This case is a companion to our decision today in Commercial Union Insurance Co. v. Swiss Reinsurance America Corp., 413 F.3d 121, No. 04-1709, 2005 WL 1503121 (1st Cir. 2005) Both involve reinsurance and the same reinsurer—Swiss Reinsurance America Corporation ("Swiss Re") and related cedents (the "cedent" being the insurance company that is reinsured). The cedent in No. 04-1709 is Commercial Union; in this case it is American Employers' Insurance Company ("American"), which is now affiliated with Commercial Union. One of the two questions presented in this case overlaps with that in Commercial Union; the other differs.

We begin with the facts and litigation history. American provided excess insurance coverage to Pennsalt Chemical Company pursuant to three multiple-year umbrella insurance policies—the "A-15 policies"—together covering the period January 1, 1964, through January 1, 1971.1 The policies covered liability for bodily injury and property damage that might be incurred by Pennsalt, over and above coverage provided by a primary insurer (who plays no role in the present litigation).

American's own liability to Pennsalt was subject to limits set forth in the American policies: $2 million for "each occurrence" under the first two American policies and $5 million per occurrence under the third. All three policies stated that the per-occurrence limit "is the total limit of the company's liability under this policy for ultimate net loss as a result of any one occurrence." All three policies, with minor variations, defined "occurrence" to mean:

(a) an accident, or (b) an event, or continuous or repeated exposure to conditions, which unexpectedly results in personal injury, property damage, or advertising liability ... during the policy period.... [A]ll personal injury and property damage ... arising out of one event or continuous or repeated exposure to substantially the same general conditions existing at or emanating from one premises location shall be deemed to be one occurrence.

For the period 1962 to 1969, a second insurance company provided additional excess coverage to Pennsalt, under the so-called "S-16 policies," triggered only when the policy limits of the A-15 policies were exceeded. The coverage was less generous to the insured because it paid only a percentage of the insured's excess liability. By the time the present controversy arose, Pennsalt's rights as the insured belonged to a successor company—Elf Atochem North America ("Elf")—and American was responsible for payments under the S-16 policies.

When it issued the A-15 policies, American had reinsured with Swiss Re by securing three multi-year reinsurance certificates corresponding to each of American's A-15 umbrella policies covering Pennsalt. Each certificate was skeletal in form, see Ostrager & Vyskocil, Modern Reinsurance Law & Practice § 5.03 (1996), comprising one two-sided page and a few attached endorsements. The front page described the contract generally: "[Swiss Re] [d]oes hereby reinsure [American] ... with respect to [American's] policy hereinafter described, in consideration of the payment of the premium, and subject to the terms, conditions and amount of liability set forth herein."

The certificate then identified the relevant American policy being reinsured, stating its limits (e.g., "$2,000,000 each occurrence and in the aggregate, in excess of underlying limits"), and then stating the "reinsurance accepted," i.e., the portion of American's payout that Swiss Re would indemnify. In the first two certificates this field read "50%"; for the third certificate, it read: "$2,000,000 each occurrence which is 40% quota share part of $5,000,000 which in turn is excess of underlying limits."

Each certificate contained on the back side several standard conditions. These included a follow-the-form clause: The liability of [Swiss Re] specified in Item 4 of this Certificate shall follow that of [American], and except as otherwise specifically provided herein, shall be subject in all respects to the terms and conditions of [American's] policy.

Such clauses are designed to make the terms of coverage under an excess policy or reinsurance policy (here Swiss Re's certificate) conform—subject to certain qualifications—to the coverage provided by an underlying policy (here, American's excess policies). Aetna Cas. & Sur. Co. v. Home Ins. Co., 882 F. Supp. 1328, 1337 (S.D.N.Y.1995) ("Where a following form clause is found ... a policy of reinsurance will be construed as offering the same terms, conditions and scope of coverage as exist in the reinsured policy...."); Ostrager & Vyskocil, supra, § 2.03[a], at 2-9; see also Associated Int'l Ins. Co. v. Blythe, 286 F.3d 780, 782 (5th Cir.2002).

