A barge owned by National Marine, Inc., cast adrift when it slipped its moorings at a dock operated by Transport Distributors, Inc. (TDI), in the Chicago Sanitary and Ship Canal, collided with another dock and two boats, causing damages that have been stipulated in an amount slightly in excess of $100,000. The owners of the damaged property brought suit in admiralty against National Marine, which impleaded TDI, which in turn impleaded the subsidiary of National Marine that had operated the tugboat (Lemont Harbor and Fleeting Services)--but we can ignore that detail and treat parent and subsidiary as one. After a bench trial, the judge ruled that both National Marine and TDI had been negligent and assessed the former's fault at twice the latter's, meaning that National Marine would have to pay two-thirds of the damages. National Marine appeals, contending that TDI was solely at fault. TDI crossappeals, arguing that National Marine was solely at fault.

The plaintiffs--oddly, it may seem--take no position on the merits of either the appeal or the cross-appeal, even though a conceivable outcome would be a ruling that neither defendant had been negligent. The defendants appear to agree, however, that at least one of them was negligent; and the plaintiffs must be satisfied that either one is good for the entire damages (which are relatively modest), so that if National Marine succeeds in shifting the entire burden of liability to TDI, or vice versa, the plaintiffs will be none the worse for it. It is true that TDI was not named as a defendant in the plaintiffs' complaint, or for that matter in any other pleading by the plaintiffs. But in an admiralty suit, once a defendant impleads a third party in an effort to shift the burden of liability in whole or part from its own shoulders, and demands judgment in favor of the original plaintiff against that third party, the suit proceeds as if the original plaintiff had sued the third party. Fed.R.Civ.P. 14(c), and Note of the Advisory Committee on the 1966 Amendment thereto; Home Ins. Co. v. Puerto Rico Maritime Shipping Authority, 524 F. Supp. 541, 546 (D.P.R.1981). So the plaintiffs in our case can recover their damages from TDI if the latter was negligent and its negligence a cause of the accident, and of course from National Marine on their original complaint if National Marine was negligent and its negligence a cause of the accident.

Although in form a tort case, in economic reality this is a contract case. It involves the respective duties to third parties (the plaintiffs) of two firms engaged in a voluntary undertaking to "park" the barge of one at the dock of the other. National Marine, which owns a fleet of barges, made a contract with a lumber company that is not a party to this case to transport a load of that company's lumber to TDI's dock. TDI is in the business of transloading bulk cargoes. It receives them by rail or barge, unloads them and stores them in its warehouse, and then reloads them onto trucks for further travel. Although at the time of the accident its place of business bordered on the canal and had a seawall at which barges could be moored, TDI received almost all cargo by rail. Only about once a year did it receive a shipment by barge. Because of its infrequent use of the dock, TDI did not have its own crew of maritime workers, and in fact had no maritime capabilities at all. When a barge arrived or was due, TDI would hire a maritime crane, complete with crew, to unload it.

National Marine notified TDI to expect a barge on April 22, 1988. A tugboat owned and operated by National Marine brought the barge to the dock that evening. The crew of the tug lashed the barge to the seawall with inch-and-a-half-thick hawsers made of braided plastic rope. It is unclear whether two, three, or four hawsers were used, but probably there were three because a photograph of the barge taken shortly after the accident showed three hawsers dangling from it. No one was at the dock when the tugboat's crew moored the barge. It was understood that when the barge had been unloaded, TDI would notify National Marine and the latter would send a tug to take the barge away.

Because of TDI's delay in obtaining a maritime crane and crew, the barge had not yet been unloaded when, in the early morning hours of April 27, almost five days after the barge had arrived, it broke from its moorings and began its fateful journey. During the period while the barge was moored at the dock no one inspected the moorings to make sure they were secure.

