concurring in part and dissenting in part:
Texas One has raised a fact question as to whether the damage to its buildings constituted one or nineteen loss occurrences under the Fireman’s Fund excess *279coverage policy. I disagree with the majority’s reliance on caselaw involving liability policies and with the result it reaches by employing this approach. Accordingly, I respectfully dissent from the portion of Judge Jones’s excellent and careful opinion that deals with Fireman’s Fund.
I.
A.
Texas law takes a slightly different approach for determining the number of “occurrences” under policies designed to protect the insured from liability to others (“liability policies”) from the approach it takes for determining the number of “loss occurrences” under policies designed to protect the insured from damage or loss to property owned by the insured (“loss policies”). For loss policies, Texas courts apply a “cause” test, which determines the number of loss occurrences based on the number of events that caused the loss or losses at issue. Goose Creek Consol. Indep. Sch. Dist. v. Continental Cas. Co., 658 S.W.2d 338, 340-41 (Tex.App.—Houston [1st Dist.] 1983, no writ). For liability policies, the number of occurrences is determined by finding the number of “events or incidents for which [the insured] is liable.” Maurice Pincoffs Co. v. St. Paul Fire & Marine Ins. Co., 447 F.2d 204, 206 (5th Cir.1971). This approach has been described as the “liability-triggering event” test.1
The approach to be employed is determined by the type of injury at issue. The injury insured against under a loss policy is direct physical damage to or loss of property, and the number of occurrences is determined not by the number of losses or types of damage, but by the number of events that caused loss and/or damage. Liability policies operate only when the insured has a civil liability to one or more nonparties. The injury is the liability, and therefore courts look for the number of events that cause liability.
The “liability-triggering event” test may be viewed as a specialized application of the “cause” test. The court still looks to the cause of the injury, but the injury insured against is liability, not property damage or theft. Whenever an insured party is liable to other parties for civil damages, there often will be contemporaneous and related injuries to the insured and to other parties. The test crafted by Pincojfs focuses the inquiry on the relevant injury and prevents the court from being distracted by these other injuries.
Consideration of the facts of Pincoffs elucidates this distinction. Pincoffs, the insured, bought a large amount of contaminated birdseed and sold it to eight dealers, Pincoffs, 447 F.2d at 205, who then sold it to numerous bird owners, killing many pets. The aggrieved bird owners made claims against their respective dealers, not against Pincoffs directly. The dealers then made claims against Pincoffs, which in turn sought coverage under its liability policy. Id. We concluded that there had *280been eight occurrences, because there were eight sales that caused Pincoffs’s liability for the type of injury protected under the policy.
Because liability was the injury insured against, “[i]f Pineoffs had destroyed the seed before sale, for instance, there would be no occurrence at all for which the insured would be liable.” Id. at 206. But under a loss policy covering loss or spoilage of inventory, the destruction of the seed before sale would not have prevented the occurrence of an insurable event. The loss occurrence would have been the event that contaminated the birdseed, or, if it was already contaminated before Pineoffs received it, the occurrence would have been Pincoffs’s purchase of the seed. Any subsequent sales of the birdseed would have been irrelevant to determining the number of loss occurrences.
For the above reasons, cases that determine the number of occurrences under liability policies have limited applicability when determining the number of loss occurrences under loss policies. Characteristics unique to liability policies determine what losses are relevant. Even under otherwise identical facts, insurance policies, depending on their purposes and definitions employed, may focus on different losses or interpret the underlying events differently.2
B.
Twice recently, this court inadvertently has brought confusion to this area by citing caselaw employing both the liability policy and loss policy approaches, without noting the type of policy at issue. See H.E. Butt, 150 F.3d at 530 (interpreting a liability policy); Ran-Nan Inc. v. Gen. Accident Ins. Co. of Am., 252 F.3d 738, 740 (5th Cir.2001) (per curiam) (interpreting a policy covering losses caused by employee dishonesty). Neither case, however, mandates the approach taken by the majority.
In H.E. Butt, 150 F.3d at 528, we considered the number of occurrences under a liability policy for abuse of children by an employee of insured’s grocery store. Though the H.E. Butt majority cited cases employing both approaches, including Pincoffs and Goose Creek, id. at 530, it appears to have applied a “cause” test. It determined that there had been two occurrences, because the “immediate cause[s]” of the two injuries were separate incidents committed on two different days. Id. at 531, 535. The concurring opinion also found two occurrences, but did so by focusing on the number of “liability-triggering events,” finding two occurrences because the claims against H.E. Butt arose from two separate acts of abuse.3
Both opinions focused on the same events, but for different reasons: the majority because those two events caused the *281injuries to the child, the concurrence because those events triggered the civil liability of the insured. Because the third member of the panel joined in the judgment only, H.E. Butt, 150 F.3d at 535, neither opinion’s reasoning carries prece-dential weight.4
Raur-Nan considered the number of occurrences under an employee dishonesty policy where two employees had separately stolen more than $30,000 each. See 252 F.3d at 738, 740. As the majority notes, in Ran-Nan we quoted the H.E. Butt majority’s statement that the court should look to the events “ ‘that cause the injuries and give rise to the insured’s liability^]’ ” Id. at 740 (quoting H.E. Butt, 150 F.3d at 530). We went on, however, to cite Goose Creek and apply its “cause” test, concluding that there were two occurrences because the losses were caused by two different employees in separate and independent acts of employee dishonesty.5 We made no attempt to determine what events “gave rise to the insured’s liability.” Such an approach has never been utilized when interpreting a loss policy, by this court or any Texas court.
