Lead Opinion

Opinion for the court filed by Chief Judge MIKVA, except as to Part 11(C).

Opinion for the court as to Part 11(C) filed by Circuit Judge BUCKLEY.

Opinion dissenting in part filed by Chief Judge MIKVA.

MIKVA, Chief Judge:

On September 1, 1983, a Korean Air Lines (“KAL”) Boeing 747 airliner was shot down somewhere over the Sea of Japan by one of the Soviet Union’s SU-15 interceptor aircraft, killing all 269 persons on board. *1473The approximate crash site placed the flight more than 300 nautical miles off course and in Soviet airspace. Before trial, the district court granted summary judgment for the United States on claims that it had breached a duty to warn. See In re Korean Air Lines Disaster of September 1, 1983, 646 F.Supp. 30 (D.D.C.1986). In an appeal brought by the unsuccessful plaintiffs in those cases, we upheld the district court’s decision. See Wyler, et al. v. Korean Air Lines, Inc., et al., 928 F.2d 1167 (D.C.Cir.1991). Additional background on the disaster is provided in our opinion in the companion appeal.

In this case, KAL appeals from a judgment entered against it by the district court after a jury found KAL guilty of willful misconduct and awarded damages to a group of 137 plaintiffs; the award included $50 million in punitive damages. KAL challenges both the willful misconduct verdict and the assessment of punitive damages. Although KAL raises some valid concerns about the quality of the evidence submitted to the jury, we conclude that the finding of willful misconduct was permissible. However, we vacate the punitive damage award.

I. Background

On August 31, 1983, Korean Air Lines Flight 007 (“KE007”) left New York’s Kennedy Airport bound for Seoul, South Korea, with a stop in Anchorage, Alaska. After refueling there, KE007 took off from Anchorage International Airport at 1300 Greenwich Mean Time (“G.M.T.”), or 4:00 a.m. local time. The Anchorage Air Traffic Control (“ATC”) Center instructed KE007 to climb and maintain a flight level of 33,-000 feet and “proceed direct BETHEL when able.” BETHEL, located approximately 360 nautical miles west of Anchorage, is the navigational gateway for Route R20 of the North Pacific Composite Route System, which operates like a multi-lane highway for civilian flights across the Pacific Ocean between North America and Asia. BETHEL also serves as a navigational waypoint at which an airplane can cross-check its position against a radio fix. Route R20 has a series of navigational waypoints with precise geographical coordinates, each about 300 miles apart, that KE007 would follow on a direct path and use for course verification and reporting purposes. These waypoints were designated BETHEL, NABIE, NUKKS, NEEVA, NINNO, NIPPI, NYTIM, NOKKA and NOHO. See Map of the North Pacific (attached as an appendix to this opinion). Position reports from the crew to ATC ground controllers were required at BE-THEL and other waypoints up through NIPPI, some 1800 miles from Anchorage. Anchorage ATC transfers control of aircraft following Route R20 to Tokyo Center at waypoint NIPPI.

At the time that these events took place, Anchorage ATC had radar coverage for less than half the distance to BETHEL. The Federal Aviation Administration’s (“FAA”) Kenai radar installation provided coverage as far as Cairn Mountain, approximately 165 nautical miles west of Anchorage. FAA radar surveillance was terminated at 1327:50 G.M.T. From that point on, Anchorage ATC would rely on the crew’s position reports (calculated in-flight with the help of the waypoints and on-board systems) to track KE007’s location and adherence to R20. At 1349 G.M.T., KE007’s pilot reported reaching BETHEL and estimated that they would pass over their next reporting waypoint, NABIE, at 1430 G.M.T.

