Opinion by Judge SNEED; Dissent by Judge GRABER
SNEED, Circuit Judge:Appellant AmHS Insurance Company, Risk Retention Group (“RRG”) and Appel-lee Mutual Insurance Company of Arizona (“MICA”) provided professional liability insurance to Dr. Wesley Romberger (“Dr. Romberger”). ' Following a jury trial, Dr. Romberger was found negligent in his care and treatment of Christina Beery. RRG defended Dr. Romberger and satisfied the $7,897,543.18 judgment against him. The parties dispute how much MICA should contribute to the payment of this judgment.
RRG appeals the district court’s determination that it failed to state either a direct or subrogated bad-faith claim against MICA. Both parties appeal the district court’s decision on summary judgment ordering MICA to pay RRG an equitable contribution in the amount of $445,013.83. RRG also appeals the district court’s order establishing that prejudgment interest did not begin to accrue until September 19,1997.
We address each order of the district court in turn. We affirm the dismissal of both the direct and subrogated claims. We reverse the district court’s calculation of MICA’s contribution and remand for further proceedings. We affirm the district court’s determination of the date from which prejudgment interest began to run.
FACTS
Dr. Romberger delivered Christina Beery on September 1, 1986. He subsequently provided care, treatment, and evaluations of Christina through September 23, 1988. In July 1990, Christina Beery was diagnosed with a ventricular septal defect. In March 1992, Christina Beery, by and through her mother, sued Dr. Romberger, alleging negligent failure to detect and diagnose Christina’s heart defect.
The Beery case proceeded to trial in August 1993. MICA contributed 10% of the cost of defending Dr. Romberger while a third insurance carrier, Samaritan, contributed 90%. MICA was continually updated on the Beery litigation, but (other than its commitment to pay 10% of the defense costs) played no role in the defense of Dr. Romberger. The jury returned a verdict in favor of Christina Beery. Through two lump sum payments, RRG paid $7,897,543.18 in complete satisfaction of the judgment. The first of these payments was made in July 1996 in the sum of $4.3 million. The second payment was made in June 1997 in the amount of $3.6 million. RRG informed MICA that it had satisfied the judgment and requested contribution. Both during the Beery litigation and after RRG satisfied the full judgment, MICA offered a maximum of $150,000 toward the total settlement of the *1093case. RRG brought this action for bad faith and contribution against MICA.
I.
Both the viability of the bad-faith claims and the correct computation of MICA’s contribution depend upon whether the competing insurance carriers are primary, excess, or co-excess insurers of the Beery judgment. And, more broadly, on the intended application of the insurance policies. We must review the language of the policies to determine the status of each insurer so as to properly apportion the loss. We begin, therefore, with a brief summary of the principles and purposes of excess insurance. We then identify the relevant portions of the competing insurance policies and categorize them with reference to the “overall insuring scheme.” United Servs. Auto. Ass’n v. Empire Fire & Marine Ins., 134 Ariz. 64, 658 P.2d 712, 714 (1982).
A. Excess Insurance Policies
An “excess” or “umbrella” insurance policy serves a different purpose than a primary policy. A “true” excess policy protects the insured “in the event of a catastrophic loss in which liability exceeds the available primary coverage.” 16 Couch on Insurance § 220:32 (3d ed.1995); See also 8C Insurance Law and Practice § 5071.65 at 107 (1981) (“In this day of uncommon, but possible, enormous verdicts, [excess policies] pick up this exceptional hazard at a small premium”). A primary policy, alternatively, provides coverage from “dollar one” for a given loss.
This clear distinction can be muddied by the inclusion of an “other insurance” clause in an otherwise primary policy. The inclusion of such a clause will not convert a primary policy into “true” excess coverage. The underlying purpose of the primary policy remains the same and it must contribute to an insured’s loss before “true” excess coverage attaches. However, determining whether a given policy is primary (with an other insurance clause) as opposed to excess can sometimes be difficult. 16 Couch on Insurance § 220:32 (“[I]t is extremely difficult to draw any black letter rules of law. There is usually no way ... to avoid doing a time-consuming, complete coverage analysis.”)
