Full opinion text
OPINION OF THE COURT AMBRO, Circuit Judge. This is a dispute over reinsurance coverage. In 1998, Travelers Casualty and Surety Co. (“Travelers”) reached a $137 million settlement with its insured, Acme Corporation (“Acme”). Travelers then proceeded to allocate those $137 million dollars among three tiers of insurance coverage, only the highest of which — the so-called “excess” layer — included policies reinsured by Ace America Reinsurance Company and Insurance Company of North America (collectively, “INA”). When Travelers billed INA $13,762,395 based on its allocation, INA refused to pay, and Travelers sued to recover in the Eastern District of Pennsylvania. At issue before the District Court was whether Travelers manipulated its post-settlement allocation so as to maximize the amount allocated to policies reinsured by INA, thus excusing INA from its normal duty as a reinsurer to “follow” all coverage decisions made by its reinsured. The District Court held two bench trials, each addressing a different aspect of Travelers’ allocation, and ultimately reached what was, in effect, a split decision. The Court ruled, following the first bench trial, that Travelers had not manipulated its allocation of the settlement dollars so as to allow it to reach the excess layer of coverage (and thus tap into its reinsurance). But the Court also ruled after the second bench trial that, once Travelers reached the highest tier of coverage, it allocated more to certain policies reinsured by INA than was reasonably allowed by their policy limits. The result of those two verdicts was to leave INA responsible for only $8,226,817 of the loss initially allocated to it. The Court then issued two consequential post-trial rulings. In the first, it held that prejudgment interest on Travelers’ award should be calculated according to the Pennsylvania rate, even though the reinsurance contracts under which Travelers sued were governed by New York law. In the second, it held that post-judgment interest on the prejudgment interest did not begin to accrue until the District Court issued its order quantifying the amount of prejudgment interest due. Both parties appealed. We affirm both trial verdicts as well as the ruling concerning when post-judgment interest on the prejudgment interest began to accrue. However, because we believe that Travelers’ award of prejudgment interest should be calculated according to the higher New York rate, we remand on that issue only so that the prejudgment interest can be recalculated. I. BACKGROUND A. The Follow-the-Fortunes Doctrine and the Reinsurance Relationship Because the events that gave rise to this dispute occurred in the context of a relationship between an insurer (Travelers) and its reinsurer (INA), we begin with some background into the reinsurance relationship. Reinsurance is a mechanism “ ‘by which one insurer insures the risk of another insurer.’” N. River Ins. Co. v. Ace Am. Reins. Co., 361 F.3d 134, 137 (2d Cir.2004) (quoting People ex rel. Cont’l Ins. Co. v. Miller, 177 N.Y. 515, 70 N.E. 10, 12 (1904)). The insurer pays the rein-surer a premium in exchange for which the reinsurer assumes “a portion of the [insurer’s] potential financial exposure under certain direct insurance policies it has issued to its insured.” Id. Obtaining reinsurance allows an insurer to diversify its risk exposure, thus increasing its “capacity to insure other customers and decreasing] the likelihood that ... insolvency will result from any large claim.” N. River Ins. Co. v. CIGNA Reins. Co., 52 F.3d 1194, 1199 (3d Cir.1995). A crucial feature of the reinsurance relationship is that “[r]einsurance involves contracts of indemnity, not liability.” Unigard Sec. Ins. Co. v. N. River Ins. Co., 4 F.3d 1049, 1054 (2d Cir.1993). That is, in providing reinsurance, the rein-surer . acquires no direct liability to the original policyholder; rather, the reinsurer assumes an obligation to indemnify the insurer for payments it makes under the reinsured policies. Id. Indeed, a reinsurance agreement typically contains two specific provisions designed to prevent the reinsurance relationship from encroaching on coverage disputes between the insurer and its insured: a “follow-the-form” provision, in which the reinsurer agrees to rein-sure the policies as written, and a “follow-the-fortunes” provision, in which the rein-surer agrees to “follow” the coverage provided by the insurer. See CIGNA, 52 F.3d at 1199-1200. Of these two provisions, the most crucial is the follow-the-fortunes provision. See Barry R. Ostrager & Mary Kay Vyskocil, Modem Reinsurance Law & Practice § 2.03[d] (2d ed.2000), at 2-17 (noting that the “follow-the-fortunes” provision lies “at the heart of the reinsurance agreement”). The follow-the-fortunes doctrine significantly restricts a reinsurer’s ability to challenge the coverage decisions that led to its liability to the insurer. This is so for a basic reason — “[i]f the [insurer] knew that its settlement decisions could be challenged by every reinsurer, there would be little incentive to settle with the insured. The costs and risks of litigation avoided by settling with the insured would only be revived at the reinsurance stage.” Commercial Union Ins. Co. v. Seven Provinces Ins. Co., 9 F.Supp.2d 49, 66 (D.Mass.1998); see also CIGNA, 52 F.3d at 1206 (“To permit the reinsurer to revisit coverage issues resolved between the insurer and its insured would place insurers in the untenable position of advancing defenses in coverage contests that would be used against them by reinsurers seeking to deny coverage.”). Accordingly, the follow-the-fortunes doctrine “insulates a reinsured’s liability determinations from challenge by a reinsurer unless they are ... in bad faith, or the payments are clearly beyond the scope of the original policy.” Ace, 361 F.3d at 140 (internal quotation marks and citation omitted). In other words, a rein-surer seeking to avoid payment must show either that the coverage decisions that led to the reinsurer’s liability to the insurer were made in bad faith, or that the coverage provided clearly fell outside the scope of the policies the reinsurer agreed to rein-sure. See Mentor Ins. Co. (U.K.) Ltd. v. Brannkasse, 996 F.2d 506, 517 (2d Cir. 1993). Otherwise, the reinsurer must simply cover the losses allocated to it. B. Acme v. Travelers and Travelers v. INA In April 1996, Travelers acquired Aetna Casualty and Surety Company (“Aetna CS”). . At the time, Acme was seeking coverage under insurance policies issued by Aetna CS in the 1970s and 1980s. Acme sought coverage primarily for two sets of claims being brought against it: (1) breast implant claims, relating to safety testing of silicone breast implants that Acme had performed for its parent, Acme Parent Corporation (“Acme Parent”); and (2) chemical products claims, relating to chemical products manufactured by Acme, including the pesticide commonly known as “DBCP.” The Aetna CS policies potentially implicated by the breast implant claims and the chemical products claims made up three distinct layers of coverage — primary policies (bearing the designator “AL”), buffer policies (bearing the designator “XS”), and excess policies (bearing the designator “XN”). Because the distinct features of each layer’s policies became central to the dispute between Travelers and INA that followed, it is worth describing those policies in some detail. 