The certificates also contained "follow-the-fortunes" provisions, often described as "follow-the-settlements" provisions. They preclude the reinsurer from challenging a cedent's decision to settle so long as the settlement is "reasonably within the terms of the [cedent's] policy, even if not technically covered by it" and so long as the cedent has acted in good faith. See Ostrager & Vyskocil, supra, §§ 9.01-.02; Travelers Cas. & Sur. Co. v. Certain Underwriters at Lloyd's of London, 96 N.Y.2d 583, 734 N.Y.S.2d 531, 760 N.E.2d 319, 328 (2001). The clauses here read:

All claims involving this reinsurance, when settled by [American], shall be binding on [Swiss Re], which shall be bound to pay its proportion of such settlements promptly following receipt of proof of loss.

In the early 1990s, Elf began to notify American of potential hazardous waste losses involving property damage and bodily injury caused by pollution at various sites. On August 29, 1994, American filed a declaratory judgment action in New Jersey Superior Court against Elf in response to Elf's demand for indemnification of property and injury claims arising out of pollution at a Bryan, Texas, site. The case soon expanded to involve additional insurers, parties, issues, and property and injury claims at 82 sites.

Discovery over the next four years focused on the Bryan, Texas, site. In April 1998, Elf presented American with a proposal—which American did not accept—to settle many of its claims against American at 37 sites for $95 million. An underlying premise of Elf's calculations (not explicitly stated but clear from the figures in the document) was that the per-occurrence limit in the American policies applied once per policy period to continuous or repeated exposures like the pollution damage at issue here.

Most courts read such per-occurrence language in such a once-per-policy-period manner, see Commercial Union, slip op. at 10, 413 F.3d at ___. New Jersey law arguably has adopted an "annualization" approach that for these kinds of continuing or recurring losses applies per-occurrence limits to each year of a multi-year policy. See, e.g., Benjamin Moore & Co. v. Aetna Cas. & Sur. Co., 179 N.J. 87, 843 A.2d 1094, 1103 (2004). Such an approach would ordinarily enlarge the insured's coverage and increase the insurer's liability. Id. at 9-11, 413 F.3d at ___-___.

In June 1998, American's coverage counsel—an outside law firm—provided a status report to American's reinsurers. It said inter alia that American would argue against annualization of the per-occurrence limits but noted "certain factors"2 that might support an annualization reading of the policies—a reading which Elf had not urged in settlement but which Elf was free to argue in the event the matter had to be litigated. In substance, coverage counsel assumed annualization as a possibility because, as an American representative said later, the aim was to tell the reinsurers "what the exposure could potentially be."

One of the wild-card issues throughout the litigation and settlement negotiations was which jurisdiction's law would be taken to apply to the construction of the policies as to individual waste sites and to "allocation" issues where more than one policy would cover the losses at a site (for example, when several successive policies of one or more insurers covered various years). In a case-management conference in March 1999, the judge presiding in the New Jersey case indicated that the law of the situs state would "govern the case" as to that site, although he later obliquely suggested that allocation would occur "pursuant to Owen," a reference to a leading New Jersey allocation case. See Owens-Illinois, Inc. v. United Ins. Co., 138 N.J. 437, 650 A.2d 974 (1994).

For most of 1999, mediation progressed under the auspices of the New Jersey court. Efforts initially focused on what the parties deemed to be the 10 most important of the 37 relevant sites. Elf furnished information as to remediation costs at these sites, and American hired an expert environmental consulting firm to make its own assessment of the total damages. The environmental consultant's work ultimately produced much lower estimates of remediation costs than Elf did.