The cause of the barge's breaking free is unclear, although several hypotheses can be rejected. Apparently the lines broke, rather than being untied or unraveling. The one line that was recovered and is in the record is frayed at the point where it broke, rather than being cut through cleanly. So the hypothesis of intervention by mischievous boys or other vandals can pretty well be excluded, and anyway is not argued. Weather conditions were normal, though there was some wind blowing away from the dock; and there appear to have been no extraordinary perturbations, such as a huge wake stirred up by a passing ship, that might have imparted such force to the barge as to cause a normal line to break even in normal weather. Three hypotheses as to why the lines broke cannot be said to be excluded by the record, although they vary in their plausibility. One is that there were only two lines, and two is not enough--the force exerted by the barge even in a mild wind would cause them to snap. But there was expert testimony that two lines are enough, and anyway there appear to have been at least three. Another hypothesis, also pressed by TDI, is that the line was composed of a type of plastic that if used too long eventually decays from exposure to sunlight, and that the flecks of powder that come off on one's hand when one touches the rope in the record are the telltale clue that the rope had indeed been used too long and had decayed. Still another theory, this one pressed by National Marine, is that the breaking of the lines was caused by chafing, that is, by their rubbing against the seawall. But against this there was evidence that the seawall and the barge were at the same height, rather than the barge's being lower, which would be necessary to generate friction between the seawall and the ropes.

Unfortunately the district judge's findings of fact do not resolve the disputed questions bearing on the cause of the accident. The entirety of the relevant discussion is as follows:

National Marine was negligent because its employees failed to moor Barge NL264 safely. Ropes provided and tied to the dock by National Marine parted, thus causing the subsequent crash between Barge NL264 and the property of each of the three plaintiffs.

Third-party defendant, TDI, was negligent because its employees failed to inspect the barge moored at its dock. Had employees of TDI inspected the moored barge on a regular basis, it is possible that they would have detected unsafe mooring, such as chafing of the ropes, too few ropes or improperly tied ropes, and that the breakaway would have been prevented.

Notice that the judge is noncommittal as to whether the accident occurred because of improper mooring, which would be National Marine's responsibility, or because of chafing, which would seem to be TDI's responsibility, not only because chafing would be detectable by even a superficial inspection but also because the effect of chafing is gradual, so presumably would not have caused the rope to part had it not been for TDI's delay in unloading the barge. Therefore if chafing was indeed the cause of the accident it is hard to see why two-thirds of the liability should come to rest on the barge's owner. Moreover, the first paragraph that we quoted suggests that the judge may have thought that the fact that the ropes broke demonstrated without need for any more evidence that the barge had been negligently moored--a conclusion that would equate negligence to strict liability.

Although the findings are deficient, both National Marine and TDI believe that we can resolve the case without a remand for better findings. Each believes it demonstrable from the legal authorities and the trial record that the other is solely responsible for the accident. National Marine, claiming that the facts demonstrate that TDI was the bailee of the barge, argues that when a barge breaks away from its moorings the bailee is presumed to be negligent and that TDI has not rebutted the presumption. TDI claims that it has rebutted the presumption, that anyway it was not a bailee, and that the relevant presumption is that the owner of a vessel is presumed negligent if the vessel drifts out of control.

The language of presumptions and bailments obscures rather than illuminates the ultimate question, which is that of the relative fault of the parties and cannot be answered without a remand for additional findings and perhaps more evidence. The accident could in principle have been prevented by either the owner of the barge or the operator of the dock. The critical issue, in the absence of an explicit contract between the parties, is their relative costs of prevention. To frame that issue it is helpful to distinguish between preventive measures that could have been taken before the tugboat, having moored the barge, departed, and preventive measures that could have been taken afterward. Before, the accident could have been prevented by someone's making sure that the barge was fastened by enough sound ropes soundly tied; and it is the custom in the inland maritime industry that the owner of the barge is responsible for taking these measures rather than the operator of the dock. The latter would have to maintain a crew on hand whenever a barge was expected and would in any event know less than the owner did about the weight of the barge and therefore about the amount and strength of rope necessary to hold it, though he would probably know more than the owner about local conditions, such as changes in water level or the possibility of waves churned up by passing vessels, that might strain the moorings.

