclude claims made against the insured for personal injuries or property damage (other than damage to the product of the insured) resulting from improper or inadequate performance, design, or specification.”

This exclusion, too, does not bar coverage for the underlying asbestos-in-building claims. Subpart (i), like the “own product” exclusion, concerns the repair or replacement of the insured’s own products. As explained above, the underlying claims allege damage to property other than NGC’s product, rendering this exclusion inapplicable. Subpart (ii) excludes only claims for loss of use of the product in the absence of damage to other property, whereas the underlying claims involve allegations of damage to other property. Subpart (iii), like the “warranty” exclusion, bars coverage only for a product’s improper performance, and it is inapplicable for the same reasons as the warranty exclusion. The exclusion is thus intended to preclude coverage only for the loss of use, or the repair and replacement of, a product that is defective in the sense that it fails to perform its intended function. The failure of asbestos-containing products to perform as building materials is not at issue; rather, the issue is the contamination of a building with hazardous asbestos fibers, and the property damage caused by that contamination.

In sum, we conclude that none of the exclusions is applicable to any portion of the costs incurred by the asbestos-in-buüding claimants.

D. Number of Occurrences

Certain of NGC’s primary policies contain a deductible feature, typically providing that “[t]he deductible amount applies to all damages and claim expenses for all coverages combined as the result of any one occurrence.” For example, from 1976 through 1980, AMICO’s primary liability policies required NGC to assume responsibility for the first $250,000 in damages resulting from any one occurrence (with a $500,000 liability limit per occurrence and a $1 million aggregate policy limit).

The District Court determined that, under the ease law of both New York and Texas, the multiple asbestos-related property damage claims being asserted against NGC— involving thousands of buildings, constructed by many owners around the country over the course of half a century or more — arose out of a single occurrence, requiring only one deductible per applicable policy period. That single occurrence, the Court ruled, was NGC’s decision to manufacture and sell asbestos-containing products.

AMICO, joined by several of the Insurers,argues that each construction , or renovation of a building involving the installation of NGC’s products is a separate occurrence requiring NGC to pay a separate deductible. AMICO contends that our holding in Grace, that installation is the occurrence giving rise to coverage,. cannot be reconciled with the District Court’s conclusion that the sole occurrence is NGC’s overall course of manufacturing and selling asbestos products over many decades. We agree.

In determining the number of occurrences for deductible purposes, New York inquires whether .multiple claims result from “‘an event of an unfortunate character that takes place without one’s foresight or expectation.’ ” See Arthur A. Johnson Corp. v. Indemnity Insurance Co., 7 N.Y.2d 222, 228, 196 N.Y.S.2d 678, 683, 164 N.E.2d 704, 707 (1959) (quoting Croshier v. Levitt, 5 N.Y.2d 259, 268, 184 N.Y.S.2d 321, 328-29, 157 N.E.2d 486, 491 (1959)). Texas applies a similar “cause” test. See, e.g., Carpenter Plastering Co. v. Puritan Ins. Co., No. 87-2435, 1988 WL 156829, at *4 (N.D.Tex. Aug. 23, 1988). Texas appears to find multiple occurrences under circumstances similar to those in this case, see, e.g., Maurice Pincoffs Co. v. St. Paul Fire & Marine Ins. Co., 447 F.2d 204 (5th Cir.1971) (separate shipments of contaminated seed to eight dealers constituted eight separate occurrences).

However, in Champion International Corp. v. Continental Casualty Co., 546 F.2d 502, 506 (2d Cir.1976), cert. denied, 434 U.S. 819, 98 S.Ct. 59, 54 L.Ed.2d 75 (1977), a case involving New York law, we held that claims based on installation of defective vinyl panels in more than a thousand vehicles arose from a single occurrence, the insured’s delivery of the panels to the manufacturers of the vehicles. Similarly, other courts have found one occurrence in the mass tort context, a result that maximizes coverage where the insured is confronted with numerous claims that, considered separately, would not exceed the deductible amount. See, e.g., Uniroyal, 707 F.Supp. at 1383 (rejecting insurers’ argument that each spraying of Agent Orange was separate occurrence subject to “per occurrence” deductible, and holding that single occurrence was delivery of herbicide by policyholder to United States military); Owens-Illinois, Inc. v. Aetna Casualty & Surety Co., 597 F.Supp. 1515, 1527 (D.D.C.1984) (manufacture and sale of asbestos-containing products constitutes single occurrence, notwithstanding number of claims asserted against manufacturer).