In June 1999, Elf presented another settlement demand seeking $120 million from American for the 37 sites;3 its spreadsheet calculations assumed high remediation costs but no annualization of the per-occurrence limits. On this premise, some of the A-15 policy limits were exceeded, but this fact did not terminate American's liability there; instead, some of the alleged liability flowed over to the S-16 policies.

In October 1999, American's coverage counsel offered American its own analysis supporting an estimated exposure of $47.6 million. A detailed analysis of the top 10 sites employed the much lower remediation-cost estimates derived from the environmental consultant's work, assumed annualization of the per-occurrence limits (based on the same factors described in the June 1998 status report), and applied an additional blanket discount of 30 percent to allow for the possible success of American's policy defenses (e.g., the insured's failure to provide prompt notice and its awareness of the harmful nature of the chemicals present at the sites). The total estimated exposure for the top 10 sites was $44.8 million.

The settlement analysis included a brief description of the remaining 27 sites ("secondary sites") identified by Elf (to which Elf had ascribed liability for American of only $14 million). The analysis said: "As we have no information on these sites and are thus unable to assess their alleged damages, as well as the possible impact on [American's] coverage, we have calculated an eighty-percent reduction of the liability [from Elf's figures]." Accordingly, the analysis assigned an expected additional liability of $2.8 million to the secondary sites.

At a mediation session in November 1999, Elf again demanded $120 million. Elf this time explicitly stated that its approach assumed one per-occurrence limit for each multi-year policy. American countered with an offer of $18 million. During the session the parties' offers converged somewhat to $78 million and $21 million (but the parties differed on the scope of the release). During a break, the American team met privately; it was noted that given American's liability to Elf under the S-16 policies for excess coverage where the A-15 policies left off, there was little difference between Elf's non-annualized approach and American's annualized approach "as far as the final liability number was concerned."

In another session in December 1999, the parties' demands and offers further converged (to $45 million and $40 million, respectively), and in January 2000, the parties agreed in principle on a settlement of $44 million to be paid in two installments. A final agreement was signed in April 2000. The agreement, which contained no mention of annualization, stated that the agreement was not "an admission with respect to insurance policy interpretation or application or an admission by any Party regarding the duties, rights or obligations arising out of the Policies."

Thereafter, American billed Swiss Re $20.5 million for what American claimed was Swiss Re's share of the settlement under the three reinsurance certificates. American's allocation of the settlement among American's policies—which in turn implicated reinsurer reimbursement—is somewhat complicated, but it basically followed the methodology laid out in the October 1999 settlement analysis. In any event, both sides agree that American's allocation of liability to Swiss Re was premised on the use of annualized per-occurrence limits and that Swiss Re's liability would be less if the limits were applied, site by site, on a once-per-policy basis.

Swiss Re resisted American's billing on the ground that a non-annualized approach should be used and also on the ground that American allocated part of the settlement payment to the 27 secondary sites.4 Swiss Re's position on the latter issue was that American had no information as to the secondary sites and therefore assigning any value to them was unreasonable or not made in good faith, thus, Swiss Re said it did not have to indemnify American for any share of the claimed liability as to these sites.

Eventually, American brought suit against Swiss Re in the federal district court in Massachusetts to recover the full amount it had allocated to Swiss Re. Swiss Re then made a partial payment but insisted that its liability was limited for the two reasons set forth above. On its calculation, its liability was reduced by $3.5 million attributable to annualization and $1.2 million attributable to the secondary sites.

Discovery was followed by cross-motions for summary judgment. On August 5, 2003, the district court entered a decision holding for Swiss Re on both disputed issues. See American Employers' Ins. Co. v. Swiss Reinsurance Am. Corp., 275 F. Supp. 2d 29 (D.Mass.2003). Following the earlier district court decision in Commercial Union (by a different district judge), the district court in the present case said that the certificates' per-occurrence language precluded annualization. As for Swiss Re's liability for the secondary sites, the court found that American's settlement made without investigation of those sites was not in good faith. American has now appealed to this court.

