Justice KITTREDGE.

In this commercial general liability ("CGL") policy dispute, we issued an opinion on January 7, 2011, finding no coverage. We subsequently granted a rehearing petition and received numerous amici briefs. Today, we withdraw our initial opinion and issue this opinion, finding the CGL policies provide coverage for the stipulated progressive property damages.

Appellant/Respondent Harleysville Mutual Insurance Company ("Harleysville") issued a series of standard CGL policies to the Respondent developers or their predecessors (collectively "Crossmann") for a series of condominium projects in the Myrtle Beach area of South Carolina. The exterior components of the condominium projects were negligently constructed, which resulted in water penetration and progressive damage to otherwise nondefective components of the condominium projects. The homeowners settled their lawsuits against Respondents. Crossmann then filed this declaratory judgment action to determine coverage under Harleysville's policies. Prior to trial, several of Crossmann's other insurers settled with Crossmann, providing coverage for the homeowners' claims. Based on the remaining parties' stipulations and our suggestion in L-J, Inc. v. Bituminous Fire & Marine Insurance Co., 366 S.C. 117, 123 n. 4, 621 S.E.2d 33, 36 n. 4 (2005), that a "CGL policy may . . . provide coverage in cases where faulty workmanship causes . . . damage to other property," the trial court determined the homeowners' claims were covered by Harleysville's policies.1 We affirm the finding of coverage.

The finding of coverage requires that we address a matter not reached in our initial and now withdrawn opinion. Harleysville appeals from the trial court's determination that its liability to Crossmann is joint and several with Crossmann's other CGL insurers. We reverse the finding of joint and several liability and find the scope of Harleysville's liability is limited to damages accrued during its "time on the risk." In so ruling, we adhere to our holding in Joe Harden Builders, Inc. v. Aetna Casualty & Surety Co., 326 S.C. 231, 486 S.E.2d 89 (1997), but overrule Century Indemnity Co. v. Golden Hills Builders, Inc., 348 S.C. 559, 561 S.E.2d 355 (2002). Using our "time on risk" framework, the allocation of the damage award against Crossmann must conform to the actual distribution of property damage across the progressive damage period. Where proof of the actual property damage distribution is not available, the allocation formula adopted herein will serve as an appropriate default method for dividing the loss among Crossmann's insurers. We remand to the trial court for further consideration of the "time on risk" allocation.

I.

Crossmann constructed multiple condominium projects from 1992 through 1999, which are at issue in this case. Crossmann utilized subcontractors to construct the condominium projects. In 2001, the homeowners filed suit against Crossmann after they discovered construction defects and resulting problems with the units. The homeowners alleged Crossmann defectively constructed the units, and as a result, the units experienced substantial decay and deterioration. Crossmann settled with the homeowners for approximately $16.8 million.

Following the settlement, Crossmann sought coverage for damages arising out of the lawsuit pursuant to their CGL policies issued by Harleysville, but Harleysville denied that coverage was triggered. Crossmann filed a declaratory judgment action to determine whether the policies covered the homeowners' damages. The parties stipulated to the facts and amount of damages and only presented the coverage and allocation questions to the trial court.

The parties' stipulations presented the coverage question on the basis of the presence or absence of an "occurrence." The parties stipulated to, among other things, the amount of damages, that the damage resulting from water intrusion constituted "property damage," that the damage began within thirty days after the certificate of occupancy was issued for each building, that the damage progressed until repaired or until Beazer Homes paid to settle the homeowners' claims, and that the parties would not argue the applicability of any policy exclusions.