After BETHEL, KE007 did not have any direct communications with Anchorage ATC. During the remainder of the flight, the crew reported reaching each successive waypoint up to NIPPI, when Anchorage transferred control to Tokyo. These reports were relayed to Anchorage ATC by another KAL flight, KE015, which had departed Anchorage 14 minutes after KE007 and also was following Route R20. At NABIE, KE007 should have been able to communicate directly with Anchorage ATC through the St. Paul Island radio transmitter, but the crew was unable to do so, having instead to relay the report through KE015. KAL elicited testimony that this was not necessarily unusual and could be caused by weather. Plaintiffs countered *1474that such a problem was rare, and they suggested that the direct broadcast was impossible because KE007 was already 90 miles off course at NABIE and out of the range of St. Paul Island’s radio. This happened again at the next waypoint, NEEVA, where KE007 should have reported by using the Sheyma Island radio station. Evidently the pilot of KE015 became suspicious that something was wrong when KE007 requested that he relay this position report to Anchorage ATC, in part because KE007’s time of arrival at that waypoint was nine minutes late and the explanation KE007’s crew gave for the delayed arrival (strong headwinds) conflicted with his own observations just four minutes behind. Plaintiffs contended at trial that the inability to use the Sheyma Island radio station and the inconsistent headwind reports resulted from the fact that KE007 was now 170 miles off course and inside Soviet airspace.

After control of KE007 was transferred from Anchorage ATC, the crew reported to the controllers in Tokyo that they had reached NIPPI and estimated their time of arrival at NOKKA as 1826 G.M.T. A subsequent accident report concluded that the wind conditions reported by KE007 at NIP-PI did not match those experienced on R20 and were more consistent with a position over 200 miles to the north-northwest over the Kamchatka Peninsula, U.S.S.R. The final direct transmission from the KE007 crew to ground controllers was a report to Tokyo at 1827:10 G.M.T. that the plane was rapidly decompressing and descending. The flight recorders and most of the wreckage were never recovered, so the details of what happened remain a mystery. However, contrary to KAL’s suggestion, there was evidence from which to calculate an approximate crash site, including intermittent acoustic signals from an underwater flight recorder beacon received by search vessels in the area along with debris from the crash found in the Sea of Japan and washed ashore on Hokkaido Island.

Plaintiffs’ liability claims in all of these actions grow from a hypothetical flight deviation that they claim was apparent on FAA radar shortly after take-off and could be extrapolated across the entire route to the approximate crash site in the Sea of Japan (after flying for about three hours in Soviet airspace and crossing over the Kamchatka Peninsula and Sakhalin Island). See Appendix. At trial, plaintiffs attempted to establish willful misconduct by the KE007 flight crew by theorizing that, due to an error in programming the Inertial Navigation System (“INS”) prior to departure from Anchorage, KE007 deviated from its plotted course to Seoul and entered Soviet airspace. The INS is a navigational device which stores preprogrammed flight plans and displays data during the flight showing present position, waypoint positions, and any course deviations from the designated route. The INS units use gyroscopes to calculate positions during flight, and they must be programmed before takeoff by inserting the exact coordinates for latitude and longitude at the particular gate where the aircraft is parked.

At trial, plaintiffs contended that the crew must have known of the misprogram-ming either before leaving Anchorage or shortly thereafter, but decided to proceed rather than turn back and face possible disciplinary action such as suspension. Plaintiffs argue that the crew’s location reports were fabricated to cover up the error in programming. Furthermore, according to the plaintiffs, the crew must have been fully aware of the serious risk of straying into Soviet airspace, given that a similar KAL flight had been intercepted and forced down five years earlier. In 1978, KAL flight 902 had strayed deep into sensitive Soviet airspace near northern Europe, over 1000 miles off course. A Soviet fighter fired on the aircraft when it took evasive maneuvers, forcing an emergency landing on a frozen lake. This incident, which caused several fatalities, was evidently discussed in subsequent KAL training programs.

Plaintiffs’ evidence consisted of radar reports covering the initial leg of KE007’s flight, an investigative report completed by the Secretary General of the International Civil Aviation Organization (“ICAO”) three *1475months after the incident (hereinafter referred to as the “ICAO Report,” although KAL emphasizes that it was never officially adopted by the ICAO Council), expert testimony from two pilots with extensive experience flying across the Pacific, and reports of prior unrelated incidents involving KAL navigation errors. At the close of plaintiffs’ case, KAL moved unsuccessfully for a directed verdict. KAL then introduced the testimony of the air traffic controllers who monitored KE007 on radar and communicated with its crew, the testimony from the Deputy United States representative to ICAO and the report of the Air Navigation Commission (“ANC”) of ICAO criticizing the Secretary General’s Report. The jury returned a liability verdict against KAL, and in a subsequent verdict awarded plaintiffs $50 million in punitive damages. The trial judge entered judgment on the verdict and denied KAL’s motion for judgment notwithstanding the verdict (“JNOV”).