1. The Insurance Policies The AL policies were issued between April 1976 and April 1987 and provided coverage for all non-products claims brought against Acme, as well as products claims brought against it outside the United States. Each of the AL policies had a per-occurrenee coverage limit, but only the policies issued between April 1985 and April 1987 had aggregate coverage limits. In addition, the AL policies had three features that became particularly significant to the reinsurance dispute that followed. First, the policies were subject to retrospective premiums from Acme. For any payment Travelers made on an AL policy, it was entitled to reimbursement from Acme up to that particular policy’s “loss limit.” The AL policies covering the period from April 1976 through April 1982 included a limit on the amount in retrospective premiums that could be collected, while the post-April 1982 AL policies included no such limit. Second, the AL policies were subject to captive reinsurance, that is, reinsurance provided by an Acme subsidiary. Each AL policy was reinsured for 95% of all losses above the loss limit, except the last policy, which was reinsured at 95.5% above the loss limit. Finally, the AL policies included an obligation to cover defense expenses in addition to an obligation to indemnify Acme for liability it incurred. For the April 1976 through April 1982 AL policies only, defense costs did not count toward the policy limits, while, for all the AL policies, captive reinsurance could not be sought for defense costs unless Travelers also made indemnity payments to Acme. The XS policies covered the period between April 1976 and April 1982. They provided United States products liability coverage and were in excess of Acme’s primary layer of products liability coverage, meaning that they were only available to Acme once its primary layer of products liability coverage (which was supplied by a Acme subsidiary) was exhausted. The XS policies were also subject to captive reinsurance. The policies spanning from April 1976 to April 1978 were 100% reinsured for bodily injury claims, while the remaining XS policies were 95% reinsured for all claims. The XN policies provided the final layer of coverage. These policies covered both products and non-products claims, and were in excess of all Acme’s insurance coverage (including its coverage under the AL and the XS policies). While none of the XN policies was subject to either retrespective premiums or captive reinsurance, the XN layer was the one layer that did possess non-captive reinsurance, including reinsurance from INA. INA assumed, through facultative reinsurance certificates, a portion of nine of the XN policies. Each certificate issued by INA contained both a follow-the-form provision and a follow-the-fortunes provision. 2. The Settlement Negotiations Acme initially sought coverage for the tens of thousands of breast implant claims brought against it under the AL policies, and sought coverage for the chemical products claims under both the XS policies and the XN policies. At first, Aetna CS declined coverage, and then, starting in February 1996, made a series of settlement offers to Acme, each of which was rejected. Upon its acquisition of Aetna CS, Travelers restarted settlement negotiations from scratch. Negotiations were handled, from Travelers’ end, by Timothy Yessman, who was Senior Vice President of Travelers’ Special Liability Group, and Susan Stonehill-Clafin, who was General Counsel for Travelers’ Environmental Litigation Group. Yessman acted as lead negotiator, while Stonehill-Clafin provided legal advice. Up to that point, Acme had not yet settled, or received an adverse judgment on, any of the claims against it, but had incurred substantial defense costs. Accordingly, the parties focused initially on reaching a “coverage-in-place” deal. Under such an arrangement, Travelers would agree to pay a fixed sum to cover Acme’s past losses, and, for Acme’s future losses, the parties would work out a formula for matching the specific claims against Acme to the specific insurance policies. Travelers would then make payments pursuant to that formula — subject to a finite cap — -as Acme’s ultimate liability developed. In its negotiations with Acme about how to characterize the claims for which coverage was being sought, Travelers was adamant about two points — that the breast implant claims were products claims, and thus were not covered by the AL policies, and that they arose out of a single occurrence (namely, a single act of negligent testing on Acme’s part). According to testimony Yessman later provided, it was the number of occurrences issue that was viewed as the most critical. Because the AL policies possessed per-occurrence limits, but were not (for the most part) subject to aggregate limits, Travelers’ greatest concern was that the breast implant claims would be characterized as non-products claims arising out of multiple occurrences. Under such a scenario, it was possible that Travelers’ exposure under the AL policies would be exponentially greater, at least if the liability for each occurrence was below the per-occurrence limit. While the negotiations with Acme were ongoing, Yessman had Mark Wigmore, a Vice-President and Associate General Counsel in Travelers’ reinsurance department, produce a memo (the “Wigmore Memo”) that became central to the litigation that followed. The Wigmore Memo explored the reinsurance implications of different coverage scenarios for the breast implant claims. The Memo noted a number of issues of potential concern, only two of which are particularly germane to this appeal. First, the Memo suggested that, because Travelers could not collect captive reinsurance on payments made to cover defense costs unless it also paid out in indemnity, it was possible that “Acme [would] litigate each and every case to the fullest extent, without making any settlements, in order to avoid its ... reinsurance obligations.” (J.A. at 615.) Second, the Memo mentioned that, if the breast implant claims were determined to be non-products claims arising out of multiple occurrences, Acme might never get out of the AL layer of coverage and into the reinsured XN layer, since, if the liability for each occurrence was low enough, it was possible that it would never exhaust the per-oceurrence limits of the AL policies not subject to aggregate limits. The Memo also speculated that, if Travelers were to bill its reinsurers based on a single occurrence characterization, “[cjollection [from them would] likely ... be more difficult,” but that, if Acme agreed to that characterization in any settlement, Travelers “would have a strong position” in “litigation or arbitration” with its reinsurers. (J.A. at 616.) The final settlement meeting between Travelers and Acme took place on July 7, 1998, at which point both parties changed their approach. Acme proposed moving to an all-cash net settlement. That meant that, rather than (as with a coverage-in-place deal) coming up with a formula for how to treat Acme’s future losses, Travelers would simply pay Acme a lump sum— forgoing both retrospective premiums and captive reinsurance — in exchange for Acme releasing all of its future claims under the policies. Travelers accepted the proposal and, with the new framework in place, the parties quickly agreed on a figure of $137 million. They then decided that, of that $137 million, $80 million would be dedicated to the breast implant claims, $20 million would be dedicated to the chemical products claims, and the remaining $37 million would go to claims that are not at issue in this case. In addition, the parties agreed that the breast implant claims would be treated as non-products, single occurrence claims, while the chemical products claims would be treated as products claims. Beyond that, they did not come to any agreement about how to allocate the settlement to the specific policies potentially implicated. 