Because the instant controversy arises under Arizona law, we rely on the Arizona Supreme Court’s standards for determining when a particular policy is “true” excess insurance. Under Arizona law, a “true” excess policy applies “when the same insured has purchased underlying coverage for the same risk.” St. Paul Fire & Marine Ins. Co. v. Gilmore, 168 Ariz. 159, 812 P.2d 977, 980 (1991). The underlying primary policy “operate[s] as a kind of deductible and ‘an insured pays a reduced premium to the excess carrier expressly because that carrier will be obligated to pay a claim only after a certain amount has been paid’ by the insured’s primary carrier.” Id. (quoting Maricopa County v. Fed. Ins. Co., 157 Ariz. 308, 757 P.2d 112, 114 (1988)). In addition, “true” excess coverage is “written under circumstances where rates were ascertained after giving due consideration to known existing and underlying ... primary policies.” Id. (quoting Loy v. Bunderson, 107 Wis.2d 400, 320 N.W.2d 175, 179 (1982)).
With these standards in mind, we turn to the policies at issue in this appeal.
B. Competing Policies
1. The Samaritan Policy
The Samaritan Policy provided primary insurance to Dr. Romberger. The relevant portion of the Samaritan Policy reads:
(7) Other Insurance: The insurance afforded by this policy is primary insurance, except when stated to apply *1094in excess of or contingent upon the absence of other insurance
b. With regard to physician Insureds, the insurance provided by this policy shall be primary, and it shall not be reduced by the amount of any other insurance the physician Insured may have.
The Samaritan Policy’s coverage of Dr. Romberger commenced October 1, 1986. It contained a policy limit of $1 million per occurrence, $12 million in the aggregate. It is undisputed that the Samaritan Policy provided the first layer of coverage applicable to the Beery judgment. It is also undisputed that Samaritan’s total liability with regard to the Beery judgment is $1 million (including costs of litigation) and that Samaritan has contributed its policy limit.
2. The MICA Policy
The MICA Policy is entitled a “Modified Claims Made Insurance Policy.” It provided coverage to Dr. Romberger from September 1, 1983 through October 25, 1986. The limits of MICA’s policy are $1 million for each occurrence and $1 million in the aggregate.
The MICA policy also provided primary coverage to Dr. Romberger. It covered “accident[s], aet[s] or omission[s] which might give rise to a suit” within the policy period. Claims and actions resulting from any act covered by the policy were also covered. Dr. Romberger paid a total of $90,362 in premiums for this coverage.
The MICA policy, however, also contains an “other insurance” clause, which reads:
This insurance shall not apply unless and until the limits of all other sources of funds have been exhausted. Such sources shall include:
(a) Other insurance;
(b) An insurance plan of a health care institution; and
(c)Any similar source of payment.
3. The RRG Policy
The RRG Policy differs from both the Samaritan and MICA policies in that Dr. Romberger was not its purchaser. Samaritan Health Systems, his insurer, purchased it as “umbrella” coverage for multiple healthcare institutions and contract physicians (such as Dr. Romberger) that are insured, in the first instance, by Samaritan. The RRG policy offered four layers of coverage with a total policy limit of $24 million. Samaritan paid premiums to RRG totaling $7,534,977.
The first layer of RRG coverage was entitled “excess insurance” and provided a policy limit of $1 million for each occurrence, but with no aggregate limit. It became effective October 1, 1986, and was written specifically as excess of the underlying Samaritan Policy. While this layer of coverage did not include a designated, separate “other insurance” clause, it did define loss so as to provide that the insurer had no liability until “deductions for all other recoveries, salvages, or other insurance” were made.
The second layer of RRG coverage, also effective October 1, 1986, was entitled “Hospital Umbrella Liability Policy.” It provided $10 million in insurance “excess of Underlying.” Among the “Underlying” was the Samaritan Policy, the first layer of RRG’s coverage, and several other insurance policies listed in the schedule of underlying coverage. The MICA policy, which Dr. Romberger purchased, was not listed.
Although this second layer of coverage stated that it was excess of the listed underlying policies, it more narrowly defined its coverage. Specifically, this layer of RRG coverage purported to apply only to the “ultimate net loss in excess of the applicable underlying limit.” “Applicable *1095underlying limit” was, in turn, defined as “the total of the limits of the underlying insurance ... and the limits of any other valid and collectible insurance.... ”
In addition, this second layer of coverage included an “other insurance” clause, which reads:
8. OTHER INSURANCE
a. The insurance afforded by this policy shall be excess insurance over any other valid and collectible insurance available to the insured, whether or not described in the Schedule of Underlying Insurance ... and applicable to any part of ultimate net loss, whether such other insurance is stated to be primary, contributing, excess or contingent.
The third layer of RRG coverage became effective October 1, 1987. It provided $5 million in coverage for losses in excess of $10 million. This policy is entitled “Excess Umbrella Liability.” By its express terms, this insurance did not provide coverage until $10 million in underlying insurance had been exhausted.