Relying on L-J, Inc. v. Bituminous Fire & Marine Insurance Co., 366 S.C. 117, 123 n. 4, 621 S.E.2d 33, 36 n. 4 (2005), and the ambiguity in the language of the CGL policies, the trial court ruled that the progressive damage "that resulted from, and was in addition to, the subcontractors' negligent work itself" was caused by an "occurrence." The trial court issued its order prior to our recent decision in Auto Owners Insurance Co. v. Newman ("Newman"), 385 S.C. 187, 684 S.E.2d 541 (2009). However, as discussed below, Newman further supports this result. The trial court also ruled that Harleysville was jointly and severally liable and was not entitled to a set-off based on other insurers' pre-trial settlements with Crossmann.2 Additionally, the trial court found that Crossmann was entitled to an award of post-judgment interest but not prejudgment interest. Both parties have appealed the trial court's order.

II.

"A declaratory judgment action is neither legal nor equitable, and therefore, the standard of review is determined by the nature of the underlying issue." Newman, 385 S.C. at 191, 684 S.E.2d at 543. "When the purpose of the underlying dispute is to determine whether coverage exists under an insurance policy, the action is one at law." Id. "In an action at law tried without a jury, the appellate court will not disturb the trial court's findings of fact unless there is no evidence to reasonably support them." Id. In this case, the parties have stipulated to the facts, and thus we are presented with a question of law. Where the action presents a question of law, as does this declaratory action, this Court's review is plenary and without deference to the trial court. J.K. Constr., Inc. v. W. Carolina Reg'l Sewer Authority, 336 S.C. 162, 166, 519 S.E.2d 561, 563 (1999).

III.

Commercial General Liability Policies and the Coverage Question in a Progressive Damage Case

A.

We affirm the trial court's finding of coverage based on an "occurrence." An occurrence was once simply defined as an "accident." However, in 1966, the occurrence definition was expanded to include "continuous or repeated exposure to substantially the same general harmful conditions."3 This Court, among others, has struggled to discern the meaning of the expanded occurrence definition in the context of progressive damage cases. The lack of a clear meaning, we believe, leaves us with an ambiguity, which we must construe against the insurer. See Super Duper Inc. v. Pennsylvania Nat'l Mut. Cas. Ins. Co., 385 S.C. 201, 210, 683 S.E.2d 792, 796 (2009) ("Ambiguous terms must be construed in favor of the insured."). Accordingly, we construe the ambiguous definition of occurrence in favor of the insured, Crossmann, and find the insuring language of the policies was triggered by the damages caused by repeated water intrusion.4

B.

While we adhere to the result in Newman, because progressive damage cases often are highly complex and involve many stakeholders, we elect to clarify the applicable legal framework for determining whether coverage is triggered.

In Newman, a homeowner brought a suit against a builder alleging breach of warranty, breach of contract, and negligence. The homeowner established that a subcontractor negligently applied stucco to the side of her house and, as a result, progressive damage ensued as water seeped into the home causing damage to the home's framing and exterior sheathing.

We held that the costs of replacing the defective application of the stucco were not covered by the builder's CGL policy, but the damage caused by the continuous moisture intrusion resulting from the subcontractor's negligence did fall within the CGL's expansive definition of an occurrence. We analyzed Newman solely through the lens of whether there was an occurrence, specifically a "continuous or repeated exposure to substantially the same harmful conditions." 385 S.C. at 194, 684 S.E.2d at 544-45 (citing Travelers Indem. Co. of America v. Moore & Assocs., Inc., 216 S.W.3d 302, 309 (Tenn.2007)).

We believe a more complete understanding of the coverage issue in this kind of progressive property damage case should involve the policy term "property damage." The standard CGL policy defines "property damage" in two different ways, as follows:

With respect to the first quoted definition of "property damage," the critical phrase is "physical injury," which suggests the property was not defective at the outset, but rather was initially proper and injured thereafter. We emphasize the "difference between a claim for the costs of repairing or removing defective work, which is not a claim for `property damage,' and a claim for the costs of repairing damage caused by the defective work, which is a claim for `property damage.'" See United States Fire Ins. Co., 979 So.2d at 889-90 (citing cases adopting this approach); see also Wm. C. Vick Constr. Co. v. Pennsylvania Nat'l Mut. Cas. Ins. Co., 52 F.Supp.2d 569, 582 (E.D.N.C.1999) ("[T]he property allegedly damaged has to have been undamaged or uninjured at some previous point in time."), aff'd, 213 F.3d 634 (4th Cir.2000) (unpublished table decision); L-J, Inc. v. Bituminous Fire & Marine Ins. Co., 366 S.C. 117, 124, 621 S.E.2d 33, 36 (2005) ("In the present case, the complaint did not allege property damage beyond the improper performance of the task itself."); Travelers Indem. Co. of America, 216 S.W.3d at 311 ("[W]e hold that claims alleging only damages for replacement of a defective component or correction of faulty installation do not allege `property damage.'").