3. The Final Settlement Agreement Although Travelers and Acme reached an agreement in principle during the July 7, 1998 meeting, the settlement was not finalized until September of that year. The primary issue in dispute was Travelers’ proposed allocation of the agreed-upon sum among the different policies. According to Robert Miley, who, along with William Kingston, was primarily responsible for drafting the settlement agreement on behalf of Travelers, it was not that Acme objected to Travelers’ proposed allocation so much as that it “wondered whether ... it [ie., the allocation language] needed to be ... in the settlement agreement.” (Trial Tr. vol. 1, 216, Jan. 11, 2005.) In a draft dated July 27, Travelers included language indicating that, of the $20 million dedicated to the chemical products claims, $5 million would be allocated to the TIC policies, while the remaining $15 million would be allocated to the XN policies. That draft also included language providing that no amount could be allocated to the post-April 1982 AL policies. In a draft returned to Travelers on September 1, Acme crossed out most of the allocation language, put a question mark next to the line indicating that the post-April 1982 AL policies were not to be used, and added language stating that, with the exception of any allocations specifically set forth in the agreement, each party reserved the right to allocate the settlement as it pleased. In a draft returned on September 4, Acme continued to designate the allocation language “IN DISPUTE,” and specifically crossed out the section regarding allocation to the post-April 1982 AL policies. The settlement agreement became final in mid-September. It provided that Travelers’ payments to Acme were “net of any reinsurance obligations the Acme Insurance Subsidiaries have or may have to Travelers and net of any retrospective premium or other obligations Acme has or may have to Travelers.” It also included the language indicating that the $20 million for the chemical products claims would be divided between the TIC policies ($5 million) and the XN policies ($15 million). In addition, the final version included a paragraph providing that “[n]o payments ... shall be allocated to any [AL] Primary Policies with a policy period commencing on or after April 1, 1982, or to any ... XS Policies because the payments of the Settlement Amount are net payments and such Policies have been exhausted by virtue of the settlement.” Lastly, the final version provided that “[w]ith the exception of the agreements explicitly set forth in ... this Agreement, Acme and Travelers each reserve to themselves the right to allocate any or all of the Settlement Amount to any Policy; Acme will not be deemed to concur in any such allocation by Travelers, and Travelers will not be deemed to concur in any such allocation by Acme.” In the subsequent litigation with INA, Travelers conceded that the allocation language was included in the settlement agreement at its behest. 4. The Post-Settlement Allocation and Reinsurance Billing Once the settlement agreement was finalized, Miley and Kingston proceeded to allocate the settlement among the different policies. As agreed, of the $20 million dedicated to the chemical products claims, $5 million was allocated to the TIC policies and $15 million to the XN policies. Travelers characterized the $80 million dedicated to the breast implant claims entirely as indemnity, not defense coverage. In allocating that $80 million, Travelers began with the AL layer of policies, but, in accord with the agreement, confined itself to the pre-April 1982 AL policies. In allocating within that layer, it employed the so-called “fill the bathtub” method. Starting with the earliest of the policies, Travelers allocated to each eligible AL policy up to its single-occurrence limit (minus the amount owed in retrospective premiums) before moving on to the next policy. This resulted in a total of $24 million being allocated to the AL policies. The remaining $56 million of the $80 million dedicated to the breast implant claims was then allocated to the XN policies in accordance with the fill-the-bathtub method (starting with the XN policies with the lowest attachment points). Two of the XN policies implicated by Travelers’ allocation of the breast implant claims settlement had three-year policy periods. In exhausting those three-year policies (each of which was reinsured by INA), Travelers treated their per-occurrence limits as applying separately to each policy year, a decision that tripled the amount that could be allocated to those policies. Ultimately, six of the nine XN policies reinsured by INA had settlement dollars allocated to them. Pursuant to its allocation, Travelers billed INA $11,604,328 for the breast implant claims and $2,158,067 for the chemical products claims. Travelers agrees that the following decisions likely increased the amount of its coverage it was able to allocate to INA — (1) treating the breast implant claims as arising out of a single occurrence; (2) bypassing the post-April 1982 AL policies in allocating the $80 million dedicated to the breast implant claims; (3) bypassing the XS policies in allocating the $20 million dedicated to the chemical products claims; (4) not allocating the $80 million for the breast implant claims exclusively to defense costs (even though, at the time of the settlement, Acme had not yet incurred any liability on those claims); and (5) allocating to the three-year XN policies on the assumption that their per-occurrence limits applied separately to each policy year. The decision to annualize the per-occurrence limits alone resulted in an increase of $5,535,578 to the amount of loss assigned to INA. At any rate, INA refused to pay any of the amount Travelers allocated to it. C. The District Court Proceedings In January 2001, Travelers brought its breach of contract action against INA, contending that INA was barred, under the follow-the-fortunes doctrine, from challenging Traveler’s resolution of its coverage dispute with Acme. In insisting that it was not obligated to pay, INA did not question the propriety of the underlying $137 million settlement. Rather, it challenged only Travelers’ post-settlement allocation of that settlement. At the close of discovery, both parties moved for summary judgment. In August 2004, the District Court denied both parties’ motions for summary judgment and the case proceeded to two separate bench trials. The first trial (“Phase I”) addressed whether Travelers had engineered its post-settlement allocation to maximize the amount of the settlement that ended up in the reinsured XN layer of coverage. The second trial (“Phase II”) addressed the propriety of Travelers’ decision, once it reached the XN layer, to treat the three-year policies as subject to three separate per-occurrence limits. Prior to the Phase I bench trial, INA made a motion in limine seeking to preclude testimony relating to discussions with, or analyses prepared by, Travelers’ in-house