O'SCANNLAIN, Circuit Judge:

We consider unresolved issues of California insurance law to decide what constitutes an "occurrence" under a third-party insurance contract and when such coverage is triggered among successive policy years.

I

Chemstar, Inc. ("Chemstar") is a supplier of lime plaster. Its predecessor, Genstar Lime Company ("GLC"), manufactured Type S Hydrate lime, which GLC marketed for use as an interior plaster finish in homes. From 1984 to 1986, GLC sold lime containing high amounts of magnesium oxide pellets, which are called "periclase." During the lime manufacturing process, periclase reacts with water through a process called "hydration" to become magnesium hydroxide, a stable compound. Some periclase may not hydrate during manufacturing but may later react with moisture in the air until it eventually changes into magnesium hydroxide.

Lime that contains unhydrated periclase is unfit for home interior use. Where such lime is encased in a plastered surface, hydrating periclase may expand and pop through the smooth plaster, causing unsightly pits. Pitting is not inevitable, however, because periclase may never expand. Further, because pitting does not weaken plaster structurally, periclase-laden lime is suitable for exterior use.

When GLC sold high-periclase lime from 1984 to 1986, it wrote on bills of lading that the lime was "for exterior use only." It did not inscribe this warning on bags containing the lime. Consequently, while GLC's direct purchaser was on notice that high-periclase lime was unsuitable for interior use, subsequent recipients of the lime, who would not have received a GLC bill of lading, did not necessarily receive notice. Given these circumstances, the district court concluded that "it was almost inevitable that some lime with high periclase concentrations would be used indoors." Chemstar, Inc. v. Liberty Mut. Ins. Co., 797 F. Supp. 1541, 1545 (C.D.Cal. 1992).

Contractors did, in fact, use high-periclase lime in the interiors of homes in southern California, causing plaster pitting in at least 28 of these homes. The 28 homeowners brought claims against GLC and Chemstar between 1985 and 1988.

Chemstar turned to its insurers for indemnity and defense of the homeowners' claims. The policies allegedly covering these losses were all based on a standard comprehensive general liability policy and contain essentially the same terms. The 1985-86 policy of defendant Liberty Mutual Insurance Co. ("Liberty") is typical. It states that Liberty:

[will] pay on behalf of the Insured all sums which the insured shall become obligated to pay by reason of the liability-imposed on the Insured . . . [f]or damages because of . . . destruction of tangible property during the policy period. . ., caused by an occurrence.

Chemstar filed a complaint asserting a claim against Liberty and other insurers seeking, among other things, a declaration of its rights to coverage. The insurers, in turn, filed various counter-and cross-claims. The district court bifurcated this action into two proceedings, Phases I and II. In Phase II, the subject of this appeal, the district court allocated among Chemstar's insurers Chemstar's property damage and defense costs.

On July 23, 1992, the district court issued a partial summary judgment order, ruling that all 28 plaster pitting claims arose from one occurrence and had occurred when plaster pitting manifested in the first home. After a later two-day bench trial, the district court found that plaster pitting first manifested during the 1985-86 policy year. On March 10, 1993, the district court entered its Phase II Declaratory Judgment Order, incorporating its previous dispositive orders and factual findings.

Chemstar, Liberty, United Insurance Company ("United"), American Home Assurance Company ("American") and Employers Insurance of Wausau ("Wausau") appeal various portions of the district court's July 23, 1992 summary judgment order.

II

Except as noted below, California law governs this diversity action. Intel Corp. v. Hartford Acc. Indem. Co., 952 F.2d 1551, 1555 (9th Cir. 1991). The panel applies California law as it believes the California Supreme Court would apply it. Id. Although California appellate court decisions are persuasive precedent, the panel is not bound by them if it believes that the California Supreme Court would decide otherwise. Hancock Labs., Inc. v. Admiral Ins. Co., 777 F.2d 520, 525 n. 10 (9th Cir. 1985).

The district court allocated pitting damages at all 28 homes to the March 15, 1985 — March 15, 1986 policy period. It based this decision on three determinations: (1) plaster pitting at all 28 homes arose from one occurrence; (2) plaster pitting occurred at the time that it first manifested, which was during the March 15, 1985 — March 15, 1986 coverage year; and (3) plaster pitting at all 28 homes constituted one, progressive loss. We examine each of these three determinations.