Further, we note it is only after "property damage" has been alleged that the question of "occurrence" is reached. With respect to the components of a project that sustained physical injury, we look to the definition of occurrence, which is ambiguous and must be construed in favor of the insured, and find coverage was triggered.

Returning to Newman and viewing those facts through the lens of both "property damage" and "occurrence," we clarify that the costs to replace the negligently constructed stucco did not constitute "property damage" under the terms of the policy. The stucco was not "injured." However, the damage to the remainder of the project caused by water penetration due to the negligently installed stucco did constitute "property damage." Based on those allegations of property damage and construing the ambiguous occurrence definition in favor of the insured, the insuring language of the policy in Newman was triggered by the property damage caused by repeated water intrusion.6

In sum, we clarify that negligent or defective construction resulting in damage to otherwise non-defective components may constitute "property damage," but the defective construction would not. We find the expanded definition of "occurrence" is ambiguous and must be construed in favor of the insured, and the facts of the instant case trigger the insuring language of Harleysville's policies. We note, however, that various exclusions may preclude coverage in some instances. Because the parties in the present case stipulated not to raise the issue, we do not address any policy exclusions and exceptions.

IV.

Harleysville next challenges the trial court's decision to impose "joint and several"7 liability. We adopt the "time on risk" framework, and therefore overrule our decision in Century Indemnity Co. v. Golden Hills Builders, Inc., 348 S.C. 559, 561 S.E.2d 355 (2002). In our view, the "time on risk" approach best conforms to the terms of a standard CGL policy and to the parties' objectively reasonable expectations. In particular, the "time on risk" approach requires a policyholder to bear a pro rata portion of the loss corresponding to any portion of the progressive damage period during which the policyholder was not insured or purchased insufficient insurance. The parties entered the following stipulations relevant to this issue:

The trial court found the allocation issue was controlled by Century Indemnity Co. v. Golden Hills Builders, Inc., 348 S.C. 559, 561 S.E.2d 355 (2002). The trial court determined that Century Indemnity mandated a "joint and several" approach and, therefore, ordered Harleysville to indemnify the full $7.2 million in stipulated damages. We reverse and remand for further proceedings.

An analysis of the proper method for allocating a loss among successive insurers must begin with the threshold question of what must happen in order to trigger the potential for coverage under a particular policy. See Montrose Chemical Corp. of California v. Admiral Ins. Co., 10 Cal.4th 645, 42 Cal.Rptr.2d 324, 913 P.2d 878, 880 n. 2 (1995) (en banc) (explaining that "trigger of coverage" is a description of "what must take place within the policy's effective dates" in order for there to be a potential of coverage (emphasis omitted)). Only after determining how policies are triggered will it be possible to decide which policies were triggered by a progressive injury8 and, correspondingly, how much of the loss caused by the injury is covered by each. The threshold issue of the trigger of coverage was resolved in South Carolina by the case of Joe Harden Builders, Inc. v. Aetna Casualty & Surety Co., 326 S.C. 231, 486 S.E.2d 89 (1997).