Review of a grant of summary judgment is de novo. Commercial Union Ins. Co. v. Walbrook Ins. Co., 7 F.3d 1047, 1050 (1st Cir.1993). The district court found (and the parties do not dispute) that the reinsurance certificates are governed by the substantive law of either New York or Massachusetts;5 in both states, contract interpretation—at least absent extrinsic evidence—is a question of law for the court.6 What law governs the underlying American policies held by Pennsalt is debatable—perhaps different states' laws apply for losses at different sites.

We start, as in the Commercial Union case, with the cedent's liability to its own insured under the cedent's policies, for the reinsurer is only liable for losses legitimately incurred by the insured. Taken alone, the term "occurrence" is arguably ambiguous as to whether per-occurrence limits should be annualized; but in the A-15 policies quoted above, further language—favoring Swiss Re—states that "repeated exposures" at a site constitute one occurrence and that exposures comprising "the same general conditions" (conceivably, even different pollutants from the same manufacturing operation) are also "one occurrence."

To thus define "occurrence" does not automatically defeat an annualization approach; that approach might still be defended as reflecting actual intention or extrinsic policy. But the broader the definition of occurrence, the easier it is to think of disparate leaks or spills over the policy period as comprising a single event; and in fact the policy refers to injuries occurring "during the policy period." Anyway, American does not point to evidence of intent or develop policy arguments to offset language. Instead it relies upon the settlement with Elf.

American says that it settled Elf's claims against it by a settlement based on annualization; that this was a good faith, reasonable settlement; and that Swiss Re is bound by its own follow-the-settlements clauses to accept the settlement as the basis for American's liability to Elf—and therefore Swiss Re's to American, subject only to possible defenses contained in the Swiss Re certificates.

Swiss Re's first response is to deny the premise that the American settlement with Elf is based on annualization. Unlike Commercial Union, where the insured's initial demand was based on annualization, slip op. at 13, here Elf did not rely on annualization when it proffered its demands. Nevertheless, Elf would have been free to assert annualization if the case had been litigated and American, in framing its own counteroffer, could take this into consideration.

More important, the settlement American actually paid was based on a calculation of its own liability that did assume annualization. American's coverage counsel's analysis, positing an expected liability figure of $47.6 million, reduced other figures (e.g., Elf's claimed remediation costs) but assumed-for stated reasons-that Elf would prevail as to annualization. American's ultimate payment of $44 million was very close to this expected figure, and American approved this level of payment partly because of the earlier projection.7

If the case had been litigated, Elf would not have had the benefit of the primary policy language available to the insured in Commercial Union but American did face a risk—not easy to assess on the present record—that in litigation the New Jersey courts might choose to apply New Jersey law, which is arguably pro-annualization. Whether the risk was sufficiently small to make the settlement unreasonable is a different issue which Swiss Re is free to litigate on remand.

It is perhaps of more concern that American's own liability may not have been much affected by whether annualization were employed or not because of its own spill-over liability based on the S-16 policies. But there is no law that says that insurer and reinsurer interests have to be perfectly aligned to trigger a follow-the-settlement clause, and it remains open to Swiss Re to challenge American's good faith on remand if it has evidence to support such a claim.

Although the issue may be close, American's characterization of the settlement is supportable. There is no case law directly in point and arguments both ways can be made, but American did calculate its ultimate obligation using annualization and the settlement roughly matched this figure—this is not a post-hoc characterization or a unilateral post-settlement allocation without grounding in the settlement process itself.8 There is considerable advantage in taking the insurer's own contemporaneous calculus as a starting point and then letting the objections be tested primarily under the rubric of reasonableness and good faith—which brings us to Swiss Re's second objection.