or outside counsel. In its motion, INA asserted that, under the so-called “sword/shield” doctrine, Travelers could not both invoke a privilege to shield its communications with its attorneys (as it had throughout discovery) and defend its conduct with reference to advice received by counsel. The District Court partially denied and partially granted the motion, limiting Travelers’ testimony on its advice from counsel to the topic, rather than the content, of those communications. Ultimately, the Court ruled that Travelers could refer generally to its use of counsel in making certain decisions to show that it proceeded in a “businesslike” manner, but could not attribute any particular decision to the advice of counsel. (J.A. at 30.) The Phase I trial was held in January and February of 2005. The District Court heard testimony from both parties’ experts, as well as the Travelers employees who worked on the settlement with Acme (including the post-settlement allocation and the reinsurance billing) all of whom denied that reinsurance recovery considerations motivated any of their decisions. In December 2005, the District Court ruled in favor of Travelers on the issues addressed in the Phase I trial. The Court summed up its findings as follows: Although there is certainly enough evidence in the record to raise the suspicions of [INA], I generally find Travelers’ witnesses to be credible. I further find that Travelers did not allocate the sum plaintiff owed under the settlement agreement to maximize its potential reinsurance recovery from [INA], that Travelers did not act in bad faith, and that its various actions were reasonable, businesslike decisions made in good faith. (J.A. at 57.) The Phase II trial was held in February 2006. After once again hearing testimony from both parties’ experts, as well as from Travelers’ employees, the District Court ruled in INA’s favor. The Court concluded that, under Michigan law (which governed the insurance policies Aetna CS had issued to Acme), “the three-year XN policies clearly and unambiguously have a single per-oecurrence limit for the entire policy period.” (J.A. at 96.) It therefore held that Travelers’ interpretation of the those policies’ per-occurrence limits was not binding on INA as its reinsurer. The Court then (in an order summing up both its Phase I and its Phase II rulings) entered judgment in favor of Travelers in the amount of $8,226,817, a figure derived by subtracting from Travelers’ initial billing of $13,762,395 the $5,535,578 attributable to its decision to annualize the per-occurrence limits of the three-year XN policies. Next, the District Court issued two consequential post-trial orders. First, it held that, though the reinsurance contracts between Travelers and INA were governed by New York law, it was required, as a federal court sitting in diversity in Pennsylvania, to calculate the prejudgment interest owed on Travelers’ award under the (lower) Pennsylvania rate, as (according to the District Court) such a calculation falls under Pennsylvania’s procedural, rather than its substantive, law. Second, the Court held that post-judgment interest on Travelers’ award of prejudgment interest did not begin to accrue until the Court issued its order quantifying the amount of prejudgment interest owed to Travelers, rather than (as Travelers contended) beginning to run when the Court first awarded Travelers a sum subject to prejudgment interest. Both parties timely appealed. II. JURISDICTION AND STANDARD OF REVIEW The District Court had jurisdiction under 28 U.S.C. § 1332(a)(1). We have jurisdiction under 28 U.S.C. § 1291. In considering the appeals from the District Court’s Phase I and Phase 2 bench trials, “we review [the] [District [C]ourt’s findings of fact for clear error and its conclusions of law de novo.” McCutcheon v. Am.’s Servicing Co., 560 F.3d 143, 147 (3d Cir.2009). When “confronted with mixed questions of fact and law, we apply the clearly erroneous standard except that the District Court’s choice and interpretation of legal precepts remain subject to plenary review.” Gordon v. Lewistown Hosp., 423 F.3d 184, 201 (3d Cir.2005). To the extent that the District Court’s conclusions rested on credibility determinations, our review is particularly deferential. See Anderson v. Bessemer City, 470 U.S. 564, 575, 105 S.Ct. 1504, 84 L.Ed.2d 518 (1985) (“When findings are based on determinations regarding the credibility of witnesses, ... even greater deference to the trial court’s findings [is required]; for only the trial judge can be aware of the variations in demeanor and tone of voice that bear so heavily on the listener’s understanding of and belief in what is said.”). Nonetheless, a “trial judge may [not] insulate his [or her] findings from review by denominating them credibility determinations, for factors other than demeanor and inflection go into the decision whether or not to believe a witness.” Id. Finally, insofar as what is being challenged is an evidentiary ruling, not a specific finding of fact or conclusion of law, we review “for abuse of discretion.” Stecyk v. Bell Helicopter Textron, Inc., 295 F.3d 408, 412 (3d Cir. 2002). Our review of the District Court’s determination that Pennsylvania law applies to the calculation of Travelers’ award of prejudgment interest is de novo. See Hammersmith v. TIG Ins. Co., 480 F.3d 220, 226 (3d Cir.2007) (explaining that a “District Court’s choice of law determination” is subject to plenary review). The same is true of our review of the District Court’s ruling with respect to when post-judgment interest on Travelers’ prejudgment interest began to run, as post-judgment interest is governed by 28 U.S.C. § 1961(a) and “a question of statutory interpretation ... requires de novo review.” Pell v. E.I. DuPont de Nemours & Co., 539 F.3d 292, 305 (3d Cir.2008). III. CHOICE OF LAW The parties agree that New York law governs the reinsurance certificates INA issued to Aetna CS (and which Travelers is seeking to enforce). They further agree that Michigan law governs the policies under which coverage was provided to Acme. Whether Pennsylvania or New York law applies to the calculation of prejudgment interest is in dispute and will be addressed below. IV. DISCUSSION INA is challenging the District Court’s Phase I ruling. Travelers challenges the Court’s Phase II ruling, along with its determinations that Pennsylvania law applies to the calculation of prejudgment interest and that post-judgment interest on the prejudgment interest did not begin to run until the Court’s order quantifying the amount in prejudgment interest owed to Travelers. We discuss each of these challenges in turn. A. The Phase I Ruling The Phase I trial concerned the bad faith exception to the reinsurer’s duty to go along with the coverage provided by the insurer. More specifically, the Phase I trial addressed INA’s allegation that Travelers manipulated its post-settlement allocation to allow it to reach the reinsured XN layer of coverage. The District Court ultimately rejected this challenge, holding that INA had failed to show bad faith on Travelers’ part. We affirm. 