The fourth and final layer of RRG coverage became effective October 1, 1988. It provided $8 million in coverage for losses in excess of $15 million. However, this final layer of RRG coverage did not become effective until after Dr. Romberger’s care of Christina Beery ended. The district court held that this layer of coverage was not applicable to the Beery judgment. MICA did not appeal this ruling.
II.
RRG, Samaritan’s excess carrier, first appeals the dismissal of both its subrogated and direct claims for bad faith against MICA. The district court held that the RRG and MICA policies provided equal-level insurance. Under Arizona law, an insurer owes no duty of good faith to a co-equal insurer. On this basis, the district court dismissed RRG’s bad-faith claims under Fed.R.Civ.P. 12(b)(6). Hartford Accident & Indem. Co. v. Aetna Cas. & Sur. Co., 164 Ariz. 286, 792 P.2d 749 (1990) (bad-faith claim may be brought only by an excess insurer against a primary insurer).
But did RRG provide, as it contends, “true” excess coverage? This is a question of law that requires analysis of the policy as a whole. Nichols v. State Farm Fire & Cas. Co., 175 Ariz. 354, 857 P.2d 406, 407 (1993) (interpretation of an insurance policy is a question of law to be determined by the court); United Servs. Auto. Ass’n, 653 P.2d at 714 (“[W]e look to the language of each policy in light of the circumstances of each contracting party to determine the intent within the framework of an overall insuring scheme.”). Thus, we must review de novo the district court’s determination that RRG and MICA provided equal-level coverage of the Beery judgment. Johnson v. Cont’l Ins. Co., 198 Ariz. 160, 7 P.3d 966, 968 (2000) (“Interpretation of an insurance contract is a question of law that we decide independently of the trial court’s legal conclusions.”).
A. Proceedings Below
The district court, in dismissing RRG’s bad-faith claim, relied on State Farm Mut. Auto. Ins. Co. v. Bogart, 149 Ariz. 145, 717 P.2d 449 (1986). In Bogart, the Arizona Supreme Court held that “other insurance” clauses in otherwise equal-level policies are “mutually repugnant” and therefore void. Id. at 453. The Bogart decision, however, is applicable only in disputes “between two insurers that provide 'primary coverage for the same occurrence, one of which seeks to avoid all liability by reason of ... its ‘other insurance’ clause.” Id. at 454 (emphasis added). Bogart says nothing about adjudicat*1096ing disputes between a primary insurer and a “true” excess insurer. Indeed, an “other insurance” clause in an otherwise primary policy cannot affect the rights of a “true” excess insurer. See Transport Indem. Co. v. Carolina Cas. Ins. Co., 133 Ariz. 395, 652 P.2d 134, 142 (1982) (policy providing primary coverage with “other insurance” clause must pay before policy “that extends only excess coverage”).
MICA concedes that its policy is a primary policy with an “other insurance” clause. The MICA policy, therefore, is only excess of other primary policies that do not contain “other insurance” clauses (i.e., the Samaritan policy). MICA, however, must share the insured’s loss on a pro rata basis with other primary policies that do contain “other insurance” clauses. See Bogart, 717 P.2d at 453. And, of course, it must tender its full policy limit before a “true” excess carrier is required to pay.
The dispute here is whether RRG’s policies became primary upon the exhaustion of the underlying Samaritan policy or remained excess to all primary policies. This issue cannot be resolved through application of a bright-line rule. Rather, we must read the policies as a whole “in order to give a reasonable and harmonious meaning and effect to all of [their] provisions.” Droz v. Paul Revere Life Ins. Co., 1 Ariz.App. 581, 405 P.2d 833, 835 (1965); see also Gilmore, 812 P.2d at 983 (“[T]he type of policy is determined by the type of coverage provided, not by the label affixed by the insurer.”).
B. RRG’s Policies
If RRG intended its policies to attach only upon the exhaustion of the underlying Samaritan policy and charged premiums consistent with that risk, it should be held to provide coverage upon such an occurrence. See 20th Century Ins. Co. v. Liberty Mut. Ins. Co., 965 F.2d 747, 755-757 (9th Cir.1992) (looking to “intended application of each policy” rather than “judicially created labels” such as “primary, secondary, etc.” and holding that excess insurer was excess only of specified primary carrier); U.S. Fire Ins. Co. v. Aetna Cas. & Sur. Co., 781 S.W.2d 394, 398 (1989) (excess policy does not “automatically overlay every applicable primary policy that contains an ‘other insurance’ clause”); Canal Ins. Co. v. United States Fidelity and Guar. Co., 149 Ariz. 578, 720 P.2d 963, 965 (1986) (excess policy became primary when specific underlying policy was exhausted).