A

Chemstar's insurance policies cover property damage arising from an "occurrence." See Whittaker Corp. v. Allianz Underwriters, Inc., 11 Cal.App.4th 1236, 14 Cal.Rptr.2d 659, 663 (1992). The district court held that the underlying cause of the plaster pitting was GLC's failure to warn end-users that high-periclase lime was unsuitable for indoor use. Chemstar, 797 F. Supp. at 1547. According to the district court, this failure to warn stemmed from an inadequate quality control system at GLC. Id. at 1545, 1548 n. 12.

Wausau argues that pitting at the 28 homes did not result from a single cause but, instead, arose from a combination of distinct causes. Because plaster pitting was not inevitable in every home in which high-periclase lime was installed, Wausau asserts that pitting resulted from circumstances unique to each home rather than an underlying failure to warn.

The district court rejected this argument. Declining to individuate causes of the pitting at each home, it stated that GLC's:

failure to warn was a continuing cause that was not interrupted by any other cause. As noted before, if GLC had employed an adequate quality control system and warned of the proper uses of its product, the other causes offered by American, Wausau, and United would not have resulted in property damage.

Id.

We agree with the district court that there was no intervening, proximate cause after GLC's failure to warn. In Mead Reinsurance v. Granite State Insurance Co., 873 F.2d 1185, 1187 (9th Cir. 1989), a municipality (the "City") had been sued in twelve separate section 1983 lawsuits, and the City looked to its primary and excess insurers for indemnity. The City and primary insurer argued that eleven of these twelve suits stemmed from only one occurrence because, in these cases, the City's liability was premised upon a single police policy condoning excessive police force. Id. at 1187. The excess insurer argued that the eleven actions involved different occurrences because the allegations of the complaints involved different police actions, different injuries, and different classes of plaintiffs. Id.

The district court entered summary judgment against the excess insurer. It found that the eleven section 1983 complaints alleged the same excessive force by its policy were premised upon the City's deliberate indifference to excessive force by its police department. Finding the differences in the allegations of the complaints to be irrelevant, the district court concluded that, under California law, the eleven claims arose from the same occurrence. This court affirmed. Id. at 1187-88.

While we are concerned that Mead failed to cite any California authority for its decision, we acknowledge that it supports the district court's conclusion as to the proximate cause of the plaster pitting. As in Mead, the fact that the 28 incidents of pitting involved different homes, claimants, sources of lime, and times does not preclude a finding that the incidents arose from the same underlying cause. Although Wausau argues that it was not inevitable that pitting would occur despite Chemstar's failure to warn, the same principle is true for Mead: the use of excessive force was not inevitable, despite a policy condoning it, in every instance of police conduct; instead, the use of force presumably depended upon the actions of individual police officers and the circumstances in which force was used.

A case cited by Wausau Eureka Fed.Sav. Loan Ass'n v. American Cas. Co. of Reading, Pennsylvania, 873 F.2d 229 (9th Cir. 1989), is distinguishable from this action. In Eureka, Eureka Savings Loan Association ("Eureka") and two of its former officers sought a declaratory judgment that an insurer's directors and officers liability policy covered losses that Eureka incurred in over 200 loan transactions. The insurer argued that losses from the 200 transactions constituted a single loss because they resulted from an "aggressive lending strategy." Id. at 234. The district court rejected this contention, treating each transaction as a separate loss. We affirmed, stating:

[T]he fact that all loan losses arguably originated from one loan policy does not require finding only one loss. In this case there were numerous intervening business decisions that took place only after the loan policy was initiated that required the exercise of independent judgment.

Id. at 234-35 (underscore added). The court added that it "d[id] not foreclose the possibility . . . that . . . separate borrowers may be aggregated as a single loss in an appropriate fact situation." Id. at 235.

Unlike the loan transactions in Eureka, no intervening business decisions occurred between Chemstar's failure to warn and the pitting. Rather, it qualifies as "an appropriate fact situation" in which distinct losses arise from one occurrence. We thus conclude that Mead, rather than Eureka, governs this case.

B

Having determined that the plaster pitting stemmed from a common "occurrence," we consider when the plaster pitting occurred. Chemstar's insurance policies only cover property damage that results from an occurrence within the respective policy periods.