1. The Insurer’s Duty of Good Faith in the Post-Settlement Allocation Context The primary purpose of the follow-the-fortunes doctrine is to prevent the reinsurance relationship from interfering with coverage disputes between the insurer and its insured. See CIGNA, 52 F.3d at 1199 (explaining that the doctrine “prevents reinsurers from second guessing good-faith settlements and obtaining de novo review of judgments of the [insurer’s] liability to its insured”). As such, there is some dispute over whether that doctrine should apply to post-settlement allocations, especially where, as here, the allocation decisions being challenged were not the product of active bargaining between the insurer and the insured. See Employers Reins. Corp. v. Newcap Ins. Co., 209 F.Supp.2d 1184, 1191 (D.Kan.2002) (declining to apply the doctrine to an allocation decision that the insurer and insured left unresolved); Graydon S. Staring, Law of Reinsurance § 18.10 (4th ed.2007), at 18-56 (suggesting that the doctrine should only apply to a particular allocation decision “if [it] was necessarily and genuinely part of the claims settlement process”). Nonetheless, the majority view, which INA does not contest, is that the doctrine does apply to post-settlement allocations. See, e.g., Travelers Cas. & Sur. Co. v. Gerling Global Reins. Corp. of Am., 419 F.3d 181, 188-89 (2d Cir.2005); Ace, 361 F.3d at 140-41; Nat’l Union Fire Ins. Co. of Pittsburgh, Pa. v. Am. Re-Ins. Co., 441 F Supp.2d 646, 652-53 (S.D.N.Y.2006); Commercial Union, 9 F.Supp.2d at 67-68. We join those courts here. A contrary holding would risk doing precisely what the follow-the-fortunes doctrine aims to prevent — interfering in settlement negotiations between insurers and their insureds by discouraging a particular type of settlement (here, an all-cash deal). However, “applying the follow-the-[fortunes] doctrine to post-settlement allocation decisions does not leave a rein-surer without protection.” Ace, 361 F.3d at 141. Those allocations must still have been in “good faith” to be binding on the reinsurer. Id. We have broadly characterized the insurer’s duty of good faith to its reinsurer as a duty not to take advantage of the reinsurer’s dependence on the decisions made by the insurer. See CIGNA 52 F.3d at 1216 (“[T]he duty of good faith requires the reinsured to align its interests with those of the reinsurer.”). In the post-settlement allocation context, this means that an insurer breaches this duty when it makes allocation decisions primarily for the purpose of increasing its reinsurance recovery. See Allstate Ins. Co. v. Am. Home Assurance Co., 43 A.D.3d 113, 837 N.Y.S.2d 138, 144 (N.Y.App.Div.2007) (“A reinsurer is not bound by the follow-the-fortunes doctrine where the reinsured’s settlement allocation ... reflects an effort to maximize unreasonably the amount of collectible reinsurance.”); see also Hartford Accident & Indem. Co. v. Columbia Cas. Co., 98 F.Supp.2d 251, 259 (D.Conn. 2000) (denying summary judgment to an insurer “in light of the inferences of unreasonableness or self-service that can be drawn” from the details of its post-settlement allocation). We make clear, however, that the insurer’s negative duty not to make allocation decisions primarily in order to increase reinsurance recovery does not translate into a positive duty on the part of the insurer to minimize its reinsurance recovery. See Gerling, 419 F.3d at 193 (“[An insurer] choosing among several reasonable allocation possibilities is surely not required to choose the allocation that minimizes its reinsurance recovery to avoid a finding of bad faith.” (emphasis in original)). What this means for the reinsurer’s burden of persuasion is that, to establish a breach of the duty of good faith, it is not sufficient simply to demonstrate that a particular allocation decision increased the insurer’s access to reinsurance, at least not where the insurer is able to point to some legitimate (ie., non-reinsurance-related) reason for the challenged decision. See id. (“An allocation that increases reinsurance recovery — -when made in the aftermath of a legitimate settlement and when chosen from multiple possible allocations — would rarely demonstrate bad faith in and of itself.”). To prevail, the reinsurer must either provide direct evidence that the insurer was motivated primarily by reinsurance considerations, or show that the after-the-fact rationales offered by the insurer are not credible. INA attempted to do both in this ease. 2. INA’s Challenges INA challenges the District Court’s Phase I ruling essentially on three grounds. First, it asserts that three of the specific allocation decisions Travelers made — to bypass the post-April 1982 AL policies in allocating the breast implant claims settlement, to allocate the chemical products claims settlement without performing any independent analysis of how those claims matched up to its policies or without allocating any of the settlement amount to the XS policies, and to allocate the entire portion of the settlement dedicated to the breast implant claims as indemnity — are inexplicable except as part of a scheme to maximize Travelers’ reinsurance recovery. Second, INA argues that the Wigmore Memo is direct evidence that Travelers improperly considered reinsurance implications in performing the allocation. And, finally, INA claims that the District Court’s ruling was based on evidence that should have been excluded— namely, evidence that Travelers sought and received legal advice about how to handle its insurance coverage dispute with Acme. i. The Bypassing of the Post-April 1982 AL Policies in Allocating the Breast Implant Claims Settlement In allocating the $80 million of the settlement dedicated to the breast implant claims, Travelers allocated $24 million to the pre-April 1982 AL policies and nothing to the post-April 1982 AL policies. The District Court held that this decision was not evidence of bad-faith maximization because it followed from the “net” nature of the settlement Travelers reached with Acme. (J.A. at 50.) While we disagree with some of the details of the District Court’s analysis, we end up at the same place — ie., concluding that bypassing the post-April 1982 AL policies was reasonable in light of the net nature of the settlement. The District Court’s specific reasoning was that, because the post-April 1982 AL policies were subject to both retrospective premiums and captive reinsurance, “there was no real risk transfer” under those policies, and thus “any allocation to [them] would have contravened the ‘net’ deal with Acme.” (J.A. at 50.) Strictly speaking, however, it is not true that the net nature of the settlement prevented Travelers from allocating to these policies. As INA notes, the only thing Travelers was required to do on account of the net settlement was not bill Acme for either retrospective premiums or captive reinsurance. Indeed, the pre-April 1982 AL policies were also subject to retrospective premiums and captive reinsurance. Yet that did not prevent Travelers from doing, with respect to those policies, what INA contends it was required to do with respect to the post-April 1982 AL policies — allocate to them up to their per-occurrence limits (minus the amount Acme would have owed it in retrospective premiums), and then simply not ask for any captive reinsurance back. The District Court concluded that Travelers was justified in treating the pre-andpost April 1982 AL policies differently for allocation purposes because the latter, but not the former, had unlimited retrospective premiums available to them. But, as both Miley (who was largely responsible for the post-settlement allocation) and Travelers’ expert, Robert Hall, conceded at trial, this difference would not have come into play under a single-occurrence