II. Analysis

On appeal, KAL argues (1) that there was insufficient evidence of willful misconduct to support the jury’s verdict; (2) that several pieces of evidence used to prove willful misconduct (the ICAO Report, the expert testimony based on that report, and the prior incidents) should not have been admitted; and (3) that the award of punitive damages was improper. We find merit in only the last of these contentions.

A. Evidence of Willful Misconduct

Initially, KAL contends that the district court’s final jury instruction misstated the standard for willful misconduct. KAL requested a definition of willful misconduct as the “intentional performance of a wrongful act,” but allegedly the court defined it only as the “intentional performance of an act.” KAL has taken the actual instruction out of context:

Willful misconduct is the intentional performance of an act with knowledge that the act will probably result in an injury or damage, or in some manner as to imply reckless disregard of the consequences of its performance.

This is precisely the formulation approved by this court in KLM Royal Dutch Airlines v. fuller, 292 F.2d 775, 778 (D.C.Cir.), cert. denied, 368 U.S. 921, 82 S.Ct. 243, 7 L.Ed.2d 136 (1968).

KAL’s primary argument on appeal is that there was no evidence from which a reasonable jury could find willful misconduct. Both sides mischaracterize what evidence was adduced at trial and for what purposes. In view of what was actually presented to the jury, we conclude that the evidence was not so one-sided that KAL was entitled to a JNOV or new trial. Cf. In re Korean Air Lines Disaster of September 1, 1983, 704 F.Supp. 1135, 1136-51 (D.D.C.1988) (hereinafter “In re KAL Disaster”) (detailing how each piece of plaintiffs’ evidence could fit into the puzzle and what permissible inferences a jury might draw therefrom). The hard evidence of what happened is sparse: Anchorage radar showed that KE007’s initial flight path was generally in the direction of waypoint BE-THEL (though plaintiffs introduced additional radar evidence of a deviation appearing at this early stage in the flight), and wreckage was' discovered in the Sea of Japan shortly after a distress call from the crew. Although the exact crash site was never established, KAL does not deny that KE007 must have strayed hundreds of miles off course to end up in or near the Sea of Japan. The question is at what point did the flight go off course and why. Plaintiffs’ theory, based in part upon the inconclusive ICAO Report, posits that the deviation was caused by an initial INS programming error that caused KE007 to stray progressively further off course during the flight. KAL emphasizes the flimsiness of this explanation but fails to offer one of its own; it contends that KE007 was on course throughout the flight across the Pacific until, for some entirely mysterious reason, it veered dramatically off course after waypoint NIPPI.

KAL argues that the radio transmissions from KE007’s crew at each waypoint verify that they remained on course along Route *1476R20. Plaintiffs respond by pointing to suspicious circumstances such as KE007’s inability to communicate directly with air traffic control in Anchorage (having instead to relay its reports through KE015), and reporting wind conditions that were inconsistent with those encountered by KE015. KAL replies that the communications and reported wind conditions were in no way unusual, adding that these items could not provide evidence of the supposed course deviation. But all that plaintiffs were trying to prove at this point was that the uncorroborated reports from the crew were not persuasive evidence of their being on course. Plaintiffs suggested that the crew had a motive to lie because they knew they would face suspension if they admitted their error and turned back. A jury could legitimately discount the accuracy of those reports, in which case KAL’s version of the probable flight path would be at least as speculative as plaintiffs’ extrapolation. Given this lacuna, and the undisputed fact that the airliner was shot down hundreds of miles off course, a jury could credit plaintiffs’ coherent theory of what happened, derived from circumstantial evidence such as the early radar reports of a mild course deviation before BETHEL and the arguably inconsistent wind reports.