Nevertheless, the case law of New York indicates that, although a single “occurrence” may give rise to multiple claims, see, e.g., Weissblum v. Glens Falls Insurance Co., 40 Misc.2d 964, 244 N.Y.S.2d 689 (Sup.Ct.App. Term, 1st Dep’t 1963) (lights broken at different times by different persons during construction work constituted single occurrence), courts should look to the event for which the insured is held liable, not some point further back in the causal chain. See Arthur A. Johnson Corp., 7 N.Y.2d at 230, 196 N.Y.S.2d at 684, 164 N.E.2d at 708 (separate occurrences where rainfall caused separate collapses of two retaining walls on construction site). This is consistent with the policy language, which defines “occurrence” as “an accident, including continuous or repeated exposure to conditions, which results in bodily injury or property damage, neither expected nor intended from the standpoint of the insured.” In this ease, the “accident” resulting in property damage is most reasonably understood not as NGC’s general decision to manufacture asbestos but rather as the installation of the ACMs. Each installation created exposure to “a condition which resulted in property damage neither expected nor intended from the standpoint of the insured,” and for each installation, there was a new exposure and another occurrence.

Notwithstanding the broad reading of “occurrence” in Champion, our holding in that case does not require a single-occurrence result here. In Champion, the insured was not the manufacturer of the defective product but the wholesaler, who delivered the product to 26 manufacturers, who in turn distributed the product to consumers. We declined to consider the possibility that each delivery of the product to a manufacturer might constitute a separate occurrence, because the 26 manufacturers had not sought indemnification from Champion. Instead, we limited our analysis to two possibilities: the 1,400 vehicles in which the products were installed, and what we viewed as a single delivery of the products. See 546 F.2d at 505; see also id. at 507 (Newman, J., dissenting) (noting third possibility of 26 deliveries to different manufacturers); Michigan Chemical Corp. v. American Home Assurance Co., 728 F.2d 374, 383 n. 10 (6th Cir.1984) (“[T]he court in Champion did not advert to the possibility that 26 occurrences had taken place.”). Champion does not require a ruling of one occurrence in the present case.

Nor does Champion require us to focus on NGC’s sale or delivery of its asbestos products, rather than on the installation or presence of those products in buildings. Unlike Champion and other eases relying on sale or delivery as the relevant event that exposes the insured to liability, see, e.g., Pincoffs, 447 F.2d at 207, this ease involves not a distributor but the manufacturer of the defective product. In Champion, the insured was exposed to liability merely because it had delivered the defective product; in this case, by contrast, NGC’s liability, as reflected in the underlying complaints, results not from its delivery of asbestos-containing products, but rather from its manufacture of those products, resulting in the presence of ACMs each time the products were installed on the property of third parties. Consequently, each location at which NGC’s products are present, reflecting a separate installation of those products, is the site of a separate occurrence requiring imposition of another deductible.²²

NGC argues that although “occurrence” should be interpreted to mean each separate installation for purposes of determining policy limits, see Grace, 23 F.3d at 627 (installation of asbestos in building represents “occurrence of damage-in-fact”), we should find only one occurrence for deductible purposes because the deductible amount will in many cases exceed the amount sought with respect to each installation. Though we recognize the maxim of construing ambiguous provisions in favor of policy-holders, we do not find the “per occurrence” deductible provisions ambiguous, and we decline to interpret those provisions in a manner contrary to the intent of the parties, simply to achieve a result favorable to the insured. The policy language and case law provide no basis for aggregating events widely separated in time and space into one “occurrence.”

IV. Issues Related to Coverage for All Claims

A. “Known Loss” Defense

Several insurers contend that they are exempted from indemnity obligations because of the “known loss” defense — the insurance law principle that an insured may not obtain insurance to cover a loss that is known before the policy takes effect. See, e.g., Bartholomew v. Appalachian Insurance Co., 655 F.2d 27, 28-29 (1st Cir.1981). The issue is presented by CSIC, which issued 1983 and 1984 policies to NGC for $15 million of excess coverage. Though CSIC’s exposure is ended by our affirmance of Judge Martin’s estoppel ruling as to the Excepted Insurers (other than ACIC), see Part 11(A)(7), supra, the issue remains alive as to Stonewall and OR, which have joined CSIC’s “known loss” defense brief. The issue is also pressed on appeal by CU, though its argument differs from that of CSIC. See Brief for Appellanf/Cross-Appellee CU at 69-70.