allocation. That is because, under a single-occurrence allocation, the per-occurrence limits would have been exhausted long before the retrospective premiums available under the pre-April 1982 AL policies were exhausted. Indeed, the position advanced at trial by Miley and Hall was that Travelers should have bypassed all the AL policies, not (as the District Court held) that Travelers was correct to bypass some, but not all, of those policies. Nonetheless, we do not agree with INA that, because Travelers was not required to bypass the post-April 1982 AL policies, it was a breach of good faith for it to do so. Because Travelers was under no duty to minimize its reinsurance recovery, the mere fact that it could have, consistent with its agreement with Acme, allocated to all the AL policies does not mean that it was required to do so. The question we think more apt is whether the net nature of the settlement made bypassing the AL policies a reasonable option (even if it was one that, for reasons that are unclear, Travelers only took halfway). See Ger-ling, 419 F.3d at 193 (explaining that the follow-the-fortunes doctrine only requires that an allocation be “reasonable,” not that it be the one allocation “among several reasonable allocation possibilities” that minimizes the burden on the reinsurer). The position advanced by Travelers’ experts at trial was that, because the AL policies did not provide for any significant risk transfer (at least not on a single-occurrence, indemnity-only, allocation), those policies were essentially exhausted by stipulation once the parties agreed to a net deal, thus authorizing Travelers to move on to the next layer of coverage. INA’s experts took the opposite position— namely, that the net deal simply waived Travelers’ right to collect captive reinsurance, but did not alter the basic fact that an insurance policy is only exhausted when money is allocated to it up to the applicable coverage limit. Fortunately for us, we need not wade into this quasi-metaphysical debate over what exhausting an insurance policy “really” requires. In this context, it is enough to note that INA has not shown that Travelers’ position is unreasonable. The theory put forward by Travelers — that the net nature of the deal authorized it to allocate the settlement as if there were a prior step in which, for the policies subject to captive reinsurance, Travelers made payments to Acme and then received that money back — strikes us as plausible. Travelers’ experts, whom the District Court found credible, testified that what Travelers did was consistent with industry practice. Given the very limited nature of the review authorized by the follow-the-fortunes doctrine, that is sufficient, even though there was contrary testimony by INA’s experts. In sum, we agree with the basic direction of the District Court’s analysis, if not all of its details. The decision to bypass the post-April 1982 AL policies was reasonable in light of both the net nature of the deal with Acme and the specific characteristics of those policies. Accordingly, we cannot say that the decision is evidence of bad-faith maximization on Travelers’ part. ii. The Allocation of the Chemical Products Claims Settlement INA’s second challenge focuses on Travelers’ handling of the $20 million of the settlement dedicated to the chemical products claims. INA makes two arguments here. One of those arguments essentially reprises the one considered in the previous section, only this time directed toward Travelers’ decision to bypass the XS layer of coverage in allocating the settlement dollars dedicated to the chemical products claims. The response is basically the same — the XS policies, although they provided coverage for products claims and attached at lower points than the XN policies, were subject to near-total reinsurance from Acme subsidiaries. Accordingly, it was not unreasonable for Travelers to view those policies as exhausted in virtue of the net settlement and skip straight to the XN layer. INA’s other challenge to the handling of the chemical products claims is new, however. It points out that there is nothing in the record to indicate that Travelers ever conducted a detailed analysis of the chemical products claims being brought against Acme before agreeing to allocate $20 million of the settlement to them. Thus, INA argues, Travelers’ handling of these claims could not have been professionally reasonable, as it was not based on an analysis of either Acme’s possible exposure under those claims or how those claims potentially matched up with the specific coverage available. Instead, Travelers began with a number — $20 million — and worked backward from there. We agree with the District Court that, under the circumstances, this approach was not unreasonable. Yessman testified that his main focus was on settling the breast implant claims for an amount Travelers could live with. Prior to the acquisition of Aetna CS by Travelers, Aetna CS had offered Acme a $300 million coverage-in-place deal just for the breast implant claims, which Acme rejected. When Travelers took over Aetna CS’s policies, it set up initial reserves of $100 million for the breast implant claims. Given that, we cannot say that the decision to dedicate $20 million to the chemical products claims, in order to resolve the breast implant claims for only $80 million, was unreasonable, even if it were not grounded in the kind of analysis of the chemical products claims that one expects to see. iii. The Indemnity-Only Allocation Travelers allocated the $80 million dedicated to the breast implant claims exclusively as indemnity. It did this despite the fact that, at the time of the settlement, Acme had spent substantial sums defending itself from those claims ($112 million), but had not yet incurred any liability on them. INA asserts that this is unambiguous evidence of bad-faith maximization on Travelers’ part, since allocating the settlement exclusively as indemnity increased Travelers’ ability to get out of the AL layer of coverage and into the XN layer. We, however, are not convinced that the District Court erred in holding to the contrary. We note initially that, while Acme had yet to incur any liability on the breast implant claims when the settlement was reached, that does not mean that the only reasonable option was to allocate that portion of the settlement exclusively as defense costs. Travelers’ settlement with Acme released all of Acme’s past and future claims under the policies. It was thus not unreasonable for Travelers to view the settlement as covering Acme’s yet-to-materialize liability. This is significant because the only scenario in which paying defense costs would have prevented Travelers from getting out of the AL layer at all — at least on a single-occurrence allocation — was one in which it paid out nothing, or very little, in indemnity. Still, the decision to allocate this portion of the settlement entirely as indemnity, as opposed to dedicating a portion to Acme’s liability and a portion to Acme’s defense costs, likely allowed Travelers to allocate more to the XN layer of coverage than it otherwise could have, as defense costs would not have counted toward the limits of the pre-April 1982 AL policies. Moreover, that decision appears to have been largely unilateral on Travelers’ part, as there is no record of Acme and Travelers ever agreeing to a framework for handling the defense costs versus indemnity