Following Joe Harden, we were confronted in Century Indemnity with a question regarding the scope of a triggered insurer's obligation to its insured. In Century Indemnity, only one policy was at issue, and we were asked to determine whether it would cover "only the amount of property damage that occurred during the policy period" or "all sums [the policyholder] becomes legally obligated to pay[,] if property damage occurs during the policy period." 348 S.C. at 564, 561 S.E.2d at 357 (emphasis in original). These two alternatives are a classic statement of the difference between a "time on risk" and a "joint and several" approach to allocating losses in progressive property damage cases. We determined that the policy must cover "damage that occurred during the policy period and . . . any continuing damage." Id., 561 S.E.2d at 358 (emphasis in original). This result was consistent with the "joint and several" approach to allocation because it made one policy responsible for the entire loss caused by a progressive injury.

The question addressed in Century Indemnity was one which has caused much consternation among the courts of this country. Because Century Indemnity gave only a conclusory analysis to this complex issue and because it relied on an incorrect interpretation of Joe Harden in doing so, we overrule Century Indemnity and confront the allocation issue anew. After a detailed review of the policy language and the relevant case law, we find that each triggered insurer must indemnify only for the portion of the loss attributable to property damage that occurred during its policy period. Accordingly, we adopt a "time on risk" approach to the allocation of damages caused by progressive injuries.

A. Harleysville's CGL Policies

We begin with the basic principle that insurance policies are contracts to be interpreted in accord with contract law. Accordingly, our discussion begins with the language of the policies themselves. Harleysville's CGL policies provided, in relevant part, that they would cover "those sums that the insured becomes legally obligated to pay as damages because of `bodily injury' or `property damage' to which this insurance applies." The insurance applied "only if: (1) The `bodily injury' or `property damage' [was] caused by an `occurrence'. . .; and (2) The `bodily injury' or `property damage' occur[red] during the policy period."9

This policy language, or language that is substantially the same, is typical of a standard CGL policy. Accordingly, the interpretation of Harleysville's policies would be controlled by Joe Harden and Century Indemnity. See Century Indemnity Co., 348 S.C. at 561, 561 S.E.2d at 356 ("We accepted the following question[] certified by the United States Fourth Circuit Court of Appeals: 1. Does a standard commercial general liability insurance policy . . . provide coverage for continuing damage that begins during the policy period?"); Joe Harden Builders, Inc., 326 S.C. at 232, 486 S.E.2d at 89 ("This case is before us on certification from the United States District Court to interpret the language of a standard occurrence insurance policy."). We turn now to an analysis of those cases.

B. Joe Harden

In Joe Harden, we were asked to answer the following certified question: "Where defective construction causes progressive property damage that is otherwise covered by insurance, is the property damage deemed to occur: 1. When the concrete frame is constructed out of plumb; 2. When the masonry contractor knowingly builds the defective brick wall; 3. When the failure of the brick wall is manifest; 4. When the owner actually discovers the failure of the brick wall; or 5. At some other time?" 326 S.C. at 232-33, 486 S.E.2d at 89-90. Clearly, this question focused on what must happen in order to trigger coverage under a particular policy. We surveyed four common theories.

First, we looked to the theory that coverage is triggered at the time of the "injury-causing event." We explained that, "[u]nder this theory, coverage is triggered at the time of the underlying injury-causing event, even though no damage has yet occurred, and the policy in effect at the time of this underlying event covers all the ensuing damage." Id. at 234, 486 S.E.2d at 90. The final phrase of this explanation— "covers all the ensuing damage"—is the logical result of the theory itself. If coverage results from the injury-causing event, and there is only one event, responsibility for full coverage would rest with a single policy. Coverage responsibility could only be spread among multiple policies if the court considered the "underlying event" to have been ongoing or repetitive, such that it spanned multiple policy periods.

The next theory we considered was that coverage is triggered "when [the] injury manifested." We said, "[u]nder this theory, damage is deemed to occur at the time it is manifested or discovered, thus triggering coverage for all the ensuing damage under the policy in effect at the time of manifestation, even if some damage actually occurred earlier but was undetected." Id. Like the first theory, the logical result here would be to place full responsibility on a single policy, unless the manifestation of an injury could somehow be construed as ongoing or repetitive.