Alternatively, if RRG wrote its policies as “true excess,” its premiums would reflect the reduced probability that it would ever be called on to provide coverage. Maricopa County, 757 P.2d at 114. “True excess” policies should not be asked to contribute until all primary policies have been exhausted. Am. Family Mut. Ins. Co. v. Cont'l Cas. Co., No.2001 WL 184770, *3, 23 P.3d 664 (2001) (quoting United Servs. Auto Ass’n, 653 P.2d at 714) (“ ‘[I]nsurers who issue residual protection only are last to pay so long as that is their expressed intent.’ ”).
Because three separate RRG policies with different terms covered Dr. Romber-ger during his treatment of Christina Beery, each must be examined to determine whether the RRG policies are excess of MICA’s primary policy.
1. RRG’s First and Second Layers of Coverage
We hold that the first two layers of RRG coverage were intended to be specific excess insurance that attached upon the exhaustion of the underlying Samaritan Policy. These layers of coverage, like MICA’s coverage, were applicable to any loss in excess of the Samaritan Policy limit. Thus, MICA must share any loss over $1 million on a pro rata basis with RRG’s *1097first two layers of coverage. In reaching this conclusion, we rely on several aspects of the RRG policy that, taken as a whole, reveal its place in the overall coverage scheme.
a)RRG had no knowledge of the MICA policy when it provided coverage to Dr. Romberger
In Gilmore, the Arizona Supreme Court noted that “true” excess coverage is “written under circumstances where rates were ascertained after giving due consideration to known existing and underlying basic or primary policies.” 812 P.2d at 980 (citation and internal quotation marks omitted). The first two layers of RRG coverage were written and priced with consideration only of enumerated underlying policies. The MICA policy was not one of the listed underlying policies.
The first layer of the RRG policy provides the following coverage: “$1,000,-000/no aggregate in excess of $1,000,000/ $12,000,000.” The “schedule of coverage endorsement” indicates that the “$1,000, 000/$12,000,000” refers to the listed underlying insurance. The listed insurance, in turn, includes the Samaritan policy (with limits of $1 million per occurrence and $12 million aggregate) and several other insurance policies. This list does not include the MICA policy. By its own terms, this layer of RRG coverage applies to losses resulting from an occurrence and exceeding $1 million. It was written as excess of a specific underlying policy (the Samaritan policy) which provided primary insurance in the required amount of $1 million per occurrence and $12 million in the aggregate.
Similarly, the second layer of RRG coverage contains a list of enumerated underlying policies and provides $10 million in coverage “excess of Underlying.” There is no suggestion anywhere in the policy that RRG knew of the MICA policy when writing this layer of coverage. RRG, therefore, did not price the policy based on the existence of an additional $1 million in underlying coverage provided by MICA.
To repeat, neither of the first two layers of RRG coverage required the insured to maintain any additional coverage beyond that provided by the primary Samaritan policy. These provisions make clear that RRG was neither aware of nor gave consideration to the existing MICA policy. Rather, RRG’s first two layers of insurance provided coverage that would attach immediately upon the exhaustion of the underlying Samaritan policy. They should be enforced as intended.
b) RRG and MICA did not cover the “same risk”
The Arizona Supreme Court has also noted that “true” excess coverage applies only “when the same insured has purchased underlying coverage for the same risk.” Gilmore, 812 P.2d at 980. In the present case, the RRG and MICA policies did not apply to the “same risk.” RRG covered Dr. Romberger for 24 of the 25 months that he provided treatment to Christina Beery. MICA insured Dr. Rom-berger for only two months after the birth of Christina Beery. The two policies overlapped for only thirty days. The thirty-day overlap does not support RRG’s contention that it is a “true” excess insurer of the Beery judgment. The relevant inquiry is whether the two insurance companies insured the “same risk.” The risk assumed by the two insurers in this case was markedly different.