It is well settled that to trigger the deference due under a follow-the-fortunes clause the cedent's settlement must be made in good faith, Travelers, 734 N.Y.S.2d 531, 760 N.E.2d at 328; see also Christiania Gen. Ins. Corp. of N.Y. v. Great Am. Ins. Co., 979 F.2d 268, 280 (2d Cir.1992), and this extends to the adoption of the settlement premise that American now seeks to impose on Swiss Re, see Seven Provinces, 9 F.Supp.2d at 66 ("The reinsurer cannot dispute ... good faith interpretations of policy terms."). Despite American's objection, we think that reasonableness is also a requirement under governing law,9 recognizing that a reasonable settlement may not precisely conform to every technicality in the policy.

Swiss Re is free on remand to challenge reasonableness and good faith. It may argue, for example, that the chances were very small in an Elf-American lawsuit that American would be held liable based on annualized limitations—a consideration that clearly bears on the reasonableness of paying out funds based on that premise. So, too, Swiss Re may try to show that American had little interest in contesting annualization in light of the S-16 policies and reinsurance coverage and argue that this undermines a claim of good faith. The initial assessment is for the district court.

As a final objection, Swiss Re suggests that whatever American's liability to Elf, its own liability to American remains limited by its certificates. This is so in principle: although follow-the-settlements clauses make a cedent's liability determinations "binding" on the reinsurer and follow-the-form clauses presume that the reinsurance dovetails with the underlying liability insurance, this presumption can be overridden by clear language in the certificate that cuts off liability under the reinsurance policy even where the cedent is liable to the insured. See Travelers, 734 N.Y.S.2d 531, 760 N.E.2d at 328 n. 9 ("[A] `follow-the-fortunes' clause does not supersede specific language in a reinsurance contract...."); Bellefonte Reinsurance Co. v. Aetna Cas. & Sur. Co., 903 F.2d 910, 913 (2d Cir.1990) (reimbursement of defense costs paid under cedent's settlement "would be contrary to the parties' express [reinsurance] agreement").

But, as in Commercial Union, there is no clear-cut anti-annualization language in the Swiss Re certificates. They use only the bare-bones term "per occurrence" which, standing alone, neither adopts nor refutes annualization. See Commercial Union, slip op. at 15-17. And absent a clear limitation in the certificate, the principle of congruent liability between cedent and reinsurer—adopted by Swiss Re and American in the certificates' follow-the-form and follow-the-settlements clauses, see Aetna, 882 F.Supp. at 1337—suggests that Swiss Re's liability should follow the gloss (assuming it is reasonable and made in good faith) given to the underlying policies by the settlement. See Commercial Union, slip op. at 15-16.

The second part of this appeal concerns Swiss Re's refusal to pay a portion of the $2.6 million of the settlement attributed by American to the sites other than the top 10 sites at which costs and ultimate liability were thoroughly investigated. Swiss Re claims, and the district court agreed, that American violated its duty of good faith by allocating any of the settlement payment to these sites without collecting, from Elf or otherwise, any site-specific information as to damages or American's own coverage obligations as to the sites under the circumstances.

We cannot agree with the district court that American's settlement of the claims based on the 27 secondary sites was a "bad faith" settlement simply because American used a mechanical 80 percent discount in the October 1999 settlement analysis—from which the challenged allocation derives—instead of requiring remediation cost data from Elf and then making its own site-by-site adjustments. Yes, a mechanical discount approach requires explanation, but a colorable explanation has been provided, as we now explain.

The top 10 sites, which were the main subject of the settlement, represented by far the largest amount of damages claimed by Elf against American. American did acquire specific information from Elf about the remediation costs at those 10 sites, including repricing by its own expert environmental consultant; and it made its own judgments about what its liability might be for those damage claims under its Pennsalt policies. American's successive offers to Elf, and the final amount paid, represented American's discounted versions of Elf's claims.