The third theory—which we ultimately adopted—was different from the first two. We explained:

Id. at 235, 486 S.E.2d at 91 (emphasis added). What makes this theory unique versus the first two is that it is logically consistent with indemnity under multiple policies: it "allow[s] coverage under any policy in effect during th[e] entire" progression of the damage. Notably missing from our explanation, then, was the mention of any one policy covering "all ensuing damage."

Though we adopted this continuous trigger theory, thereby allowing coverage under multiple policies, we expressed disapproval regarding the aspect of the theory that allowed coverage under policies that preceded the first actual injury:

Id. at 236, 486 S.E.2d at 91 (emphasis added). For that reason, we adopted what we called a "modified" continuous trigger. The modification was that, rather than defining the damage period as beginning with the injury-causing event and ending with manifestation, we defined the damage period as the term during which actual injuries occurred.

Before we made this modification, however, we described the fourth general theory regarding the trigger of coverage—that coverage is triggered "at the time of an injury-in-fact"—as follows: "Under this theory, coverage is triggered whenever the damage can be shown in fact to have first occurred, even if it is before the damage became apparent, and the policy in effect at the time of the injury-in-fact covers all the ensuing damages." Id. (emphasis added). Like the first two theories we discussed and rejected, this theory is logically consistent with indemnity only by a single policy: an injury can only have "first occurred" on a single occasion. See Montrose Chemical Corp., 42 Cal.Rptr.2d 324, 913 P.2d at 895 n. 17 (noting an insurer's argument that "once an injury-in-fact is established, even retrospectively, all potential coverage is cut off . . . and only the insurer on the risk at the time the injury-in-fact first `occurs' is liable to indemnify the insured").

In order to make an injury-in-fact trigger consistent with coverage under multiple policies, we had to recognize the ongoing or repetitive nature of the injuries in a progressive damage case. We did precisely that by (a) adopting the continuous trigger theory (theory three), and (b) modifying it to require an injury-in-fact during each policy period. Thus, we stated:

326 S.C. at 236, 486 S.E.2d at 91 (emphasis added). The Industrial Steel Container Co. case we cited in support explains that where a court considers there to have been ongoing injuries-in-fact, there is the potential for coverage under more than one policy. 399 N.W.2d at 159 ("We view this `actual injury' rule to be sufficiently broad to recognize that in cases involving long exposure to a toxic substance there can be damage with more than one manifestation and more than one insurance policy can afford coverage. We reject the argument that there can be only one occurrence in a case where property damage results from continuous or repeated conditions of exposure.").

Our holding in Joe Harden was this: "We hold coverage is triggered at the time of an injury-in-fact and continuously thereafter to allow coverage under all policies in effect from the time of injury-in-fact during the progressive damage." 326 S.C. at 236, 486 S.E.2d at 91. This statement was a complete answer to the certified question before the Court. Unlike theories one, two, and four—which contemplated that a single policy would cover "all ensuing damage"—theory three included no statement as to the scope of coverage that would be provided by each triggered policy. Thus, the holding in Joe Harden did not answer the allocation question with which we are now presented.

Yet the Joe Harden Court then went one step further, making a statement in dictum that has been the source of much confusion. We said: "[T]his theory of coverage will allow the allocation of risk among insurers when more than one insurance policy is in effect during the progressive damage." Id. at 237, 486 S.E.2d at 91. This statement could be taken as a mere summation of the fact that theory three allows coverage under multiple policies. Nevertheless, it may be mistakenly construed to mean that (1) risk can only be allocated among insurers, and no portion of the risk can be borne by policyholders who allow insurance to lapse during some portion of the damage period; and (2) where there is only one insurance policy in effect during the progressive damage, that policy must bear the full risk. We attribute no such weighty meaning to this dictum.

We turn now to Century Indemnity, in which we were squarely presented with a question as to how much coverage each triggered insurer must provide.