c) RRG failed to ascertain the total level of primary insurance
RRG provided its coverage to Samaritan several years after MICA’s policy took effect. RRG had the opportunity to, and should have, “taken steps to avoid the *1098confusion, uncertainty and now the litigation produced by the overlapping policies.” Executive Risk Specialty Ins. Co. v. Lexington Ins. Co., 106 F.Supp.2d 181, 189 (D.Mass.2000) (applying Arizona law). In Executive Risk, the court concluded that a specific excess policy was excess to an unrelated primary policy with an “other insurance” clause. However, the court reached this conclusion only after examining the “context within which the policies were written.” Id. at 183. Because the primary insurer wrote its policies with knowledge of overlapping coverage, it “should bear the risk of any doubt it could have avoided.” Id. at 189. In this case, RRG could have avoided the present dispute by ascertaining the total level of existing primary coverage prior to issuing its policy. This factor weighs against RRG.1
d) The RRG policies contain no relevant coverage exceptions
The fact that the RRG policies include provisions purporting to accept liability only “after making deductions for all other recoveries, salvages or other insurance” does not alter our analysis.2 These provisions cannot be given the effect of exceptions to coverage rather than excess clauses. “By definition, an exception or exclusion provides that there is no coverage regardless of the existence of other insurance.” Fremont Indem. Co. v. New England Reinsurance Co., 168 Ariz. 476, 815 P.2d 403, 406 (1991). In this case, RRG clearly provided insurance covering the Beery judgment. RRG intended, however, for its coverage obligation to change depending on the existence of coverage by other valid insurance.
These provisions are not exclusions; they are typical excess insurance clauses. “The gist of the hybrid escape-excess clause is to permit escape if the loss is less than any other insurance protection and to provide excess insurance if its coverage exceeds the other valid insurance.” Id. at 407. That RRG inserted its “other insurance” clauses in the definition of loss rather than as separate provisions is irrelevant. “[W]e cannot agree with the theory ... that such a clause is transformed into an exception simply because of its location in an insuring agreement as opposed to another portion of a policy. As a general rule, insurers cannot gain an advantage merely by rearranging ‘other insurance’ provisions.” Id. at 406 (citations omitted). Mutually repugnant “other insurance” clauses are void regardless of where in the insuring agreement they are located. Jefferson Ins. Co. v. Glens Falls Ins. Co., 88 A.D.2d 925, 450 N.Y.S.2d 888, 889 890 (1982) (ultimate net loss provision in one policy and other insurance clause in competing policy “cancel out each other” and each insurer contributes pro rata to settlement).
e) RRG’s alternative conclusion is not supported by Arizona precedent
RRG does not contend that it wrote its policy as excess of the MICA policy. Rather, it argues that a policy that is excess to any primary policy is excess to each and every primary policy covering the same loss. RRG relies on two cases in support of this proposition: Ariz. Joint Underwriting Plan v. Glacier Gen. Assurance Co., 129 Ariz. 351, 631 P.2d 133 (1981) (“Glacier ”) and United Servs. Auto. Ass’n., 134 Ariz. 64, 653 P.2d 712. In both Glacier and United Services, the Arizona *1099appellate court held that two primary policies should pay their full policy limits before an excess policy is compelled to pay. Neither case supports RRG’s position because neither case involved competing policies both of which were indisputably excess to a single primary policy.
In Glacier, the court held, as we hold today, that a specific excess policy attaches upon the exhaustion of its underlying primary policy. “AJUP was hable for the excess of the total applicable limits of its underlying insurance, the MICA policy.” Glacier, 631 P.2d at 135. In Glacier, however, the underlying primary policy was not exhausted until a co-equal primary policy contributed its pro rata share. Consequently, the excess insurer’s liability did not attach until both the underlying policy and an unrelated primary policy had contributed. This fact distinguishes Glacier from the present litigation.
Here, the underlying Samaritan policy tendered its entire $1 million policy limit before any other insurer’s obligation attached. As was the case in Glacier, the excess carrier’s coverage (in this case, RRG) attached upon the exhaustion of its underlying primary policy (Samaritan). Unlike Glacier, however, the underlying Samaritan policy is the only first-level policy applicable to the insured’s loss. MICA’s duty to contribute to the Beery judgment arose at the same time that RRG’s obligation attached — after the Samaritan policy was exhausted. MICA, therefore, should contribute pro rata with the RRG policy to cover any loss above Samaritan’s $1 million policy limit.
Similarly, in United Services, the court allocated the loss among two primary insurers and an excess policy. The United Services court analyzed “the language of each policy in light of the circumstances of each contracting party to determine the intent within the framework of an overall insuring scheme.” United Services, 653 P.2d at 714. This analysis revealed that it was the “expressed intent” of the excess policy that it be the “last to pay.” Id.
Our analysis of the competing RRG and MICA policies is similar. For the reasons already noted, we conclude that it was not RRG’s “expressed intent” that its first two layers of coverage overlay all primary policies. Reading the first two layers of RRG coverage within the context of the “overall insuring scheme,” we hold that RRG, in authoring and pricing these policies, gave consideration only to that coverage listed in its revised schedule of underlying coverage. That is, RRG wrote its policies to provide coverage in excess of the underlying Samaritan policy.3 It now wishes to treat these policies as excess of all insurance that may fortuitously apply to a given loss. Gilmore, 812 P.2d at 981. We decline to provide RRG with this windfall.