In the District Court, the “known loss” defense was raised by CSIC in a motion for summary judgment considered by Magistrate Judge Bemikow. Discussing the issue in the context of property damage, he recommended rejecting the defense except for cases filed, or for which NGC received a pre-suit demand letter, before the policies incept-ed. Judge Martin accepted this recommendation, observing that, despite the extent of awareness about potential asbestos claims, “all that can fairly be said is that there was a known risk that there would be losses extending into the policy period and, therefore, obtaining insurance was a sensible business judgment.”

CU’s related argument that triggering must end at manifestation because at that point the injury is known appears not to have been raised in the District Court. The jury verdict form offered the jurors several choices to describe the triggering they found to have occurred. The choice they selected was worded, “Bodily injury, sickness or disease occurs at the time of the first exposure to asbestos and continues or progresses continuously thereafter.” There is no indication that CU requested this choice to be reworded to end at manifestation. The Jury Trial BI Judgment, entered on the jury’s verdict, stated the ending date of triggering as “the earlier of the claimant’s filing of a claim or death.” The Bench Trial BI Judgment also stated the ending date of triggering to be “the earlier of the claimant’s filing of a claim or death,” but also added “unless the insurer can show that after the claimant’s first diagnosis but prior to the earlier of the filing or death date, the disease had stopped progressing.” There is no indication that CU objected to the wording of these judgments.

NGC broadly contends, citing City of Johnstown v. Bankers Standard Insurance Co., 877 F.2d 1146 (2d Cir.1989), that the “known loss” defense exists only with respect to first-party insurance and does not even apply to third-party liability insurance. It also contends that this defense is merely another way of arguing that the losses were “expected or intended” and therefore not within the definition of “occurrence,” an argument we have already rejected. See Part 11(C).

City of Johnstown, however, does not purport to rule that the “known loss” defense is unavailable in third-party insurance. Though the classic statement of the defense is the observation that first-party insurance cannot be validly purchased for a home that has already burned down, it is not readily apparent why the defense should not also apply in the third-party context. City of Johnstown did not hold that insurance can be validly purchased to indemnify against a third party’s specific loss that the insurance buyer knows has already occurred. What City of Johnstown decided is that an insured’s knowledge of a risk of losses does not bar indemnity coverage. Id. at 1152-63; see Gulf Chemical & Metallurgical Corp. v. Associated Metals & Minerals Corp., 1 F.3d 365, 369-70 (5th Cir.1993) (following City of Johnstown).

Nor do we agree with NGC’s contention that the “known loss” defense is merely another way of claiming that the injury was “expected or intended” within the meaning of the “occurrence” definition. The “expected or intended” claim requires consideration of whether, at the time of the acts causing the injury, the insured expected or intended the injury, an inquiry that generally asks merely whether the injury was accidental. See City of Johnstown, 877 F.2d at 1150. The “known loss” defense requires consideration of whether, at the time the insured bought the policy (or the policy incepted), the loss was known. The contentions may overlap, but they are distinct, as the separate discussions in City of Johnstown recognized.

Though NGC was aware, prior to the inception of many of the policies, that its products risked asbestosis and cancer diseases and had received a large number of claims, it was highly uncertain, as Judge Martin found, as to the prospective number of injuries, the number of claims, the likelihood of successful claims, and the amount of ultimate losses it would be called upon to pay. NGC was fully entitled to replace the uncertainty of its exposure with the precision of insurance premiums and leave it to the insurers’ underwriters to determine the appropriate premiums.²³ See Gulf Chemical, 1 F.3d at 369-70 (applying Texas law); Montrose Chemical, 10 Cal.4th at 663, 42 Cal.Rptr.2d at 332, 897 P.2d at 9.

The cases relied on by CSIC are inappo-site. Bartholomew v. Appalachian Insurance Co., 655 F.2d 27 (1st Cir.1981), involved an attempt to obtain coverage for a loss for which suit had been filed before the policy took effect. Id. at 29. Appalachian Insurance Co. v. Liberty Mutual Insurance Co., 676 F.2d 56 (3d Cir.1982), involved what the court considered to be a single occurrence that had injured women who had filed an EEOC complaint prior to the policy period. Carpenter Plastering Co. v. Puritan Insurance Co., CIV.A. No. 3-87-2435-R, 1988 WL 156829 (N.D.Tex. Aug. 23, 1988), involved a claim for damages resulting from a specific sale of defective wall panels, which the insured knew had begun to crack before the policy periods began.