After mooring, the sort of accident that happened here can be prevented, or at least the probability of its occurring can be greatly reduced, by regular inspection of the ropes to make sure that they are not chafing, or otherwise fraying, or loosening, or coming untied. Most docks that receive barges are operated by firms that employ people who know something about the mooring of barges, and it is cheaper for those employees to inspect the barges moored at the docks than it is for the owner of the barges to send an inspector from his place of business or employ an inspector at every seawall or pier at which they might dock. TDI, however, employed no one competent to make such an inspection, and it asks us to infer that National Marine was the lower-cost accident avoider after the mooring as well as before. There are many objections to this argument. One is that in the absence of a contract to the contrary National Marine was entitled to rely on the custom of the industry, which is that the operator of the dock inspects the moorings during the period in which a barge is docked in order to make sure they are secure. National Marine had no reason to suppose that TDI was not in compliance with industry custom. If because TDI did not have an inspection capability National Marine would in fact have been the lower-cost after-mooring accident avoider, TDI should have notified National Marine, which might (perhaps for a price) have taken extra measures to moor the barge securely, knowing that it would not be inspected. This is the contract principle of Hadley v. Baxendale, 9 Ex. 341, 156 Eng.Rep. 145 (1854), applied to torts in such cases as EVRA Corp. v. Swiss Bank Corp., 673 F.2d 951 (7th Cir.1982), and Rardin v. T & D Machine Handling, Inc., 890 F.2d 24 (7th Cir.1989). Without a requirement of notice the law would be encouraging a wasteful duplication of safety measures.

Notice to National Marine was not TDI's only recourse if it lacked the human, informational, or other resources necessary to do the necessary inspecting itself. Since the need for inspection increases with the length of time the barge is docked, TDI could have made greater efforts to obtain a maritime crane and crew promptly. What in any event is clear is that the fact that a potential injurer fails to equip himself with any means of avoiding injury cannot exonerate him from liability for negligence. Vaughan v. Menlove, 3 Bing. (N.C.) 468, 132 Eng.Rep. 490 (Com.Pl.1837). "The law considers ... what would be blameworthy in the average man, the man of ordinary intelligence and prudence, and determines liability by that. If we fall below the level in those gifts, it is our misfortune; so much as that we must have at our peril." O.W. Holmes, Jr., The Common Law 108 (1881). If you hand over your car to an attendant at a parking lot, and he drives it into a wall while trying to park it, the lot cannot defend on the ground that it had hired an attendant who did not know how to drive and therefore was incapable of preventing the accident. On a par is TDI's argument that the fact that after the barge was discovered missing from its moorings National Marine rather than TDI sent a tug to rescue it shows that TDI was not a bailee. All it shows is that TDI had no capability for rescuing a runaway barge. TDI gets no points for the limitations of its competence.

However, the fact that TDI may not have come up to the customary standard of care in its industry, or, even without regard to that point, may have been able to avoid the accident at a cost less than the expected accident cost (simply by notifying National Marine that it had no intention of ever inspecting the barge's moorings), does not establish its sole liability, as National Marine appears to believe. Under a traditional common law regime, where there is no contribution among joint tortfeasors, National Marine could not even have impleaded TDI, but would have had to bear the sole liability because the plaintiffs had decided to sue it alone and ignore TDI, provided of course that the plaintiffs were able to prove both that National Marine was negligent and that the accident would not have occurred had it not been so. The rule in admiralty, however, is that joint tortfeasors are entitled to allocate a plaintiff's damages among themselves in accordance with their relative fault. United States v. Reliable Transfer Co., 421 U.S. 397, 95 S. Ct. 1708, 44 L. Ed. 2d 251 (1975). That rule allowed National Marine to implead TDI. But the fact that TDI was at fault, even grievously so, would not entitle National Marine to shift the entire burden of liability to TDI's shoulders unless, of course, National Marine was not negligent at all. Even if it was not negligent for failing to send inspectors to check the moorings after the barge was delivered to TDI's docks, it may have been negligent in the original mooring of the barge.