The discounts from Elf's initial per site claims varied substantially among the top 10 sites. They ranged, so far as we can compute them, from a discount of about 86 percent in the case of one site down to about 29 percent at the least discounted site. The overall discount reflected in the October 1999 settlement analysis appears to be in the 60- to 70-percent range, if one compares American's calculated liability (and later payout) with the total liability that Elf attributed to American in its June 1999 settlement demand for losses at the same 10 sites.10

Elf's claims against American for the 27 secondary sites ($14 million) were much less than its claims for the top 10 sites ($128 million). In preparing its analysis, American did not go through the same process of getting remediation figures from Elf for the 27 sites and adjusting them downward individually. Instead, American simply discounted the amounts claimed by a flat 80 percent and built this amount into the analysis that guided American's settlement conduct. The discount was intended to reflect both reduced remediation costs and the previously used 30-percent discount for possible coverage defenses (which were thought to apply equally to the secondary sites). Two American deponents so testified.

The main attack by Swiss Re, and the only ground adopted by the district court, was that this mechanical discount approach inherently amounted to bad faith because there was no investigation of the claims as to the individual 27 sites. But terms like "good faith" and "reasonableness" do not prescribe one approach for all settlements; how much individualized study is needed depends on the situation. If a house burns down and the insured offers to settle for $500, an insurer could reason that the sum is a fraction of litigation costs and arson investigation.

Here, American treated the top 10 sites as a very crude proxy for the rest. The Elf claims for all of the sites were a common baseline that might be assumed to be biased in Elf's favor to somewhat the same degree. American worked out a range of discounts for the top 10 sites and then appears to have adopted a larger but somewhat similar discount for the remaining sites. The discount was very large and the sites represented small potatoes compared to the claims for the top 10 sites. And Elf settled for a sum close to American's version of its expected liability.

Seemingly the choice of 80 percent was somewhat arbitrary, just like the initial 30-percent coverage-defense discount; at least American has never pointed to any detailed derivation for the 80-percent figure. But there is some flavor of "extrapolation" here (evidenced by application of the same coverage defense reduction), and it was a large discount and one even more favorable to American than most of the more calibrated discounts developed by American for the top ten sites. Perhaps Swiss Re can offer some basis for arguing the 80-percent discount was unreasonably small but the fact that it is a round number based on judgment rather than a mathematical equation is not enough.

On the surface, this looks like a good deal for American and so for Swiss Re. Of course, if individualized remediation-cost assessments were free, there would be no excuse for not making them. But they do impose costs—here for the expert consultants, the time of executives and lawyers to evaluate results and bargain further, and the delay. Certainly American could have settled only as to the 10 sites, but this would simply have left the risks of greater liability and litigation costs as to the 27 sites for the future.

Settling numerous claims based on detailed information about only a subset of those claims is consistent with modern practice in similar cases. See, e.g., E.I. Du Pont De Nemours & Co. v. United States, 297 F. Supp. 2d 740, 743 (D.N.J.2003) (court limiting discovery to one of fifteen waste sites, to permit parties to develop issues to foster adjudication of claims and settlement); Aluminum Co. of Am. v. Aetna Cas. & Sur. Co., 140 Wash.2d 517, 998 P.2d 856, 862 (2000) (court designating three of thirty-five sites as "test sites" for similar reasons). There is no rule of law, although Swiss Re strives to create one, that there is some fixed quantum of investigation or individualized computation required for a reasonable, good faith settlement. The commentators and cases on which Swiss Re relies support nothing other than the common-sense assumption that one will look into a claim before paying it.11 How far one looks, with what tools, and with what costs in money and delay, obviously depend on circumstances.

In this case, the district court's decision rests solely on the lack of an investigation directed separately at the 27 sites, and that ground will not sustain the judgment for Swiss Re. To the extent Swiss Re has alternative grounds for objecting to the good faith or reasonableness of the settlement as to the 27 sites, it is free to raise them on remand. Comparing the modest stakes with the cost of modern litigation, the parties would be well advised to settle.

The judgment of the district court is vacated and the matter is remanded for further proceedings consistent with this opinion.

It is so ordered.