The District Court’s findings concerning the extent of unknown liabilities support rejection of the “known loss” defense. CU’s related contention that triggering, even if occurring in successive policy periods, must end at manifestation is rejected for lack of objection to the wording of the jury verdict forms or the judgments, and because the injury-in-fact approach permits coverage for injuries continuing to occur to unidentified victims after manifestation, so long as a particular victim’s manifestation was not known to the insured.

B. Aggregate Policy Limits of Short-Term Policies

Certain policies issued to NGC by its historic insurers were either canceled before the stated policy period had expired (e.g., Stonewall) or extended for a period beyond the stated expiration date (e.g., Republic), leaving them in effect for a fractional annual period (“short-term policies”). The District Court decided that, as a general rule, the full annual aggregate limits of these short-term policies were available to NGC for whatever fractional period they were in effect, because there was no provision in the policies for the reduction of aggregate limits in the event of cancellation. The District Court reasoned that if NGC canceled a policy halfway through the year and the occurrence happened during the part of the year in which the policy was in effect, the insurer should be liable for the full aggregate limit.

However, the District Court created an exception for the policy of one excess insurer, Home Insurance Co. (“Home”), that was canceled in mid-year and replaced with another policy issued by the same insurer with higher aggregate limits. The District Court distinguished the circumstances of this policy on the ground that the others involved “either a total change in the insurance carrier or a change in the level at which the carrier became liable,” rather than merely a change in the aggregate limit. The District Court subsequently enlarged this exception to include the policy of another insurer, CU, that was excess to the Home policy. On appeal, Stonewall challenges the District Court’s general ruling and wishes to be included in the exception from that general ruling, Republic challenges the District Court’s general ruling, and NGC asks this Court to affirm the District Court’s general ruling but reverse the exclusion of CU from that general ruling. We affirm all of the rulings regarding short-term policy limits.

For an advance premium of $171,-680, Stonewall issued an excess umbrella policy to NGC, effective April 25, 1978, to July 1,1979, with occurrence and aggregate limits of $5 million. The Stonewall policy states that its liability limits of $5 million are available “in the aggregate for each annual period during the currency of this policy.” The term “annual period” is not defined anywhere in the Stonewall policy, nor does the cancellation section contain any provision stating the consequences of a fractional period upon aggregate limits. The Stonewall policy was canceled “pro rate” effective January 1, 1979, and $71,920 of the premium was refunded to NGC. NGC then purchased a second excess umbrella policy from Stonewall, effective January 1, 1979, to January 1, 1980, for a premium of $155,000. Like the first Stonewall policy, the second Stonewall policy had occurrence and aggregate limits of $5 million.

The District Court’s ruling with respect to Stonewall was proper. Nothing in the language of Stonewall’s policies provides for proration of annual aggregate limits where the policy is in effect for only a fraction of a year. Stonewall’s policies are silent on the consequences of cancellation, making this another ambiguity to be resolved against the insurer. See Breed v. Insurance Co. of North America, 46 N.Y.2d 351, 357, 413 N.Y.S.2d 352, 357, 385 N.E.2d 1280, 1284 (1978) (ambiguities should be construed in favor of policy-holder); Southern Life & Health Insurance Co. v. Simon, 416 S.W.2d 793, 795 (Tex.1967) (same). Moreover, in exchange for a reduction in premiums, Stonewall received a reduction in the amount of time it was on the risk.

With respect to Republic’s policy, Republic originally issued an excess umbrella policy to NGC covering the period January 1, 1977, to January 1,1978. The Republic policy states that its limits of $5 million are available “in the aggregate for each annual period during the currency of this policy.” As in the Stonewall policies, the term “annual period” is not defined in the Republic policy, nor does the Republic policy contain any provision stating the consequences of a fractional annual period upon the policy’s aggregate limits. The advance premium paid by NGC for the policy was $129,032. In consideration of an additional premium of $53,004, coverage under the policy was extended to April 25,1978.