issue, nor was that issue addressed in the final settlement agreement. Yet we are not inclined to depart from the District Court’s finding that the decision to allocate the $80 million exclusively as indemnity was sufficiently grounded in Travelers’ prior interactions with Acme that it cannot be characterized as solely an attempt to maximize Travelers’ reinsurance recovery. Yessman and Stonehill-Clafin both testified that, during their early negotiations with Acme, Travelers made a coverage-in-place offer that included $15 million for past defense costs, and $26 million for future defense costs and liability, only to be told that the $15 million for past defense costs was “within ballpark,” but the $26 million figure was “too low.” (Trial Tr. vol. 1, 438, Jan. 12, 2005.) From that, Yessman reportedly concluded that recouping past defense costs was not that important to Acme, a position that both Yessman and one of Travelers’ experts, Jerold Oshinsky, testified was consistent with their experience with large policyholders, who tend to be more interested in achieving coverage certainty going forward than in recovering past losses. In addition, Travelers knew that Acme would be getting defense coverage from another carrier, the Fireman’s Fund, and, as one of INA’s own experts noted, it is typical for a policyholder in Acme’s position to choose one insurer among several to cover defense costs. Given this context, we cannot say that the District Court’s conclusion that “[t]he sum for the breast implant claims was reasonably and in a businesslike manner paid as indemnity and not as defense costs” was clearly erroneous. (J.A. at 44.) iv. The Wigmore Memo INA contends that each of the specific allocation decisions it cites as questionable must be read against the backdrop of the Wigmore Memo, which it characterizes as direct evidence that Travelers allocated the settlement with an eye toward maximizing its reinsurance recovery. Travelers conceded at trial that it is a breach of the duty of good faith for an insurer to take reinsurance implications into account in making coverage decisions. The District Court concluded that the Wigmore Memo was not evidence that Travelers had generally acted in bad faith. It did so on two grounds. First, the Court essentially accepted Yessman’s testimony that the Memo’s purpose was to provide him with a general estimate of Travelers’ potential net exposure on the breast implant claims, which Travelers wanted in relation to an indemnity agreement it had entered into with Aetna U.S. Healthcare when it acquired Aetna CS, and that he did not use the Memo to aid him in his negotiations with Acme. Second, the Court accepted the testimony of Stonehill-Clafin, Miley and Kingston (who, according to Travelers, were the ones responsible for the details of the settlement agreement and the post-settlement allocation), all of whom testified that they never saw, or heard of, the Wigmore Memo prior to the subsequent litigation, and that, as a general matter, they were walled off completely from any information relevant to Travelers’ reinsurance recovery prospects. In its brief, INA cites various reasons why we should not find Yessman credible on this subject, despite the District Court’s having found otherwise. It is unnecessary to detail those reasons here. That is because, even were we to agree that Yessman’s testimony was not credible — and also agree that Yessman’s exposure to the Wigmore Memo contaminated the other Travelers’ employees who worked on the settlement and allocation, and whose credibility INA does not contest — that would not change the result. The Wigmore Memo’s two main points of emphasis are: (1) that, if the breast implant claims were characterized as arising out of multiple occurrences, the coverage was likely to stay in the AL layer; and (2) that, because Acme’s subsidiaries were not obligated to provide reinsurance so long as only defense costs are covered, there was the possibility that Acme would continue to litigate each of the claims without settling. But neither of those points is particularly pertinent to what INA is challenging here. First, as noted above, Travelers had sufficient reasons, apart from concerns about reinsurance recovery, to want to avoid a multiple-occurrence scenario, since such a scenario would have greatly increased its exposure under the AL policies. Thus, the Wigmore Memo is not itself sufficient to show that Travelers insisted on a single-occurrence characterization of the breast implant claims for reasons related to reinsurance recovery (and, at any rate, INA does not raise the number of occurrences issue on appeal). Second, the specific point the Memo raised about defense costs was that, until Travelers covered any losses stemming from liability, it could not access the captive reinsurance. The Memo said nothing about what is at issue here — namely, the effect of defense coverage on Travelers’ ability to recover from its outside reinsurers. In sum, we do not discern clear error regarding the District Court’s finding that “[t]he Wigmore Memo was not used as a ‘roadmap’ by Travelers to negotiate [the] settlement with Acme, or to allocate the settlement dollars to the insurance policies.” (J.A. at 57.) v. The Sword/Shield Doctrine Finally, INA argues that, even if it is not entitled to a reversal of the District Court’s Phase I order, that order should still be vacated, as (according to INA) the District Court based its ruling on evidence that should have been excluded. More specifically, INA contends that it was improper for the District Court to draw inferences in Travelers’ favor based on its consultations with outside and in-house counsel. INA’s argument is that, because Travelers invoked the attorney-client and work-product privileges to shield the substance of those consultations, it should not have been allowed to defend its conduct with reference to those consultations. See United States v. Bilzerian, 926 F.2d 1285, 1292 (2d Cir.1991) (“[T]he attorney-client privilege cannot at once be used as a shield and a sword.... Thus, the privilege may implicitly be waived when [the party claiming the privilege] asserts a claim that in fairness requires examination of protected communications.” (internal citations omitted)). INA is correct that the District Court drew inferences in Travelers’ favor based on the bare fact that, in connection with the Acme settlement, it received advice from attorneys experienced with insurance-coverage disputes. In particular, it found Travelers’ receipt of such advice to be evidence that it generally handled the settlement in a “reasonable” and “businesslike” manner. (J.A. at 61.) But we do not agree that the District Court’s Phase I ruling depended on those inferences. The Court’s conclusion that INA failed to show bad faith on Travelers’ part was sufficiently supported by its findings that: (1) the specific allocation decisions challenged by INA could be accounted for without reference to intentional maximization; and (2) the Travelers employees most responsible for the details of the post-settlement allocation were screened off from the reinsurance implications of their decisions. Accordingly, it is unnecessary for us to determine whether a party that refers generally to having received advice from counsel in order to establish a pattern of businesslike conduct waives any associated privileges. Even were the District Court’s consideration of the disputed testimony