C. Century Indemnity

In Century Indemnity, we accepted the following certified question:

348 S.C. at 561, 561 S.E.2d at 356. In Century Indemnity, the policyholder purchased coverage from December 7, 1989, to December 7, 1990. The parties stipulated that moisture began to cause property damage prior to the end of the policy period, and that the damage was "continuous since that time." Id. at 562, 561 S.E.2d at 356. The CGL policy at issue was identical in all relevant respects to the policies in this case. Id. at 563, 561 S.E.2d at 357. We reasoned:

Id. at 564, 561 S.E.2d at 357-58 (emphasis in original).

This analysis fundamentally misinterpreted Joe Harden and is profoundly at odds with the insurance contract. The Joe Harden language relied upon in Century Indemnity was taken from the description of theory four, treating theory four as if it was the holding in Joe Harden. The actual holding in Joe Harden, adopting theory three and then modifying it using the injury-in-fact trigger from theory four, was that "coverage is triggered at the time of an injury-in-fact and continuously thereafter to allow coverage under all policies in effect from the time of injury-in-fact during the progressive damage." 326 S.C. at 236, 486 S.E.2d at 91 (emphasis added). The fundamental difference is this: while the theory quoted in Century Indemnity would make a single policy responsible for full indemnity, the modified continuous trigger theory adopted in Joe Harden makes multiple policies responsible.

The error in Century Indemnity is further evident in its statement that "property damage relates back in time to the time of the occurrence, that is, when the first injury occurred to the property." 348 S.C. at 563, 561 S.E.2d at 357 (emphasis added). Again, a "first injury" can only occur once. Thus, Century Indemnity would collapse all property damage into a single policy period. This would leave no logical basis for finding that later policies were also triggered. Joe Harden, on the other hand, expressly contemplated that property damage would span multiple policy periods, triggering coverage under each policy.

In sum, the actual holding in Joe Harden gave no guidance as to how much coverage would be provided by each triggered policy. Because Century Indemnity relied on a single trigger theory, rather than on the modified continuous trigger theory adopted in Joe Harden, the Century Indemnity opinion gave short shrift to this complex issue. For this reason, we overrule Century Indemnity and confront the issue anew. We turn now to an analysis of the two theories regarding the scope of coverage advocated by the parties in this case.

D. "Joint and Several" vs. Pro Rata/"Time on Risk"

In this case, Crossmann argues in favor of a "joint and several" approach to the allocation of damages, while Harleysville advocates a pro rata/"time on risk" approach. We adopt the pro rata/"time on risk" approach.

Courts adhering to the "joint and several" theory require each triggered insurer to indemnify its policyholder for the entire loss caused by the progressive injury, up to the policy limit, even if the majority of the loss occurred after the policy period expired. A key feature of this approach is that a policyholder may be indemnified in full despite its failure to purchase CGL coverage throughout the entire progressive damage period. Further, under this theory, where multiple insurance policies are triggered, the policyholder often is permitted to choose the policy from which it will seek indemnity. The chosen insurer may then seek partial reimbursement from any other insurers triggered by the progressive injury.10 Because "joint and several" jurisdictions typically allow a selected insurer to bring a separate lawsuit seeking such reimbursement, many have criticized the theory as inefficient and wasteful of judicial resources. E.g., Boston Gas Co., 910 N.E.2d at 311.

The seminal case advocating a "joint and several" approach is Keene Corporation v. Insurance Company of North America, 667 F.2d 1034 (D.C.Cir.1981). Keene rested primarily on the view that a policyholder's purchase of insurance entitled it to certainty that it had limited its liability to the cost of the policy itself. Id. at 1047-48. Advocates of a "joint and several" approach typically contend that the plain language of the insuring agreement requires an insurer to pay "all sums" or "those sums" the insured becomes legally obligated to pay, as long as some property damage occurs during the policy period. See, e.g., Goodyear Tire & Rubber Co. v. Aetna Cas. & Sur. Co., 95 Ohio St.3d 512, 769 N.E.2d 835, 841 (2002).11 As we detail below, we believe this interpretation ignores critical language limiting the insurer's obligation to pay to sums that are attributable to property damage that occurred during the policy period.