Republic argues that the annual period of its policy was simply extended by four months without creating a second annual period with a full $5 million limit available. However, in light of both the ambiguity in the policy language and the premium amount paid by NGC, which was roughly equivalent to one third of a year of full coverage, it was reasonable for the District Court to conclude that NGC intended to purchase an additional policy with the full $5 million limit applicable for that four-month period.

Finally, with respect to Home and CU,²⁴ a Home policy was issued to NGC effective January 1, 1965, to January 1, 1968, with an annual aggregate limit of $2 million in excess of primary coverage. NGC paid a premium of $26,326 for this three-year policy. Effective October 4, 1967, this policy was canceled and replaced with another Home policy, which was effective October 4,1967, to January 1, 1971, with limits of $5 million in excess of primary coverage. As consideration for these higher limits, NGC paid a larger premium of $32,330.22 for the three-year-and-three-month replacement policy.

At this same time, CU also issued a policy to NGC for $3 million of additional excess coverage. For this three-year policy, which was effective for the same period as that of the first Home policy (January 1, 1965, to January 1, 1968), NGC paid a premium of $4,970. Like the first Home policy, the first CU policy was canceled effective October 4, 1967, and replaced with a new CU policy, effective October 4, 1967, to January 1, 1971, with limits of $5 million in excess of primary coverage. As consideration for these higher limits, NGC paid a larger premium for the three-year-and-three-month replacement policy. The CU policy states that it is “subject to all the terms and conditions of policies ... issued by [Home].”

The Home policy states that the aggregate limits apply “for each annual period where applicable.” In turn, the term “annual period” is defined for each policy as “each consecutive period of one year commencing from the inception date of this Policy.” Like the other policies, nothing in the Home policy’s cancellation clause provides for any decrease in aggregate limits as a consequence of cancellation prior to the end of a period. Nevertheless, the District Court concluded that NGC had canceled and replaced the Home policy simply to increase the amount of its excess coverage from $2 million to $5 million rather than to obtain a $7 million limit for the replacement period. We do not regard the District Court’s assessment of the circumstances of the transaction between NGC and Home and of the parties’ intent to be erroneous.

NGC also contests the District Court’s decision to extend to CU the ruling with respect to Home. Notwithstanding the fact that CU’s policy was an upper-level excess policy while Home’s policy was a lower-level umbrella policy, the District Court’s determination regarding CU was proper. Home and CU used the same form and, as with the Home policy, the amount of NGC’s excess coverage increased under the CU replacement policy, suggesting that NGC was merely increasing the amount of excess coverage available for the replacement period from $3 million to $5 million, rather than adding $5 million to the $3 million limit already in place.

C. Excess Policy Defense Obligations

With respect to 27 different excess policies, the District Court made rulings as to whether each insurer (1) owed a duty to defend NGC, (2) owed a duty to reimburse defense costs incurred by NGC, or (3) owed no defense obligation whatsoever.

As to the duty to defend, the District Court held that these excess insurers must defend NGC under the 1981 Transit policy and the 1982 and 1983 ACIC policies. In reaching this conclusion, the District Court found that the Texas Amendatory Endorsement to the 1981 Transit and 1982/83 ACIC policies (“the Texas Endorsement”) abrogates only the duty to defend claims arising in Texas. Both the Transit and ACIC policies state that the insurer shall be obligated to assume NGC’s defense if the underlying insurance is exhausted. The Texas Endorsement amends the Transit policy to state that Transit “may, at the sole option of the company, assume charge of the ... defense,” and it modifies the ACIC policies to read “the Company shall not be obligated to assume” NGC’s defense. Though the Insurers con-, tend that the Texas Endorsement abrogates their duty to defend NGC, and does not merely apply to claims arising in Texas, we find no basis for overturning the District Court’s narrow interpretation of that endorsement, which was supported by a review of similar policy language, affidavits, and deposition testimony.

As to the duty to pay defense costs, the District Court concluded that the Texas Endorsement in the 1981 Transit and 1982/83 ACIC policies eliminates the duty to reimburse NGC’s defense costs for claims arising in Texas. NGC argues that the Texas Endorsement, while negating the duty to assume defense of the Texas claims, does not in any way limit the Insurers’ obligation to reimburse defense costs incurred. We agree. In arriving at its narrow interpretation of the policy language, the District Court appeared to have simply assumed that the Texas Endorsement was intended to exclude the duty to pay defense costs as well as the duty to defend with respect to Texas claims, and failed to distinguish between the two kinds of duties. Moreover, the Insurers did not argue in the District Court, and do not argue on appeal, that the Texas Endorsement limits the obligation to pay defense costs for claims in Texas. The policies are at least ambiguous on this point, and should be construed in favor of NGC and coverage. Therefore, we reverse this aspect of the District Court’s ruling.