1

Adopting a "manifestation" trigger of coverage, the district court held that plaster pitting occurred at the first date that it became reasonably apparent or appreciable to homeowners. Chemstar, 797 F. Supp. at 1551-52. Chemstar, United, and Wausau assert that the district court should have applied a "continuous trigger" of coverage to determine which insurers were on the risk at the time of the plaster pitting. Under a continuous trigger, an insurer must cover damage that either begins or progresses during its policy period. Antognini, "A Prophecy from the Wilderness: The California Supreme Court and the Trigger of Coverage," 5 Cal.Ins. Law Reg.Rptr. 196, 204 (1993) ("Prophecy"). Had the district court adopted a continuous trigger, property damage would not have been limited to the 1985-86 policy year but, instead, likely would have been spread out over several years, triggering coverage under several policies.

In Hancock Laboratories, Inc. v. Admiral Insurance Co., 777 F.2d 520 (9th Cir. 1985), a case decided under California law, we explicitly rejected both the manifestation and continuous triggers in favor of an "exposure" trigger, which none of the parties in this appeal advocate. Since Hancock, however, several California courts have selected a manifestation trigger. In Prudential-LMI Commercial Insurance v. Superior Court of San Diego County, 51 Cal.3d 674, 274 Cal.Rptr. 387, 798 P.2d 1230 (1990), the California Supreme Court adopted a manifestation trigger for first party insurance. In that case, a homeowner looked to its insurers to cover an extensive crack in an apartment unit's foundation. The crack had been caused by gradual soil expansion which could have occurred over several policy periods. A district court of appeal held that the property damage should have been apportioned to all insurers whose policies covered the period in which the crack had developed.

The supreme court rejected the intermediate appellate court's approach and adopted a manifestation trigger for two reasons. First, the supreme court believed that "the underwriting practices of the insurer can be made predictable because the insurer is not liable for a loss once its contract with the insured ends unless the manifestation occurred during its contract term." Id. 274 Cal.Rptr. at 404, 798 P.2d at 1247. This increased predictability leads to reduced insurance premiums "because insurers will be able to set aside proper reserves for well-defined coverage and avoid increasing such reserves to cover potential financial losses caused by uncertainty in the definition of coverage." Id. at 403, 798 P.2d at 1246. Second, the supreme court believed that the manifestation trigger met the reasonable expectations of insured, who "look to their present carrier for coverage" of losses. Id. at 404, 798 P.2d at 1247.

The court expressly limited its decision to first party insurance. Concerned "that there are substantial analytical differences between first party property policies and third party liability policies," it declined to identify the appropriate trigger for third party liability insurance, leaving that question open for subsequent resolution. Id. at 389, 403, 798 P.2d at 1232.

Subsequent decisions by California appellate courts have applied a manifestation trigger to third party insurance. For example, in Pines of La Jolla Homeowners Association v. Industrial Indemnity, 5 Cal.App.4th 714, 7 Cal.Rptr.2d 53 (1992), the insureds were responsible for construction of a condominium complex. After a multitude of construction defects appeared, the condominium homeowners association sued the insured, who then sought indemnity from their insurers. Another district court of appeal recently held that a manifestation trigger determined in which policy periods the defects occurred. Id. 7 Cal.Rptr.2d at 57. Tomaselli v. Transamerica Ins. Co., 25 Cal.App.4th 1269, 31 Cal.Rptr.2d 433 (1994).

Given the changed state of California law, it appears most likely that the California Supreme Court would adopt a manifestation trigger for the third party insurance claims involved in this case. The first, and most obvious, reason for this conclusion is that post- Hancock California precedent has chosen a manifestation trigger. Although not determinative, this fact weighs strongly in favor of abandoning Hancock and adopting a manifestation trigger. See Houston Gen. Ins. Co. v. AG Production Co. Chemurgic Agric. Chems., Inc., 840 F. Supp. 738, 742 n. 1 (E.D.Cal. 1993).

Second, the supreme court's reasons for adopting a manifestation trigger for first party insurance in Prudential equally favor a manifestation trigger in this case. As in first party insurance, a manifestation trigger presumably will promote certainty in the third party insurance industry because an insurer can be confident that it will not be liable for a loss once its policy period has ended and no loss has manifested. See Antognini, "Prophecy," 5 Cal.Ins. Law Reg.Rptr. at 209. This increased certainty should redound to the benefit of insureds, who will pay lower premiums.

Also, a manifestation trigger does not pose a risk of compromising an insured's reasonable expectations of coverage under the facts of this case. Here, the insured already has received coverage for the plaster pitting claims. The court's selection of a trigger of coverage thus will affect only the allocation of coverage among insurers; it will not impact on Chemstar. A manifestation trigger can no more deprive Chemstar of reasonably anticipated coverage than it can upset reasonable expectations in first party insurance. See Fireman's Fund Ins. Co. v. Aetna Cas. Sur. Co., 223 Cal.App.3d 1621, 273 Cal. Rptr. 431, 434, 435 (1990).