2. RRG’s Third Layer of Coverage
The third layer of RRG coverage repeatedly states that it applies only to losses in excess of $10 million. The schedule of coverage endorsement, the declarations page, and the limits of liability clause all note that the coverage is in excess of the $10 million in coverage provided in the underlying policies. Unlike the first and second layers of RRG coverage, this policy does not purport to attach upon the exhaustion of a specific underlying policy. Rather, coverage under this policy attaches only after exhaustion of a specified policy amount. “Exhaustion of the [$10 million] amount is a fixed policy require*1100ment; it was not satisfied and this fact cannot be altered by language in other policies.” Maricopa County, 757 P.2d at 114.
The third layer of the RRG policy, therefore, is excess of all insurance up to $10 million, including the MICA policy. The district court erred in holding that this layer of coverage was an equal layer of coverage with the MICA policy.
III.
RRG’s Bad-Faith Claims
Our determination that the first two layers of RRG coverage and the MICA policy are equal-level insurers of the judgment against Dr. Romberger controls the analysis of RRG’s bad-faith claims. Under Arizona law, an insurer has neither a direct nor a subrogated claim for bad faith against an equal-level insurer. Twin City Fire Ins. Co. v. Superior Court, 164 Ariz. 295, 792 P.2d 758, 760 (1990) (no direct claim for bad faith against other insurer); Hartford Accident & Indemnity, 164 Ariz. 286, 792 P.2d 749 (subrogated claim may only be brought by excess insurer against primary carrier). The district court correctly dismissed these claims.
A. Direct Bad-faith Claim
Under Arizona law, an insurer owes only the insured a duty of good faith. “[T]he insurer’s obligation to settle, as well as the obligation to defend, arises out of the contract between the parties.” State Farm Auto. Ins. Co. v. Civil Service Employees Ins. Co., 19 Ariz.App. 594, 509 P.2d 725, 733 (1973). Because MICA never entered into a contract with RRG, it owed no duty to RRG and cannot be liable on a direct claim for bad faith. This reasoning is equally applicable without regard to whether RRG is a “true” excess carrier or merely a co-excess insurer of the Beery judgment.
In Twin City, 792 P.2d at 758, the Arizona Supreme Court held that an excess carrier could not bring a direct claim for bad faith against a primary carrier. Although there is language in Twin City that suggests such a cause of action is possible in some circumstances, a subsequent decision from the Arizona Court of Appeals interpreted Twin City as foreclosing all direct-duty claims initiated by an excess carrier against a primary insurer. Cal. Cas. Ins. Co. v. State Farm Mut. Auto. Ins. Co., 185 Ariz. 165, 913 P.2d 505, 510 (1996) (excess insurer may assert “only a claim that derives from the primary’s contract with the insured”).4 The district court correctly dismissed RRG’s direct claim for bad faith against MICA.
B. Subrogated Bad-faith Claim
Arizona recognizes the right of an excess insurer to bring a subrogated claim against a primary carrier for bad-faith failure to settle within the primary carrier’s policy limits. Hartford Accident, 792 P.2d at 749. No such claim exists between two equal-level insurers.
Equal-level insurers have an identical duty of good faith toward the insured. A complete failure to settle is necessarily a *1101failure by both insurers. It follows that RRG cannot refuse to settle the Beery litigation on the one hand and, on the other, contend that MICA’s refusal to settle was made in bad faith. Because RRG and MICA had equivalent obligations to the insured, the proper approach to resolving their dispute is not a suit for bad faith, but one for contribution. See St. Paul Fire & Marine Ins. v. Allstate Ins. Co., 25 Ariz.App. 309, 543 P.2d 147 (1975) (co-primary insurer can bring equitable contribution claim); Mut. Ins. Co. of Ariz. v. Am. Cas. Co. of Reading, Pa., 189 Ariz. 22, 938 P.2d 71, 75-76 (1996).
No Arizona court has recognized the right of an insurer to bring a subrogated bad-faith claim against an equal-level insurer. The rationale for recognizing an excess insurer’s right to bring such an action against a primary insurer does not apply in an action between equal-level insurers. As MICA and RRG were equal-level insurers, they had the same duty to enter a good-faith settlement. The district court, therefore, correctly dismissed RRG’s subrogated bad-faith claim against MICA.
IV.
Contribution
Having determined that RRG and MICA are equal-level insurers, the district court prorated the liability for the judgment in the Beery litigation. RRG argues that the district court incorrectly ascertained the proper contribution level because it used the wrong formula. We hold that the district court used the proper formula, but erred in calculating the total amount of applicable insurance.