II.
A.
Though the majority is correct that Ran-Nan quoted from and endorsed the “liability-triggering event” test from H.E. Butt and Pincoffs, this test has no applicability to the facts at issue here. H.E. Butt and Pincoffs held that the court should look to the events that gave rise to the liability of the insured to other parties. The insured here, Texas One, has no liability to anyone. The only liability at issue is the insurer’s to the insured. As applied to this case, then, this test points to no events at all.6
Perhaps influenced by the desire to blend the “liability-triggering event” test with the relevant law as established by Goose Creek, the majority suggests that “Goose Creek instructs us ... to focus on the event that gave rise to Fireman’s Fund’s liability under the policy.” This looks similar to H.E. Butt’s admonition to look to the events that “give rise to the insured’s liability,” H.E. Butt, 150 F.3d 526, 530, the difference being, of course, that Fireman’s Fund is the insurer defendant, not the insured plaintiff.
The majority’s characterization of Goose Creek creates a test with no useful meaning. The loss occurrence definition is designed to help the parties determine the event or events that give rise to the insurer’s liability. The majority holds that the number of events that cause the insurer’s liability determines the number of loss occurrences, creating a circular definition.
B.
Despite the instruction of Ran-Nan and Goose Creek to find the “cause” of the loss, *282the majority avoids directly tackling this requirement. It comes closest when it states that “a different leak was responsible for the damage to each building, and as such ... each leak constitutes a separate occurrence[J” The loss was caused by the leaks, and, the majority’s statement implies, the leaks caused the leaks.
The leaks may have happened as a result of the passage of time and normal wear and tear, eighteen natural occurrences that were not the result of any defect. Or, as the plaintiff contends, they may have resulted from faulty workmanship or materials. This dispute is not a question of law. We must assume, for summary judgment purposes, that the leaks were caused by a single event: the installation of faulty plumbing. And we have no undisputed evidence of any intervening cause.7 It is appropriate, in these circumstances, that a jury resolve the fact question of which event or events caused the damage.
The majority draws analogy to Goose Creek, where the court found that two fires could not be considered one event.8 The court started with the proposition that the setting of the two fires caused the loss. The setting of the two fires could not be linked as one event, because the two events were “distinguishable in space and time,” and “one did not cause the other.” Goose Creek, 658 S.W.2d at 341.
We must assume that all of the leaks do share a single cause. It is not relevant that the leaks did not affect buildings other than the ones above them.9 Though “one [leak] did not cause the other,” Texas One contends they were all caused by the pluming installation, an event not “distinguishable in space and time occurred.” Unlike the two fires in Goose Creek, the plumbing installation at the nineteen buildings was “part of a process of continuum.” Id. at 340. The majority notes that the installation was set apart in time from the manifestation of the damage. Goose Creek, however, considered the impact of a delay between two causal events “distinguishable in space and time” from each other, not a delay between a casual event and the manifestations of damage.
Although it is instinctively appealing to assume that the lengthy time lag between the plumbing installation and the loss implies a break in the casual chain, nothing in the policy’s definition of loss occurrence justifies making this assumption as a matter of law. The Fireman’s Fund definition *283of “loss occurrence” contains no requirement that losses caused by a single event must manifest within a certain time frame compared to each other.
Fireman’s Fund has not even argued that the policy will not cover the loss if it arose out of an event that occurred many years earlier.10 Instead, it suggests that the court should pretend that the event is not relevant, because it took place too long ago. Without some basis in the language of the policy, I see no reason to do so. Taking Texas One’s version of the disputed facts to be true, the damage to the nineteen buildings is rightly viewed as one loss occurrence.11
Summary judgment was improper as to this issue. Because it is still disputed whether the leaks were caused by defective workmanship or materials, I would reverse and remand for a factual determination. Accordingly, I respectfully concur in part and dissent in part.
. H.E. Butt Grocery Co. v. Nat’l Union Fire Ins. Co., 150 F.3d 526, 535 (5th Cir.1998) (Benavides, J., concurring). Texas state courts have adopted the Pincoffs approach in liability policy cases. See, e.g., State Farm Lloyds, Inc. v. Williams, 960 S.W.2d 781, 784-85 (Tex.App.—Dallas 1997, writ dism'd by agr.) (relying on Pincoffs in finding that three random gunshots fired by one person constituted three occurrences because they caused three liability-triggering injuries); but see Foust v. Ranger Ins. Co., 975 S.W.2d 329, 334 & n. 3 (Tex.App.—San Antonio 1998, writ denied) (finding that multiple passes by a crop duster together formed one occurrence and distinguishing Williams because the Williams policy did not define occurrence, whereas the definition of occurrence found in the policy at issue included “repeated exposure to [the same general] conditions”).