Chemstar and United argue that a manifestation trigger is inconsistent with Chemstar's insurance policies because the policies nowhere limit coverage to property damage that "manifests," becomes discoverable, or starts during a policy period. See TBG, Inc. v. Commercial Union Ins. Co., 806 F. Supp. 1444, 1453 (N.D.Cal. 1990). We disagree. At best, the argument raises the possibility that the insurance contracts are ambiguous. California courts appear to have rejected that position, however. Further, underlying the parties' criticism is the view that the manifestation trigger upsets the reasonable expectations of insured. Yet, in Prudential, the California Supreme Court observed that an insured typically looks to its present insurer to cover losses that occur. 274 Cal.Rptr. at 404. The manifestation trigger is consistent with this expectation. Further, even if an insured tends to interpret a third party liability policy in the manner suggested by Chemstar and United, there is no need to defer to that understanding in this case. The "general principle that ambiguities in insurance contracts must be interpreted in favor of coverage is inapplicable where . . . the case concerns only the respective liabilities of two insurers." Fireman's Fund, 273 Cal.Rptr. at 435. As noted above, the trigger of coverage will affect only the respective liabilities of Chemstar's insurers because Chemstar already has received full indemnity.

Chemstar and United present no convincing reason to disregard the clear trend of California authority. The district court correctly selected a manifestation trigger to determine when the plaster pitting damage occurred.

2

Applying the manifestation trigger, the district court found that plaster pitting first manifested between March 15, 1985 and March 15, 1986. Liberty and Wausau challenge this factual finding, arguing that plaster pitting first manifested during the 1984-85 policy period. We review for clear error, giving due regard to the district court's superior opportunity to judge the credibility of witnesses. Fed.R.Civ.P. 52(a); United States v. Helmandollar, 852 F.2d 498, 501 (9th Cir. 1988); Wardley Int'l Bank, Inc. v. Nasipit Bay Vessel, 841 F.2d 259, 261 n. 1 (9th Cir. 1988).

The district court's finding was not clearly erroneous. The parties stipulated that pitting had become appreciable before March 15, 1986. Further, there was no dispute that pitting had occurred during the March 15, 1985 — March 15, 1986 policy period. The only disputed fact was whether pitting had occurred before this period. Addressing this issue, Liberty presented the testimony of homeowners and a general contractor to prove that pitting had occurred before March 15, 1985. Chemstar introduced the testimony of plaster installers and inspectors to refute this testimony. Finding the testimony of Liberty's witnesses to be unreliable and inconsistent with other evidence, the court concluded that the preponderance of the evidence showed that pitting had first manifested after, rather than before, March 15, 1985. We see no reason to overturn this credibility assessment.

Liberty and Wausau also complain that the district court wrongly gave them the burden of proving that plaster pitting occurred prior to March 15, 1985. California law places "the burden of proof . . . on . . . the insured . . . to prove that an event is within the scope of coverage." Whittaker, 14 Cal.Rptr.2d at 664.

The district court applied the correct burden of proof. As noted above, the parties already had stipulated that pitting had manifested during the March 15, 1985 — March 15, 1986 policy period. Chemstar thus had met its burden of producing evidence that damage had occurred during the policy period. Although Liberty asserts that it was forced to prove a negative — i.e., that pitting did not first manifested before March 15, 1985. In not true. The district court simply required Liberty to show (affirmatively) that pitting first manifested before March 15, 1985. In fact, it is Liberty that seeks to impose upon Chemstar the burden of proving a negative: under Liberty's approach, Chemstar would have had to demonstrate, not only that pitting manifested during the 1985-86 policy period, but also that no pitting had manifested before March 15, 1985.

We thus conclude that the district court correctly selected and applied the manifestation trigger.

C

We last consider whether pitting at all homes should be telescoped into the March 15, 1985 — March 15, 1986 period of coverage, even though pitting at some homes first became appreciable after that time. The district court held that, as a matter of law, "all of the homeowners' property damage occurred when the plaster-pitting manifested in the first home, thus triggering coverage only under those policies effective at the time of first manifestation." Chemstar, 797 F. Supp. at 1553.