The district court prorated liability between RRG and MICA according to the “policy limits” approach. Under this approach, each insurer’s liability is determined by dividing its policy limit by the total amount of coverage. In this case, the district court found that the total amount of applicable coverage was $17 million (the sum of MICA’s $1 million and RRG’s $16 million policy limits). Under the “policy limits” approach, MICA would be responsible for Yu of the total judgment paid. The judgment (minus the contribution from Samaritan of the remainder of its policy limit) totaled $7,565,106.33. The district court determined that MICA’s contribution, exclusive of interest, was $445,013.83 under the “policy limits” approach.
RRG, to reduce its share of the aggregate liability, argues that proration should be done according to the “maximum loss” rule. Under this rule, the court must determine the maximum amount each insurer was potentially obligated to pay absent the competing insurance policy. In this instance, MICA’s “maximum loss” was $1 million and RRG’s “maximum loss” was $7,565,106.33. RRG would contribute approximately $7.56 for every dollar contributed by MICA until the judgment was satisfied. Under this approach, MICA’s share would have amounted to $883,226.16. See Employers Mut. Cas. Co. v. MFA Mut. Ins. Co., 384 F.2d 111, 115 (10th Cir.1967) (where one insurer’s maximum loss was $10,000 and competing insurer’s maximum loss was $5,000, judgment should be prorated on a two-to-one basis).
In choosing the “policy limits” approach, the district court relied on several decisions, applying Arizona law, which prorate equitable contributions according to policy limits. The first of these decisions was a Ninth Circuit case predicting the course the Arizona Supreme Court was likely to take. See Weekes v. Atlantic Nat’l Ins. Co., 370 F.2d 264, 274 (9th Cir.1966). Our use of the “policy limits” approach in Weekes was followed by the Arizona Court of Appeals in both Harbor Ins. Co. v. United Services Auto. Ass’n, 114 Ariz. 58, *1102559 P.2d 178, 183 (1976) (prorating according to policy limits) and A.H. v. Ariz. Prop. and Cas. Ins. Guar. Fund, 189 Ariz. 378, 943 P.2d 738, 747 (1996) (same). In addition, in Bogart, 149 Ariz. 145, 717 P.2d 449, the Arizona Supreme Court affirmed the lower court’s use of the “policy limits” approach.
RRG, with some reason, contends that these cases do not establish that Arizona follows the “policy limits” approach. RRG argues that Weekes, Harbor Ins., and A.H v. Arizona Property all involved insurance companies with equal policy limits. RRG points out that when two insurers have the same policy limits, their contributions would be equal without regard to whether the court uses the “policy limits” or the “maximum loss” rule. The courts’ use of the “policy limits” approach in these cases, therefore, was unnecessary to the calculation of each party’s contribution. In addition, RRG argues that in Bogart neither party contested the proration formula. The Bogart court affirmed the application of the “policy limits” approach without discussion and in a case in which neither party challenged the proration formula on appeal.
RRG, instead, asks us to follow the holding in Industrial Indemnity Co. v. Beeson, 132 Ariz. 503, 647 P.2d 634 (1982). In Beeson, the Arizona appellate court specifically disapproved the “policy limits” approach and remanded for a determination of liability with instructions to use the “maximum loss” rule to calculate contribution. “We therefore believe the more equitable basis for proration should be according to the maximum loss which each company could have sustained in the particular case, absent the other insurance coverage.” Id. at 639-40.
In choosing a proration formula, we are guided by our decision in Sec. Pac. Nat’l. Bank v. Kirkland (In re Kirkland), 915 F.2d 1236, 1240 (9th Cir.1990). There we were called on to predict the course the California Supreme Court would take if faced with a bankruptcy question of first impression. In Kirkland, we relied on two specific factors. First, we surveyed the lower California courts and determined how a majority of those courts had addressed the same question. Second, we noted that the majority rule in California was also the majority approach in other jurisdictions. Both of these factors favor adoption of the policy limits approach in the present litigation.
As noted above, the weight of Arizona judicial authority supports use of the “policy limits” approach. Although no Arizona court has offered a detailed rationale for applying this rule, both the Arizona Supreme Court and Arizona appellate courts have — with one exception — held that proration according to policy limits is appropriate. The “policy limits” approach, moreover, is consistent with the law in a majority of other jurisdictions. See Am. Cas. Co. v. PHICO Ins. Co., 549 Pa. 682, 702 A.2d 1050, 1053 n. 4 (1997) (twelve of seventeen jurisdictions that have adopted an allocation formula since 1975 have adopted the policy limits approach); Ostrager & Newman, Handbook on Insurance Coverage Disputes, § 11.04 (9th ed.1998) (describing policy limits approach as the majority rule and citing cases). The district court, therefore, did not err in prorating liability according to the policy limits of the respective policies.