In contrast to the "joint and several" approach, courts adhering to a "pro rata" theory of the scope of each triggered insurer's obligation to its insured hold each insurer responsible only for some pro rata portion of the loss caused by a progressive injury, regardless of whether there are other triggered insurers available to cover the remainder. Thus, unlike jurisdictions using a "joint and several" theory, courts subscribing to a pro rata theory generally require the policyholder to bear the portion of the loss attributable to the policyholder's assumption of the risk. Boston Gas Co., 910 N.E.2d at 303 ("One important feature of a pro rata allocation is that courts adopting this type of allocation generally require the policyholder to participate in the allocation . . . for those periods of no insurance, self-insurance, or insufficient insurance." (alteration in original) (quoting J.M. Seaman & J.R. Schulze, Allocation of Losses in Complex Insurance Coverage Claims § 4:3[c], at 4-21 (2d ed.2008))). As a result, for those periods of "no insurance, self-insurance, or insufficient insurance" during the progressive damage period, a portion of the loss will be borne by the policyholder.

Pro rata theorists have developed several different methods for calculating each insurer's pro rata portion of the loss, each supported by its own notions of fairness and the nature of CGL insurance. "Time on risk" is one such method, and in our opinion, is the method most consistent with the language of a standard CGL policy.12

Our approach relies heavily on the opinion of the Supreme Judicial Court of Massachusetts in Boston Gas Company v. Century Indemnity Company, which provides the reading of the relevant policy language that we believe is correct:

910 N.E.2d at 305 (emphasis in original); id. at 306-07 (adopting this interpretation of the policy language). This interpretation is consistent with various other courts around the country, including, among others, the Second and Sixth Circuit Courts of Appeal and the courts of Minnesota, New York, and Vermont. Id. at 307 n. 34 (citing cases).

While Boston Gas dealt with an excess insurance policy, rather than a standard CGL policy, the key language is the same. As in Boston Gas, Harleysville agreed to pay for damages incurred "because of `bodily injury' or `property damage' to which this insurance applies." The insurance applies "only if . . . [t]he `bodily injury' or `property damage' occur[red] during the policy period." In other words, the insurance does not apply to property damage that did not occur during the policy period. Though the insurer in Boston Gas agreed to pay the "ultimate net loss" above a certain threshold amount while a standard CGL policy promises payment for "those sums that the insured becomes legally obligated to pay as damages," both phrases are modified by the requirement that covered damages must be "because of . . . `property damage' to which th[e] insurance applies." This requirement limits the promise of payment, obligating the insurer to pay only those damages caused by property damage that "occurs during the policy period."

Not only does the Boston Gas interpretation give effect to each part of the insuring agreement (rather than focusing solely on the terms "all sums" or "those sums"), it is consistent with the objectively reasonable expectations of the contracting parties. As the Boston Gas Court explained:

910 N.E.2d at 309-10 (quoting Public Service Co. of Colorado v. Wallis & Cos., 986 P.2d 924, 940 (Colo.1999) (en banc)).

Further, this interpretation forwards important policy goals. Specifically, it preserves the incentive for businesses to purchase sufficient insurance, which in turn promotes stability in the marketplace. See id. at 311 ("[T]he pro rata allocation method . . . engenders stability and predictability in the insurance market, provides incentive for responsible commercial behavior, and produces an equitable result."). By contrast, the "joint and several" theory "creates a false equivalence between an insured who has purchased insurance coverage continuously for many years and an insured who has purchased only one year of insurance coverage." Id. (quoting Public Serv. Co. of Colo., 986 P.2d at 939-940).

For these reasons, we reverse the trial court's order allocating the entire $7.2 million in stipulated damages to Harleysville and hold that the proper method for allocating damages in a progressive property damage case is to assign each triggered insurer a pro rata portion of the loss based on that insurer's time on the risk.13 We remand for further proceedings consistent with a "time on risk" approach. To aid the trial court in applying the "time on risk" framework, we provide guidance in the section below.