A critical element in plaintiffs’ scenario is evidence that a course deviation was evident as early as waypoint BETHEL that grew progressively larger during the flight. The FAA radar showed that halfway between Anchorage and BETHEL, KE007 was six miles north (to the right) of what would have been a direct heading toward the first waypoint. Military radar data from the uncertified though apparently accurate King Salmon station showed that KE007 was 12 miles north of a direct heading shortly before reaching BETHEL. Plaintiffs’ witnesses interpreted this evidence as indicating a drift off course, extrapolating it to suggest at least a 12 mile deviation by the time KE007 passed BE-THEL (well outside the 1-2 mile margin of error allowed). KAL argues that KE007 could proceed toward BETHEL any way it wanted (the ATC course clearance for R20 only applied once it left the coast after BETHEL), and being six miles off the most direct path at the half-way point and 12 miles off just before BETHEL would not justify an extrapolation of a 12 mile deviation at BETHEL. But plaintiffs introduced testimony that the flight would have reappeared on radar if it had suddenly turned directly toward BETHEL. A jury could reasonably believe plaintiffs’ explanation of a deviation at BETHEL rather than KAL’s evidence to the contrary, see In re KAL Disaster, 704 F.Supp. at 1138, 1142, and then extrapolate that deviation across the entire flight to the approximate crash site, just as juries are normally permitted to draw conclusions from less than complete direct evidence of what happened. This is not simply a case where the district court let the jury choose, on the basis of mere speculation and conjecture, between hypothetical scenarios of what happened. Contrast Siegel v. Mazda Motor Corp., 878 F.2d 435, 437-39 (D.C.Cir.1989) (“When the record thus contains competing, unrebutted hypotheses consistent with driver error [rather than a mechanical defect in the transmission] ... we can discern no basis upon which to say that any one of the possible explanations is ‘more probable’ than the others.”).

Contrary to KAL’s suggestion, plaintiffs did not simply use the flight path postulated by the ICAO Report and based on the Soviet version of the intercept. Even if this aspect of the ICAO Report was arguably unreliable, a jury could make its own evaluation based upon the evidence supplied by plaintiffs and then accept the ICAO conclusions of possible causes drawn from the assumed flight path. Once the jury accepted plaintiffs’ extrapolated path, the question becomes why the crew did nothing to correct the error. Because the flight recorder was never recovered, we only know that the crew claimed throughout the flight that they were on course. As explained above, plaintiffs discount these reports as fabrications. Plaintiffs’ experts testified that, assuming the course deviation posited by plaintiffs, the deviation was probably the result of an error in programming the INS. The ICAO Secre*1477tary General reached the same tentative conclusion in his report. Moreover, plaintiffs’ experts testified that the crew would have known of the deviation or, if they did not, that their ignorance was due to gross negligence in failing to check their instruments. (The jury was also informed that the ICAO Report concluded that the crew probably would not have been aware of the error.) Plaintiffs then introduced evidence of a 1978 Soviet intercept of a KAL flight to prove that this crew would have known of the hazards of straying into Soviet airspace.

Appellees add that even if one believed KAL’s claim that KE007 stayed on course up to waypoint NIPPI, the flight must have then turned suddenly in the direction of Sakhalin Island to reach the crash site. KAL responds that if this were the correct scenario, there would be no basis for a finding of willful misconduct. However, the crew could not have failed to notice such a glaring deviation, and their failure to notify ATC until after they were decompressing suggests willful misconduct. Ap-pellees are not using this as a fallback scenario so much as they are emphasizing KAL’s inability to suggest any innocent explanation for the disaster. Admittedly, plaintiffs’ case at trial did not include any “smoking guns,” but that is because no one knows exactly what happened. There was sufficient evidence here from which to decipher a pattern of conduct giving rise to liability. If the only evidence for the proposed course deviation had been the wreckage in the Sea of Japan, plaintiffs could not have prevailed simply by drawing a straight line from Anchorage and arguing that the crew would have known of this course deviation but chose to cover it up. Here there was additional, albeit not uncon-troverted, evidence that a course deviation appeared almost immediately after takeoff, consistent with a 10 degree error in setting the INS longitude coordinate at Anchorage, combined with somewhat suspicious radio reports from the crew. When “questions [of willful misconduct] depend upon inferences to be drawn from essentially circumstantial evidence ... [o]ne can hardly imagine a clearer case in which such questions should have been left to the jury.” Berner v. British Commonwealth Pacific Airlines, 346 F.2d 532, 538 (2d Cir.1965) (reversing JNOV for plaintiffs in an air crash case), cert. denied, 382 U.S. 983, 86 S.Ct. 559, 15 L.Ed.2d 472 (1966).