However, we uphold the District Court’s conclusion that several insurers have no duty to pay defense costs under the AMI-CO umbrella policies in effect from 1973 through 1979, and the excess policies that follow the form of the AMICO umbrella policies (the “AMICO policies”).²⁵ The original insuring agreement required the insurers to indemnify NGC’s “ultimate net loss,” including damages and expenses that NGC became obligated to pay. The “New York Amendatory Endorsement” amended the definition of “ultimate net loss” in the insuring agreement by deleting the reference to “expenses.” Notwithstanding NGC’s efforts in the District Court and on appeal to rely on the legislative history of the New York Amendatory Endorsement, the District Court properly found that the term “ultimate net loss,” as amended, unambiguously includes only damages and not defense costs. See Home Insurance Co. v. American Home Products Corp., 902 F.2d 1111, 1113-14 (2d Cir.1990).

The District Court also found no obligation to indemnify NGC’s defense costs under the 1974-76 Affiliated FM Insurance Company, 1980 and 1981 United States Fire Insurance Company, and 1985 International Insurance Company policies. NGC concedes that these policies impose no duty to defend, but argues that they nonetheless require the insurers to indemnify NGC for defense costs. The policies state that the insurer will pay a proportion of “expense[s] and/or costs in connection with any claim or suit” that are “incurred jointly by mutual consent.” NGC argues that this provision does not require NGC to obtain the excess carriers’ approval before NGC incurs legal fees in defending itself, but instead only permits the insurers to deny unreasonable or excessive defense costs incurred.

However, we agree with the District Court that, pursuant to this provision, the insurer has no duty to defend or pay costs, but only has the right to do so at its own election. The insurer can permit the insured’s defense to proceed and take the chance that the insured might exceed the limits of other liability coverage, or participate in the defense and agree to pay all or part of the defense costs incurred. The consent provision does not require the insurer to indemnify NGC for defense costs unless the parties mutually agree beforehand to this arrangement.

Finally, the District Court ruled that, under policies containing a duty to reimburse defense costs but not a duty to defend, the Insurers have a duty to reimburse defense costs for claims that are established to be covered through judgment and settlement, and not for claims only potentially falling within the policy’s coverage. This conclusion is supported by both the case law and the policy language, which states that NGC will be indemnified for damages and defense costs “paid as a consequence of any occurrence covered hereunder.” NGC argues that the Insurers must indemnify defense costs whenever a claim is filed and allegations in the complaint raise a possibility of coverage, see Okada v. MGIC Indemnity Corp., 823 F.2d 276, 280 (9th Cir.1986), but even Okada indicates that the insured must repay any defense costs advanced if the facts ultimately show that the claim was not covered, regardless of the allegations contained in the underlying complaint. See id. at 282. The policies do not contemplate unconditional payment of defense costs for potentially covered claims, but only payment of costs if indemnification is required.

D. Insurers’ Peremptory Jury Challenges

The District Court allowed the Insurers five peremptory challenges jointly, the same number allowed to NGC. CU contends that because some of the Insurers had adverse interests and took different positions on the issue of trigger of coverage, it was error to treat them as one defendant, particularly because CU was the only insurer involved in the jury trial. We disagree.

The statute governing allocation of peremptory challenges in civil cases, 28 U.S.C. § 1870, “specifically leaves to the trial court’s discretion the number and manner of exercise of peremptories in cases where there are ‘[s]everal defendants or several plaintiffs.’ ” Carr v. Watts, 597 F.2d 830, 832 (2d Cir.1979) (quoting section 1870). The District Court properly exercised its discretion by slightly increasing the number of statutorily-granted challenges (from three to five) and then treating all of the Insurers as one party for the purpose of exercising those challenges. CU did not object to the decision to group the Insurers, nor did it seek additional challenges.

Conclusion

The judgments of the District Court are affirmed except as to the rulings: (1) including ACIC in the group of excess insurers to whom NGC made representations of an exposure-only limit, (2) allocating liability to NGC for uninsured periods after 1985, (3) finding that only a single occurrence was involved in asbestos-in-building claims asserted against NGC, and (4) holding that the Texas Endorsement eliminated obligations to pay defense costs on claims arising in Texas; these four rulings are reversed. The bench trial finding as to the triggering of coverage for cancer claims — that asbestos-related bodily injury was not proven to be continuously occurring — is remanded for further consideration. The cases are remanded for further proceedings consistent with this opinion.