E. Application of the "Time on Risk" Approach

An ideal application of the "time on risk" approach would require the finder of fact to determine precisely how much of the injury-in-fact occurred during each policy period and precisely what quantum of the damage award in the underlying suit was attributable to that injury. Unfortunately, it is often "both scientifically and administratively impossible" to make such determinations. Boston Gas Co., 910 N.E.2d at 301 (quoting Michael G. Doherty, Comment, Allocating Progressive Injury Liability Among Successive Insurance Policies, 64 U. Chi. L.Rev. 257, 257-58 (1997)). This unique difficulty has been described as follows:

Id. at 300 n. 20 (quoting William R. Hickman & Mary R. DeYoung, Allocation of Environmental Cleanup Liability Between Successive Insurers, 17 N. Ky. L.Rev. 291, 292 (1990)).

In cases where it is impossible to know the exact measure of damages attributable to the injury that triggered each policy, courts have looked to the total loss incurred as a result of all of the property damage and then devised a formula to divide that loss in a manner that reasonably approximates the loss attributable to each policy period. The basic formula consists of a numerator representing the number of years an insurer provided coverage and a denominator representing the total number of years during which the damage progressed.14 This fraction is multiplied by the total amount the policyholder has become liable to pay as damages for the entire progressive injury. In this way, each triggered insurer is responsible for a share of the total loss that is proportionate to its time on the risk.15

This formula is not a perfect estimate of the loss attributable to each insurer's time on the risk. Rather, it is a default rule that assumes the damage occurred in equal portions during each year that it progressed. If proof is available showing that the damage progressed in some different way, then the allocation of losses would need to conform to that proof. However, absent such proof, assuming an even progression is a logical default.

In this case, a strict application of the basic "time on risk" formula might be inappropriate. There were numerous buildings involved in the underlying lawsuit against Crossmann, each with its own certificate of occupancy, and the parties have stipulated that the damage began "within 30 days after the Certificate of Occupancy was issued for each building." Further, the parties stipulated that the damage "progressed until repaired or until Beazer Homes paid to settle the underlying cases, whichever came first." Accordingly, it may be that, as to each building, each policy was "on the risk" for a slightly different proportion of the total damage period. We leave it to the sound discretion of the trial court to determine whether it is necessary to apply the "time on risk" formula separately to each individual building or whether, instead, it would be prudent to modify the default formula to arrive at a reasonable methodology for this case. Thus, we emphasize that trial courts employing the "time on risk" approach may alter the default formula set forth above where a strict application would be unduly burdensome or otherwise inappropriate under the circumstances of a particular case. However, any such alterations must remain within the bounds of a pro rata/"time on risk" approach: the formula must result in a reasonable approximation of the amount of property damage that occurred during each insurer's policy period.16

In sum, we construe the standard CGL policy to require that each insurer cover only that portion of a loss attributable to property damage that occurred during its policy period. In light of the difficulty in proving the exact amount of damage incurred during each policy period, we adopt the formula above as the default method for allocating shares of the loss. Trial courts may vary from this default formula where appropriate to the circumstances of a particular case, but they must remain faithful to the premise that each insurer is responsible only for a pro rata portion of the total loss, and each pro rata portion must be defined by the insurer's time on the risk.

CONCLUSION

We affirm the trial court's finding that coverage was triggered by an "occurrence." We overrule Century Indemnity and impose a "time on risk" approach to defining the scope of each CGL insurer's obligation to its insured in a progressive damage case. This equitable approach best harmonizes with policy language limiting coverage to the "policy period." Moreover, the "time on risk" framework lends itself to a logical default formula that is easily applied when the actual quantum of damage incurred during each policy period is not known. We remand to the trial court for further proceedings consistent with this opinion.17

AFFIRMED IN PART, REVERSED IN PART AND REMANDED.

TOAL, C.J., PLEICONES, BEATTY and HEARN, JJ., concur.