Nor can National Marine throw the entire burden of liability onto TDI merely by establishing that the latter was the bailee of the barge at the time it broke away. In maritime law as in common law, a bailee is responsible for exercising due care in the keeping of the good that has been entrusted to him. Riverway Co. v. Trumbull River Services, Inc., 674 F.2d 1146, 1150 (7th Cir.1982); Solano v. Beilby, 761 F.2d 1369, 1373 (9th Cir.1985); Global Tank Trailer Sales v. Textilana-Nease, Inc., 209 Kan. 314, 317, 496 P.2d 1292, 1295 (1972). We may assume for purposes of this discussion that if TDI failed to inspect the barge's moorings it was in violation of a bailee's duty of care. For the fact that National Marine did not pay TDI for the right to park its barge at the latter's dock would not make TDI a gratuitous (that is, unpaid) bailee, who does not owe his bailor a duty of ordinary care, but only one of slight care. Coggs v. Bernard, 2 Ld.Raym. 909, 92 Eng.Rep. 107 (1703); United States v. Lyons, 706 F.2d 321, 334 n. 21 (D.C.Cir.1983); American Enka Co. v. Wicaco Machine Corp., 686 F.2d 1050, 1053 (3d Cir.1982). Even though it was not paid directly by National Marine, TDI was offering transloading services for compensation, not to accommodate friends. It was thus a bailment for mutual benefit, imposing upon the bailee a duty of exercising ordinary care. Id. at 1053-54; Global Tank Trailer Sales v. Textilana-Nease, Inc., supra, 209 Kan. at 316, 496 P.2d at 1294-95; Hartmann v. Black & Decker Mfg. Co., 16 Conn.App. 1, 7, 547 A.2d 38, 42 (1988). But if, therefore, by failing to inspect the lines, TDI violated its bailee's duty of care to National Marine, this would only establish TDI's fault; it would not establish National Marine's freedom from fault.

And if TDI was not a bailee, nevertheless this would not, as TDI believes, exonerate it from responsibility. By operating a dock and offering to take delivery of a cargo from a barge that would be moored at the dock, TDI held itself out to the world of inland waterway shipping, a world that includes National Marine, as complying with industry custom, which is that the operator of the dock must exercise due care to prevent a vessel moored at it from breaking away. The duty of due care that resulted from this holding out did not depend on the existence of a bailment. It arose from the implied representation of compliance with an industry custom of due care. Cf. Trevino v. Union Pacific R.R., 916 F.2d 1230, 1237 (7th Cir.1990); Erie R. Co. v. Stewart, 40 F.2d 855 (6th Cir.1930). That custom became an implied term of the implied contract between National Marine and TDI. One could if one wanted say that if TDI by virtue of this implied contract was responsible for the safety of the vessel this made it a bailee. But that would merely be to relabel a conclusion as a criterion. The issue of bailment is a red herring in this case--and for another reason as well. TDI's duty to National Marine to shoulder, if itself negligent, some part of the liability to the plaintiffs does not arise from any bailment, but from the admiralty rule that joint tortfeasors are entitled to an apportionment of liability in accordance with their relative fault. If TDI's negligence was responsible in part for the damage to the plaintiffs' property, then TDI was a tortfeasor, whether or not it was a bailee, and National Marine would be entitled to some contribution from TDI to help defray their joint liability to the plaintiffs.

It is true that many cases not only presume the owner of a drifting vessel to be at fault if it collides with an anchored or otherwise stationary vessel or a structure on the shore, but also describe this as a presumption of law in a strong sense, one that shifts the burden of persuasion as well as of production. E.g., The Louisiana, 70 U.S. (3 Wall.) 164, 173, 18 L. Ed. 85 (1866); American Petrofina Pipeline Co. v. M/V Shoko Maru, 837 F.2d 1324 (5th Cir.1988); Weyerhaeuser Co. v. Atropos Island, 777 F.2d 1344, 1348 (9th Cir.1985); contra, Pennsylvania R.R. v. S.S. Marie Leonhardt, 320 F.2d 262, 264 (3d Cir.1963); see generally Thomas J. Schoenbaum, Admiralty and Maritime Law § 13-2 at pp. 454-55 (1987). Such a presumption is not easy to square with the rule that a plaintiff has the burden of proving the essential elements of his case, one of which is the defendant's negligence. It would be more consistent with the general policy of the law to regard this so-called presumption as simply a common-sense inference, akin to the inferences drawn routinely in nonmaritime accident cases in accordance with the principle of res ipsa loquitur (the thing speaks for itself). Neace v. Laimans, 951 F.2d 139, 141-42 (7th Cir.1991). That principle is applied in admiralty law as well, Lone Star Industries, Inc. v. Mays Towing Co., 927 F.2d 1453, 1457 (8th Cir.1991), and its relation to the drifting-vessel presumption has been noted. Schoenbaum, supra, § 13-2, at p. 455; cf. American Petrofina Pipeline Co. v. M/V Shoko Maru, supra, 837 F.2d at 1326 ("the presumption derives from the common-sense observation that moving vessels do not usually collide with stationary objects unless the vessel is manhandled in some way"). Because American law is too complex and should where possible be simplified, we are sympathetic to the suggestion that " 'negligence' at sea does not differ, in principle, from 'negligence' ashore," Grant Gilmore & Charles Black, Jr., The Law of Admiralty § 7-11 at p. 509 (1975), implying that special presumptions applicable only to proof of negligence in admiralty cases should be shunned. What is true is that if a vessel drifting out of control hits a stationary object, and the vessel's owner presents no evidence, the inference of his negligence arising from the circumstances of the accident will be as a practical matter irrebuttable. But there is no need to embellish this sensible observation with the name of presumption, a name that could be thought--that in this context has been thought--to require a shifting of the burden of persuasion.