1 . Although ACMC has been substituted for NGC in this action and on appeal, for convenience we shall refer to NGC.

2 . The "Jury Trial BI Judgment” sets forth rulings relating to coverage for asbestos bodily injury claims under policies issued to NGC by Commercial Union Insurance Company ("CU”). The “Bench Trial BI Judgment” sets forth rulings relating to the same set of issues but with respect to policies issued by the other insurers. Finally, the “PD Judgment" memorializes the rulings concerning coverage for asbestos property damage claims against NGC. NGC appeals from the Jury Trial BI Judgment, the Bench Trial BI Judgment, and the PD Judgment. CU appeals from the Jury Trial BI Judgment, while the other Insurers appeal from the Bench Trial BI Judgment. All Insurers appeal from the PD Judgment.

3 .Virtually the same policy language was used even until the 1980s, when insurers were well aware of the fact that asbestos-related claims were being brought against insureds, and that the nature of the injuries and damage involved in these claims was quite different from the types of harms envisioned in the 1950s.

4 . For a discussion of the process by which the insurance industry drafted these occurrence-based forms, see American Home Products Corp. v. Liberty Mutual Insurance Co., 565 F.Supp. 1485, 1500-03 (S.D.N.Y.1983), modified, 748 F.2d 760 (2d Cir.1984).

5 . Although most of the policies indicate on their face that it is the injury, and not the "occurrence” or cause of that injury, that must occur while the policy is in effect, certain policies issued by American Motorists Insurance Company ("AMICO") could be construed to require that the occurrence take place during the policy period as well. Those policies generally state that AMICO will indemnify for "personal injury ... caused by or arising out of an occurrence which takes place during the policy period anywhere in the world." NGC trial exhibit 9 (emphasis added). By contrast, CU's policies, for example, provide for indemnification for personal injury “caused by or arising out of an occurrence happening anywhere in the world." However, AMI-CO has interpreted its own policy language to require that only injury need occur during the policy period. AMICO reads its language to mean that the phrase "which takes place during the policy period anywhere in the world” modifies "personal injury," rather than the immediately preceding word “occurrence.” Moreover, AMICO's policies use the same definition of “occurrence” as do the other policies, and there is no evidence that the parties intended narrower coverage under AMICO’s policies. Therefore, we rule that those policies, like the other policies, require that only injury needs to result during the policy period. See Eagle-Picher Industries, Inc. v. Liberty Mutual Insurance Co., 682 F.2d 12, 23-24 (1st Cir.1982).

6 .See, e.g., Note, Adjudicating Asbestos Insurance Liability: Alternatives to Contract Analysis, 97 Harv.L.Rev. 739, 745-46 (1985); Special Project — An Analysis of the Legal, Social and Political Issues Raised by Asbestos Litigation, 36 Vand. L.Rev. 573, 709-30 (1983).

7 . Those insurers joining AMICO are The Constitution State Insurance Company ("CSIC") and Republic Insurance Company ("Republic").

8 . The policy in AHP used substantially the same wording of the key terms found in the policies in the pending case. The one significant variation was the addition of a specific exclusion for "injury, death or destruction caused by such continuous or repeated exposure any part of which occurs after the termination of the policy.” See AHP, 748 F.2d at 762.

9 . California appears to have left this precise issue for future resolution. Montrose Chemical, 10 Cal.4th at 676 n. 16, 42 Cal.Rptr.2d at 341 n. 16, 897 P.2d at 18 n. 16.

10 . It is arguable that the damages for which an insurer in the final year of a progressive disease process is liable should be less than those for which an insurer in an early year is liable, on the theory that the liability arising from the injury in an early year is for all subsequent consequences, whereas the liability arising from the injury in the final year is only the final year of suffering plus the death of a substantially diseased individual. However, none of the Insurers has contended on this appeal that its liability should vary according to the particular interval between exposure and death during which each was on the risk. Their only allocation contention argued on appeal is to support the District Court’s pro-ration-to-the-insured approach. See Part 11(B), infra.