The district court reached its decision by viewing all plaster pitting as one, progressive loss. California provides that, once property damage manifests within a policy period, the insurer is solely liable for that damage, even if it progresses after the policy lapses. See Pines of La Jolla v. Industrial Indem., 7 Cal.Rptr.2d at 56-57. Progressive loss typically occurs where damage to a single building or property becomes worse over a period of time. See, e.g., Fireman's Fund, 273 Cal.Rptr. at 431.

Analogizing to this situation, the district court concluded that there was:

no basis . . . for distinguishing between cases in which one occurrence gives rise to property damage in one or multiple buildings. On the contrary, it finds good reason to treat both types of cases the same. . . . [F]inding a single trigger in both cases avoids anomalous outcomes. Without the single trigger, courts may, at their whim, allocate losses among insurers by drawing arbitrary lines between, say, two condominiums in the same building, two halves of one duplex, a house and a guest house on one property, or two homes on the same street. Such anomalous outcomes . . . expose insurers to greater uncertainty: Although an insurer may escape liability in one court, it has no guarantee that the next court will not hold it liable on similar facts.

Chemstar, 797 F. Supp. at 1552.

The district court added:

[F]inding a single trigger in both cases protects the insured's access to insurance. In this case, once plaster-pitting manifested in the first home, a potential insurer who is aware of this risk would refuse to insure Chemstar if some courts treated plaster-pitting in a different home arising from the same occurrence as a separate trigger. By contrast, finding a single trigger protects the insurer from such potential liability and makes it willing to insure against other property damage arising from future occurrences.

Id. (emphasis added).

California courts have never directly addressed this question. The closest case is Whittaker Corp. v. Allianz Underwriters, Inc., [ 11 Cal.App.4th 1236,] 14 Cal.Rptr.2d 659 (1992). In Whittaker, the insured ("Whittaker") developed and marketed a sealing compound that beverage can manufacturers used to seal can ends onto can bodies. This sealing compound originally contained an anti-oxidant called No-Nox. In 1983, Whittaker switched to another anti-oxidant, ACD-50. Whittaker's three customers — American Can Co., Ball Corp. and Reynolds Metal Co. — later experienced problems with Whittaker's sealing compound, which had permitted leakage from the customers' beverage cans. Whittaker sought indemnity from two insurers under separate 1983 and 1984 policies for claims brought by these customers.

The parties agreed that American Can's claim was "a fiscal year 1983 occurrence" because Whittaker had sold the compound to American Can in 1983 and the damage had occurred in 1983. There thus was no dispute that American Can's claim was covered by the 1983 insurance policy. The Court found, however, that neither Ball not Reynolds had suffered damage from the sealing compound until 1984. It concluded that these claims were not covered by the 1983 policy, even though the American Can loss had manifested during the 1983 coverage period.

Whittaker can be read to support the view that property damage which arises from the same occurrence but which occurs at different locations does not constitute one, progressive loss. However, the Whittaker court never directly addressed this point; instead, it framed the issue before it as whether a claim arises at the time of the act that causes property damage or the time of the manifestation of the damage. Id. 14 Cal.Rptr.2d at 662. We thus are reluctant to interpret Whittaker as a definitive statement of California law on the precise question before this court.

Given this, we find the district court's reasoning to be persuasive. In addition to avoiding arbitrary distinctions, applying the progressive loss rule in a multi-location context avoids creating a gap in insurance coverage. Once property damage at one building manifests, wary insurers may become unwilling to insurance for such damage in future policy periods out of fear that the damage will manifest at other buildings. There thus is a risk that insured will become unable to obtain insurance coverage for these anticipated losses. See K. Abraham, Distributing Risk 159 (1986).

Further, applying the progressive loss rule in these circumstances will not subvert the purpose of the manifestation trigger, which is to enable insurers to ascertain their liabilities at the end of a coverage period. See Prudential-LMI, 274 Cal.Rptr. at 403-04. Just as other insurers can foresee the possibility of future losses arising from a single occurrence which already has manifested one injury, so too can one's present insurer. Consequently, once damage manifests at one location during a policy period, an insurer is on notice of the risk of damage occurring at other locations. It thus has ample opportunity to investigate the extent of its possible exposure and to adjust its insurance reserves accordingly.

III

We conclude that the district court did not err in entering its July 23, 1992 summary judgment. It properly found that there was one occurrence and that the insurer on the risk at the time of the first manifestation of pitting damage was liable for pitting losses at all 28 homes. Accordingly, the March 10, 1993 final order incorporating that judgment is

AFFIRMED.