At all events, even if there is a drifting-vessel presumption in some strong sense that differentiates it from the inference-drawing that lawyers call by the name of res ipsa loquitur, it has no application to this appeal. The issue is not the allocation of responsibility for the accident between the owner of the drifting vessel and the owners of the stationary objects that were damaged by it. The issue is the allocation of responsibility between the owner and the third party that let the vessel slip its mooring and drift into collision with the plaintiffs' property. The drifting-vessel presumption is not designed for the allocation of liability between the injurers, as distinct from the allocation of the loss between them and their victims, and although occasionally mentioned in the former context as well it does not control decision there, as well shown by Pasco Marketing, Inc. v. Taylor Towing Service, Inc., 554 F.2d 808 (8th Cir.1977), and Lancaster v. Ohio River Co., 446 F. Supp. 199, 202 and n. 1 (N.D.Ill.1978). For while as between drifting vessel and stationary object struck by it common sense suggests that the former is more likely to have been at fault in the collision than the latter, there is no similar "presumption" when the issue is whether the drifting vessel itself or the dock that, as it were, let it drift was at fault in the subsequent collision of the vessel with a stationary object.

National Marine has its own candidate for a presumption to govern in this case: if a vessel that has been docked for a substantial period of time breaks loose from its moorings and an accident results, the operator of the dock shall have the burden of proving that the breaking loose was not the result of his negligence. Id. at 202, and cases cited there. Again it is better to look at the matter in terms of inference. It is indeed natural to infer in such a case that the operator of the dock was at fault in having failed to inspect the moorings from time to time. So in the absence of other evidence the trier of fact will have a solid, perhaps as a practical matter an unshakable, basis for concluding that, as between the owner of the vessel and the operator of the dock, the sole cause of the accident was the latter's negligence. (Which further shows why the drifting-vessel presumption should not be applied in a case where the issue is the allocation of fault between the injurers, not between injurers and victim.) But in this case there is other evidence. Evidentiary presumptions, or, as we prefer to call them, inferences from experience, are designed to fill a factual vacuum. Once evidence is presented, here concerning the action and inaction of the vessel's owner and the dock's operator, presumptions become superfluous because the parties have introduced evidence to dispel the mysteries that gave rise to the presumptions. Pennsylvania R.R. v. S.S. Marie Leonhardt, supra, 320 F.2d at 264.

But even if there were no evidence, the method of decision by presumptions could not work in this case, where each party is armed with a presumption--TDI with the presumption that the owner of a drifting vessel is at fault, National Marine with the presumption that (a substantial period of time having elapsed between the mooring and the drifting) the operator of the dock is at fault. When presumptions clash, they disappear, and if there is no other evidence the case must be resolved in accordance with the simplest of default rules, that which exonerates a defendant, such as TDI, whether a defendant to a first-party or a third-party complaint, if the claim against it is not established by a preponderance of the evidence. Utica Mutual Ins. Co. v. Fireman's Fund Ins. Cos., 748 F.2d 118, 120 n. 1 (2d Cir.1984); Schulte & Bruns v. Great Lakes Stevedoring Corp., 300 F.2d 897 (6th Cir.1962).