11 . CU’s further contention — that even if triggering may be proven to have occurred in successive policy periods after exposure, such triggering must end at the point of manifestation — will be considered in Part IV(A), infra, in our discussion of the "known loss” defense.

12 . CU also contends that the trial court erred in failing to instruct the jury that a “fully developed” disease could constitute injury-in-fact. However, we find no error in Judge Martin's instruction, which merely echoed our holding in AHP, 748 F.2d at 766, in stating that "a real but undiscovered injury which is proven in retrospect to have existed at the relevant time will trigger the policy in effect at that time.” The jury was not precluded from finding that only a fully developed disease could trigger coverage; rather, the jury was informed that "[i]t is for you to decide when bodily injury, sickness or disease occurs.”

13 . The preceding two paragraphs have been circulated to all of the active judges of the Court.

14 . These insurers were CSIC, AMICO, Republic, and ACIC.

15 . Although ACIC is no longer a parly to this appeal (its coverage disputes with NGC have been settled), NGC argues that this issue is not moot because the way in which the ACIC policy responds to claims may affect the obligations of policies excess thereto.

16 . We therefore have no occasion to consider the correctness of the formula. See Owens-Illinois, Inc. v. United Insurance Co., 138 N.J. 437, 474-81, 650 A.2d 974, 993-96 (1994).

17 . The District Court relied on two decisions of the New York Appellate Division, National Casualty Insurance Co. v. City of Mount Vernon, 128 A.D.2d 332, 515 N.Y.S.2d 267 (2d Dep’t 1987), and Levine v. Lumbermen's Mutual Casualty Co., 147 A.D.2d 423, 149 A.D.2d 372, 538 N.Y.S.2d 263 (1st Dep't 1989). These decisions, however, do not resolve the issue presented in this case. In Mount Vernon, the claimant sued for wrongful incarceration; the decision sensibly held that each day of incarceration was compensable and that the insurer was liable only for the days falling within its policy period. Mount Vernon did not impose on the insured a portion of the loss for which the insurer was liable; it simply obliged the insurer to pay only the damages resulting from the discrete injury that occurred during the interval that the insurance was in effect. Levine is arguably somewhat closer to our case since it involves what the Appellate Division regarded as a continuous injury (legal malpractice in failing to act properly to have a dismissed action reinstated). But Levine holds only that the several insurers on the risk throughout the period of injury each bear a pro rata portion of the loss. Since there was no interval in which insurance was lacking or exhausted, the Court had no occasion to consider the proration-to-the-insured approach.

The District Court also relied on Judge Wein-stein’s opinion in Uniroyal, Inc. v. Home Insurance Co., 707 F.Supp. 1368, 1383 (E.D.N.Y.1988), but that decision is distinguishable. In Uniroyal, the parties had stipulated that any injury from Agent Orange occurred at or shortly after a serviceman's exposure to spraying. Uniroyal, 707 F.Supp. at 1389. That stipulation permitted “a clean differentiation” between different policy periods for the injuries of the various claimants. Id. at 1392.

18 . Forty-Bight Insulations had no occasion to consider the proration issue in periods when insurance was no longer available, since the case was decided in 1980, prior to the arrival of the asbestos exclusion clauses in 1985.

19 . This decision, of course, does not alter NGC's obligation, without insurance indemnification, to pay for claims based on exposures occurring after 1985. As to such claims, none of the Insurers is “on the risk,” and no issue of proration arises.

20 . They are CU, Stonewall, Republic, Houston General Insurance Company, Old Republic Insurance Company (“OR”), and AMICO.

21 . CU's policies also incorporate a form of the AMICO exclusion.

22 . Even if the relevant event were each delivery of NGC's asbestos-containing products to intermediaries or directly to consumers, the number of such deliveries over the years — and thus the number of deductibles under each policy — would likely be substantial and result in significantly diminished coverage for NGC.

23 . CSIC contends that the premium it received was so low compared to the amount of excess insurance purchased as to demonstrate that it could not have been insuring the losses that had already occurred, even though not specifically known to NGC. However, the relatively low amount of the premium reflected NGC's assurance that CSIC’s policies would be triggered only by exposure occurring after the inception of CSIC's policies.

24 . All claims involving Home were settled prior to this appeal. However, we must interpret the Home policy because CU’s policy is “subject to all the terms and conditions of policies ... issued by [Home].”

25 . The District Court's holding applies to policies issued by CU, Republic, and Stonewall that were in effect during this period.