error, it was harmless. See Hirst v. Inverness Hotel Corp., 544 F.3d 221, 228 (3d Cir.2008) (“Discretionary evidentiary rulings will give rise to reversible error only where a substantial right of the party is affected.” (internal quotation marks and citation omitted)). s¡: % ^ # In sum, we agree with the District Court that “there is ... enough evidence ... to raise ... suspicions” that Travelers engineered its post-settlement allocation to maximize its reinsurance recovery. (J.A. at 57.) That is why the District Court was correct to deny summary judgment to Travelers, despite the very deferential nature of the scrutiny authorized by the follow-the-fortunes doctrine. However, reviewing the District Court’s findings with the deference we must, we also agree that INA did not meet its burden at trial of showing that the allocation decisions it was challenging were driven primarily by reinsurance considerations. Accordingly, we affirm the Phase I order in Travelers’ favor. B. The Phase II Ruling The Phase II trial addressed one particular aspect of Travelers’ allocation of the settlement — its decision, in allocating the breast implant claims settlement, to treat the per-occurrence limits of the two three-year XN policies reinsured by INA as applying separately to each policy year. INA argued that it was not bound by Travelers’ decision to allocate that much of the breast implant claims settlement to those two policies. That was because (according to INA) that allocation enlarged the limits of those policies beyond what INA agreed to reinsure. The District Court agreed, ruling in INA’s favor. We affirm. 1. The Policy Language It is well settled that “a rein-surer cannot be held liable for a kind of loss that it did not agree to cover.” CIGNA 52 F.3d at 1206-07. “This distinction between reinsured and unreinsured risk is particularly important in facultative reinsurance where the reinsurer accepts only specific risks.” Id. at 1207. At the same time, however, the “ ‘follow the fortunes’ doctrine creates an exception to the general rule that contract interpretation is subject to de novo review.” Id. at 1206. Thus, to prevail in its challenge, INA had to show more than that, of the different reasonable interpretations of the relevant policy language, its is the most persuasive. Rather, INA was required to show that, under Michigan law (which both parties agree controls interpretation of the XN policies), the “underlying policy language ... unambiguously provides that” the per-occurrence limits are not subject to the treatment Travelers gave them (ie., as applying separately to each policy year). Id. at 1207 (internal quotation marks and citation omitted). Under Michigan law, an insurance policy is viewed “much the same as any other contract.” Auto-Owners Ins. Co. v. Churchman, 440 Mich. 560, 489 N.W.2d 431, 433 (1992). That is, an insurance policy is treated as “an agreement between the parties in which a court will determine what the agreement was and effectuate the intent of the parties.” Id. Where “the terms of the contract are clear,” those terms must be enforced “as written.” Frankenmuth Mut. Ins. Co. v. Masters, 460 Mich. 105, 595 N.W.2d 832, 837 (1999). In interpreting such a contract, a court may not “create ambiguity where none exists.” Auto-Owners, 489 N.W.2d at 434. The two XN policies at issue here — 01 XN 247 and 01 XN 752 — are three-year policies. The language of the two policies is identical, except that the limit provided for the former is $4.5 million, while for the latter it is $8 million. Policy 01 XN 752 reads in pertinent part as follows: LIMITS OF LIABILITY: 53.33% ($8,000,000.MAXIMUM) QUOTA SHARE OF $15,000,000. EACH OCCURRENCE 53% ($8,000,000.MAXIMUM) QUOTE SHARE OF $15,000,000. ANNUAL AGGREGATE As this makes clear, the policy is subject to two separate limits — an aggregate limit and a per-occurrence limit. Travelers concedes that the aggregate limit applied only to products claims. Thus, that limit is not at issue here because the breast implant claims were treated as non-product claims. But the aggregate limit is modified by the word “annual,” while the “each occurrence” limit is not. As the District Court noted, that is a strong indication that the aggregate limit was meant to operate annually, while the per-occurrence limit was not. This conclusion is further bolstered by the language of the “Indemnity Agreement” portion of the XN policies, which provides in pertinent part that “[Travelers] will indemnify the INSURED against EXCESS NET LOSS arising out of an accident or occurrence during the policy period.” The clear implication, again, is that the term “accident or occurrence” is linked with the entire policy period — in this case, three years — rather than being linked separately to each policy year. The XN policies do not contain a definition of the term “occurrence.” However, each XN policy incorporates the terms and conditions of its underlying controlling policy (except with respect to certain terms not at issue here). These policies were issued by the Home Insurance Company, and define an “occurrence” as follows: The term occurrence wherever used herein shall mean an accident, or a happening, or event, or a continuous or repeated exposure to conditions which unexpectedly and unintentionally results in personal injury, property damage or advertising liability during the policy period. All such exposure to substantially the same general conditions existing at or emanating from one premise’s location shall be deemed one occurrence. This definition clearly contemplates “continuous” occurrences that are capable of spanning multiple years (e.g., environmental contamination). Such a definition is inconsistent with treating an occurrence as a separate liability trigger for each policy year. In sum, there is nothing in the relevant policy language to suggest any ambiguity with respect to whether the policies’ per-occurrence limits are intended to apply separately to each policy year. On the contrary, the language indicates that the per-occurrence limits — unlike the aggregate limits — are meant to cover the entire policy period. This was the position advanced by INA’s expert, who testified (credibly, according to the District Court) that he had never heard of a company interpreting three-year policies like the ones at issue here in the manner Travelers did. Thus, the District Court’s conclusion that “under Michigan law the three-year XN policies clearly and unambiguously have a single per-occurrence limit for the entire policy period” is well founded. (J.A. at 96.) 2. Travelers’ Argument On appeal, Travelers does not appear to challenge the details of the District Court’s interpretation of the two XN policies (by, for instance, pointing to sources of ambiguity in the policy language overlooked by the Court). Rather, Travelers argues that the kind of analysis performed by the District Court — one that involved “applying general rules of contract construction to in effect predict what a Michigan court would do if presented with the issue” — is insufficient to support a judgment against it. (Travelers’ Br. at 32.) According to Travelers, its right to bind INA to its interpretation of the relevant per-oceurrence limits follows from the fact that “[w]hen Travelers made the annualization decision, there was no Michigan law on the insurance annualization issue, and there was a body of out-of-state law that had reached diametrically different conclus