There is evidence here, but the evidence does not establish with the clarity that would permit us to make the findings that the district judge failed to make which party was at fault, and how much at fault. Consider the evidence of National Marine's negligence. The evidence that the number of lines was insufficient is feeble. TDI's alternative theory is that one or more of the lines was defective, as shown by the one in the record. The evidence was in conflict on the point. TDI asks us to resolve the conflict by taking in our hands the line that was recovered and made a part of the record and noting the fine film of dust that it leaves on them. This request misconceives the relation between an appellate court and a trial court in determining the facts of a case. Appellate courts review findings of fact made by trial courts; with immaterial exceptions they do not make their own findings unless the evidence is completely one-sided. They rarely handle evidence, physically, to determine its properties, though there is no difference in principle between feeling and seeing an exhibit in order to determine a contested issue. When the author of this opinion, responding despite his doubts to counsel's invitation, handled the broken hawser, his hands indeed became dirty. But he could not tell whether this was because of dirt on the surface of the rope or because of, as it were, an internal dust constituted of the disintegrated remains of the part of the rope that gave way.

TDI's fault is equally uncertain. The district judge said that if TDI had inspected the moorings "on a regular basis" it might have detected National Marine's negligence. But this would depend on how regular is regular. If it means every five days, inspection would have missed the problem entirely if the inspectors had made their rounds in the middle of the day on April 22, before the barge arrived, and had not returned till the same time of day on the twenty-seventh, by which time the barge had broken its moorings and drifted away. The judge made no finding on the industry's custom regarding the frequency of inspections. Nor on the feasibility of TDI's having obtained a maritime crane and crew earlier, knowing that it was not going to inspect the barge.

Once the judge has made findings that identify the fault (if any) of each party, which he did not do in this case, he must determine the parties' relative fault, which he also did not do. We have suggested in previous cases that relative fault is in inverse ratio to the costs of accident avoidance to the respective parties. Wassell v. Adams, 865 F.2d 849, 854 (7th Cir.1989); Davis v. United States, 716 F.2d 418, 429 (7th Cir.1983). That is, the party that could have avoided the accident more cheaply, and is therefore more culpable because due care would have been less of a burden to him, pays proportionately higher damages. They were not admiralty cases, but there is no reason to treat admiralty cases differently in this respect; and they involved the issue of comparative negligence (the injurer's negligence relative to the victim's) rather than that of relative fault (one injurer's negligence relative to another's), but again we cannot see why that should make a difference. Suppose, purely hypothetically, that National Marine could have averted the accident by changing its mooring lines more frequently, and the added expense of doing so would have been $1,000. And that TDI could have gotten a maritime crane and crew to unload the barge on April 23, before it broke away, at an additional cost of $2,000. Then National Marine would be two-thirds responsible for the damage to the plaintiffs' property and TDI one-third--as the judge concluded, but without any subsidiary findings that might have justified such a conclusion. Or suppose that TDI could have averted the accident simply by notifying National Marine of its incompetence to inspect. That might have led National Marine, knowing that there would be no back-up safety precautions by the dock operator, to add another mooring line, at a trivial cost, that would have prevented the accident. Suppose that adequate notification could have been given by TDI at the cost of a 20-cent phone call. That might furnish a compelling argument for placing the entire burden of liability on TDI. But the judge made no finding concerning the feasibility of notification or, what is more problematic, National Marine's likely reaction, if any, to such notification.

We have mentioned industry custom several times and must in closing try to determine the weight that the district judge should give it on remand in making the necessary determinations concerning the parties' fault. One of the best-known principles of tort law--a principle that received its canonical expression in an admiralty decision written by Learned Hand, T.J. Hooper, 60 F.2d 737, 740 (2d Cir.1932)--is that compliance with custom is no defense to a tort claim. The principle (whose application in admiralty is further illustrated by Tug Ocean Prince, Inc. v. United States, 584 F.2d 1151, 1156-57 (2d Cir.1978)) is obviously sound when one is speaking of the duty of care to persons with whom the industry whose customary standard of care is at issue has no actual or potential contractual relation. For in that situation the costs of the injury can be made costs to the industry, and thus influence its behavior, only through the imposition of tort liability. R.H. Coase, The Problem of Social Cost, 3 J.Law & Econ. 1 (1960). It is different when the potential victims are the customers of the potential injurers. For then the latter, even if they are not subject to any tort liability, will have to ponder the possibility that if they endanger their customers they may lose them or may have to charge a lower price in order to compensate them for bearing a risk of injury. In such a case the market itself fixes a standard of care that reflects the preferences of potential victims as well as of potential injurers and then the principal function of tort law, it could be argued, is to protect customers' reasonable expectations that the firms with which they deal are complying with the standard of care customary in the industry, that is, the standard fixed by the market. United States Fidelity & Guaranty Co. v. Jadranska Slobodna Plovidba, 683 F.2d 1022, 1029 (7th Cir.1982). This consideration is made relevant here by National Marine's argument that by departing without notice from the industry custom regarding inspections by dock operators TDI increased the risk of an accident.