As detailed above, however, we hold that only the first two layers of the RRG policy covered the loss incurred by the Beery litigation. Therefore, we reverse the district court’s order awarding contribution to RRG and remand with instructions to calculate MICA’s contribution based on the policy limits of RRG’s first two layers of coverage and MICA’s policy.
*1103V.
Prejudgment Interest
The district court held that RRG’s contribution claim against MICA was a liquidated claim. Under Arizona law, “prejudgment interest on a liquidated claim is a matter of right.” Gemstar Ltd. v. Ernst & Young, 185 Ariz. 493, 917 P.2d 222, 237 (1996) (en banc). The rate of interest is set by statute at ten percent per annum. Ariz.Rev.Stat. Ann. § 44-1201.
Neither party contests the district court’s determination that prejudgment interest must be awarded. However, RRG argues that the trial court erroneously calculated the amount of interest. Specifically, the district court held that prejudgment interest began to accrue on September 19, 1997 — the day that RRG sent to MICA a complete copy of the relevant insurance policies. RRG contends that interest actually began to accrue on the dates that it paid the judgment in the Beery litigation. RRG paid $4.3 million in partial satisfaction of the judgment in July 1996. It made a second payment of $3.6 million in June 1997.
The district court has discretion to determine the date of commencement of prejudgment interest. Trus Joist Corp. v. Safeco Ins. Co. of Am., 735 P.2d 125, 140 (1986) (holding that “the trial court did not abuse its discretion in commencing the prejudgment interest as of [a specified date]”). The district court’s determination that prejudgment interest did not begin to accrue until RRG provided MICA with a copy of its policy was not an abuse of discretion.
Under Arizona law, prejudgment interest begins when the creditor provides to the debtor “sufficient information and supporting data so as to enable the debtor to ascertain the amount owed.” Homes & Son Constr. Co. Inc. v. Bolo Corp., 22 Ariz.App. 303, 526 P.2d 1258, 1261 (1974). RRG notified MICA of the amount paid out in the Beery judgment at the time the money was paid. The district court found, however, that this was not sufficient to enable MICA to ascertain its liability with “reasonable exactness.” Id. at 1262. Under the “policy limits” approach to prorating contribution, MICA’s liability could not be discerned without reference to the total available insurance. RRG, therefore, was under a duty to inform MICA of the total policy limit applicable to the Beery judgment. RRG did not do so until September 19,1997.
The district court’s order awarding prejudgment interest after September 19 was a reasonable application of Arizona law governing the timing of prejudgment interest. The district court did not abuse its discretion in reaching this conclusion. See Wing v. Asarco Inc., 114 F.3d 986, 988 (9th Cir.1997) (reversal justified only when the lower court’s decision “is clearly against the logic and effect of the facts as are found”) (citations and internal quotation marks omitted).
VI.
The district court’s dismissals of both the direct and subrogated bad-faith claims against MICA are affirmed. Arizona law does not recognize the right of an excess insurer to sue a co-excess insurer for bad faith. The district court’s calculation of MICA’s contribution is reversed and remanded for the limited purpose of calculating MICA’s contribution consistent with this opinion. The district court’s award of prejudgment interest is affirmed. Each party shall bear its own costs.
AFFIRMED IN PART, REVERSED IN PART, AND REMANDED
. RRG also could have written its policy to attach only after the insured incurred a specified financial loss.
. The quoted phrase is from RRG's first layer of coverage. An analogous provision is included in RRG’s second layer of coverage. See supra Section 1(B)(3).
. MICA argues in its cross-appeal that the first layer of RRG coverage is actually primary coverage that must be exhausted before any coverage under the MICA policy attaches. However, the RRG policy, read as a whole, is unmistakably excess of the underlying Samaritan policy.
. At most, Twin City provides a narrow exception to the general rule that only the insured (or one subrogated to the rights of the insured) can bring an action for bad faith. That exception provides that an excess insurer can bring a direct-duty bad-faith claim against a primary insurer only if the insured has engaged in wrongful conduct which consequently bars equitable subrogation. Twin City, 792 P.2d at 760. Neither of the conditions necessary for such a claim is present here. For the reasons noted above, RRG is not excess of MICA with regard to the Beery judgment. In addition, there has been no wrongful conduct on the part of the insured that could undermine an equitable subrogation claim.