. Definitions of "occurrence” under liability policies tend to be similar to each other and dissimilar to the definitions of "loss occurrence” under loss policies. These variances in the meanings of key terms may advise different outcomes to otherwise similar cases. See Foust, 975 S.W.2d at 334; see also Goose Creek, 658 S.W.2d at 340 ("[B]oth parties have cited numerous cases touching upon ‘accident’ or 'occurrence,' but many of the cases do not define these terms, and no case has been cited or found which attempted to define 'single event.' We conclude that all cases cited are distinguishable from the instant case and that no useful purpose would be served by discussing the holdings of these cases.”).
. H.E. Butt, 150 F.3d at 535 (Benavides, J., concurring). Though I agree with most of Judge Benavides’s reasoning, I disagree with his description of Goose Creek as applying a "liability-triggering event” test. Id. at 535-36 (citing Goose Creek, 658 S.W.2d at 339). Goose Creek concerned a loss policy, 658 S.W.2d at 339, and never refers to the insured's liability.
. The majority argued that the two approaches were not inconsistent. See H.E. Butt, 150 F.3d at 530 n. 2. The two approaches yielded the same result in that case, but would not in all cases. Though the proper test under liability policies is not before us, I do believe the concurrence is the better statement of the law.
. Ran-Nan was explicit in its endorsement of the “cause” test.
The few Texas cases that have addressed this issue apply a "cause” analysis in determining whether a set of facts involves one or several ofccurrences. This “cause” approach to analyzing the number of "occurrences” is utilized by the great majority of courts and jurisdictions nationwide.
Ran-Nan, 252 F.3d at 740 (citations omitted).
.Similarly, the plaintiffs in Ran-Nan, victims of employee theft, had no liability to anyone, which explains why this court, despite citing H.E. Butt and Pincoffs, made no attempt to apply the "liability-triggering event” test to the facts of that case.
. Some courts have suggested that an intervening cause might change the number of occurrences. See Home Indem. Co. v. City of Mobile, 749 F.2d 659, 662 (11th Cir.1984) (finding under Alabama law that "if one cause is interrupted and replaced by another intervening cause, the chain of causation is broken and more than one occurrence has taken place”); Appalachian Ins. Co. v. Liberty Mat. Ins. Co., 676 F.2d 56, 61 (3d Cir.1982) (stating that under "cause” test employed in a majority of jurisdictions, "the court asks if there was but one proximate, uninterrupted, and continuing cause which resulted in all of the injuries and damage”) (internal quotation marks and other punctuation omitted).
. Goose Creek, 658 S.W.2d at 341. As the majority appears to concede, Goose Creek is the most appropriate case to consider because it involved a loss policy with a definition of loss occurrence that is very similar to the one at issue here. 658 S.W.2d at 340 (defining a loss occurrence as "the total loss by perils insured against arising out of a single event”).
.These buildings join to form a single apartment complex, built together with, the plaintiff would contend, a single defective plumbing system. The buildings are insured together as a single property, and are not far apart as in Goose Creek. If this case involved one house, and that faulty plumbing caused damaging leaks in the kitchen and the bathrooms, would the majority find multiple loss occurrences if the leaks in the kitchen did not cause any damage to the bathrooms?
. There is a distinction here between when the losses manifested and when the event that gave rise to those losses occurred. It is a different matter whether the policy covers losses that actually manifest before a certain date. See Am. Home Assurance Co. v. Uni-tramp, Ltd., 146 F.3d 311, 314 (5th Cir.1998) (holding that an injury becomes manifest when it is apparent or capable of easy perception). Here, there is no dispute that the losses all manifested during the policy period.
Consider a hybrid between the facts of Goose Creek and this case, where an arsonist sets fire to the woods near the Texas One property, eventually causing eighteen buildings on the property to burn down. The court would have little difficulty determining that the fires were all caused by one event, meaning there was one loss occurrence. The court would be unconcerned with whether one building caught fire immediately and was consumed in minutes while another building caught fire hours later and then slowly burned for two days. It also would not matter whether the fire spread directly from the trees to all the buildings independently or whether it had spread from one building to another in a chain.
. In Unity/Waterford-Fair Oaks, Ltd. v. Fireman’s Fund Ins. Co., 2001 WL 34019525 (W.D. Tex. June 20, 2001) 2001 U.S. Dist. LEXIS 24818, the insured proved nearly identical facts to those alleged by Texas One here. The court concluded that "the damages arising out of the failure of the underground plumbing system constitute a single loss occurrence within the meaning of the Fireman's Fund policy." Id. at *6. The case was appealed, but not as to this holding. See Gen. Accident Ins. Co. v. Unity/Waterford-Fair Oaks Ltd., 288 F.3d 651 (5th Cir.2002).