This case illustrates the operation of custom in both the contractual and the noncontractual setting. The owners of the dock and boats that were damaged had no actual or potential contractual relationship with the defendants. The latter in deciding what precautions to take would not be influenced, therefore, by the possible effect on transactions they might have with those owners; and as a result, the standard of care in the inland trade with regard to preventing runaway barges might be too low because it ignored some of the accident costs to which such runaways give rise. But at the present stage of this lawsuit the focus is not upon the defendants' duties toward strangers (the plaintiffs); it is upon their duties toward each other; and they had a contractual relationship with each other. It was not a direct or explicit relationship. There was no written contract, no writing at all, no money changing hands, no receipts--not even the maritime equivalent of a parking-lot claim check. But a barge owner and a dock owner (or operator) are knitted together by their contracts with their customers as tightly as they would be by a contract between themselves. Competition among barge owners and among dock owners to provide their respective legs of a unitary transportation service at the lowest possible price, coupled with tort liability to third parties, will give both types of service provider market incentives to adopt optimal safety precautions. Both face potential liability to third parties such as the plaintiffs in this case. They minimize their liability costs by allocating the responsibility for safety measures between them efficiently. They can do this explicitly, cf. Ferentchak v. Village of Frankfort, 105 Ill. 2d 474, 480-81, 86 Ill. Dec. 443, 447, 475 N.E.2d 822, 826 (1985), or, implicitly, by abiding by the custom that the market has evolved.

Here the custom was for the barge owner (or operator) to moor the barge to the dock with a sufficient number of sound ropes, carefully fastened, and for the dock owner (or operator) to inspect the barge from time to time while it is at the dock, to make sure that the mooring lines remain securely fastened. We do not know whether National Marine violated the duty of care that custom places upon it because we do not know whether the ropes were unsound, insufficient in number (which seems highly unlikely, as we have said), or improperly fastened (also unlikely). And we do not know whether TDI violated the duty that custom imposed upon it to inspect (or in lieu therefore to notify National Marine that it was not inspecting, or to expedite the unloading), because there is no finding about what precisely the duty consists of. Since, however, these customs appear to reflect an undistorted market determination of the best way to minimize runaway-barge accidents, we think the focus of the district court's inquiry should be on the parties' respective compliance with and departures from the customs and that the judge and the parties should not feel compelled to conduct a cost-benefit analysis of barge transportation from the ground up. Cf. United States Fidelity & Guaranty Co. v. Jadranska Slobodna Plovidba, supra, 683 F.2d at 1029.

The judgment of the district court is vacated, and the case remanded to that court for further proceedings consistent with this opinion. Should the district judge think it necessary to take additional evidence he is free to do so. Circuit Rule 36 shall not apply.

VACATED AND REMANDED, WITH INSTRUCTIONS.

HARLINGTON WOOD, Jr., Senior Circuit Judge, concurring.

Judge Posner's opinion lucidly examines the facts, customs, and law and comes to a conclusion with which I fully agree.

My only hesitation is with the issue as framed. It is stated, referring to the relative fault of the parties, "[T]he critical issue, in the absence of an explicit contract between the parties, is their relative costs of prevention." Ante, at 883. We are reminded, ante, at 888, that in two prior decisions of this court, Wassell v. Adams, 865 F.2d 849, 854 (7th Cir.1989), and Davis v. United States, 716 F.2d 418, 429 (7th Cir.1983), both also written by Judge Posner, a similar issue arose. In those cases it was also suggested "that relative fault is in inverse ratio to the costs of accident avoidance to the respective parties." Ante, at 888.

That may well be a more efficient approach, but unfortunately, not being the expert in economics as are my two colleagues, I would prefer, for now, to approach comparative fault determination on the more traditional basis first broadly enunciated in United States v. Reliable Transfer Co., 421 U.S. 397, 95 S. Ct. 1708, 44 L. Ed. 2d 251 (1975). I note, however, that this present opinion specifically advises the judge and parties, ante, at 889, that they "should not feel compelled to conduct a cost-benefit analysis of barge transportation from the ground up."