B. Evidentiary Rulings

1. The ICAO Report and Plaintiffs’ Expert Testimony

KAL objects to the introduction of the ICAO Report (and the appended Soviet intercept report), contending that it was hearsay not within the public records exception of Rule 803(8), Fed.R.Evid. That rule provides, in pertinent part, that the following are not hearsay:

Records, reports, statements, or data compilations, in any form, of public offices or agencies, setting forth ... factual findings resulting from an investigation made pursuant to authority granted by law, unless the sources, of information or other circumstances indicate lack of trustworthiness.

Fed.R.Evid. 803(8)(C). “Rule 803(8)(C) is to be applied in a commonsense manner, subject to the district court’s sound exercise of discretion_” City of New York v. Pullman Inc., 662 F.2d 910, 914 (2d Cir.1981), cert. denied, 454 U.S. 1164, 102 S.Ct. 1038, 71 L.Ed.2d 320 (1982). KAL argues that there were no “factual findings” because the final report from the Secretary General — criticized for lacking hard evidence from which to draw its conclusions — was never adopted by the ICAO Council. KAL also contends that the sources of the Report’s information concerning the probable flight path, especially the so-called “Russian line” provided in the Soviet intercept report, lacked trustworthiness. Finally, KAL argues that admission of the ICAO Report contravened Rule 403, Fed.R.Evid., because the jury was misled into believing that the conclusions were authoritative.

We disagree with KAL’s characterization. of the Report as non-final. By a special resolution passed a few weeks after the disaster, the ICAO Council directed “the Secretary General to institute an investiga*1478tion ... and to provide an interim report to the Council within 30 days ... and a complete report during the 110th Session of the Council.” After first giving the Council an interim report as required by the resolution, the Secretary General ultimately provided his “final report” to the Council, although he qualified it by conceding that the report was based on incomplete evidence and promising to supplement the Report if any new information came to light. The minutes of the Council session indicate the submission of the Secretary General’s “final report” on December 12, 1983. The Council had the power to endorse or ignore the Report but apparently not revise it. Cf. New York v. Pullman, 662 F.2d at 914 (“As an interim report subject to revision and review, the report did not satisfy the express requirement of the Rule that the proffered evidence must constitute the ‘findings’ of an agency or official.”). After sending the Report to the Air Navigation Commission for review and comment to determine whether any changes should be made in ICAO’s Rules of the Air in light of the incident, the Council chose not to endorse it. Even so, the Secretary General was acting in the capacity of a public official when he conducted “an investigation made pursuant to authority granted by law” (in this case, the special resolution) and issued his final report. Cf. In re Multi-Piece Rims Prods. Liab. Lit., 545 F.Supp. 149, 151 (W.D.Mo.1982) (admitting letters from NHTSA employee acting in an official capacity and describing evaluative tests).

We are somewhat more troubled, however, with the district court’s resolution of the trustworthiness question under Rule 803(8)(C). Although the ICAO Report was based on very limited hard evidence, this alone would not undermine its finality or trustworthiness. See Beech Aircraft v. Rainey, 488 U.S. 153, 168, 109 S.Ct. 439, 449, 102 L.Ed.2d 445 (1988) (upholding the “broad admissibility” of facts, conclusions, evaluations and opinions contained in an aircraft accident report, because “the admission of a report containing ‘conclusions’ is subject to the ultimate safeguard — the opponent’s right to present evidence tending to contradict or diminish the weight of those conclusions”). In this case, the district court let KAL introduce the ANC report criticizing the Secretary General’s report. But KAL’s trustworthiness objection focuses on the Preliminary Soviet Report appended to the ICAO Report and the Secretary General’s use of the uncorroborated “Russian line” in arriving at his conclusions. Among other problems with the Soviet report, KAL claims that it was politically motivated, seeking to exonerate the Soviet Union’s actions. Standing alone, we doubt that the Soviet report would have been admissible under Rule 803(8)(C), but its inclusion as an appendix to the ICAO Report would not automatically render the Secretary General’s otherwise admissible report objectionable. See FAA v. handy, 705 F.2d 624, 633 (2d Cir.) (“As a statement by a foreign government to the federal government, incorporated in the FAA’s factual findings resulting from an investigation made pursuant to authority granted by law, the [German] telex was admissible as a public record and report under Fed.R. Evid. 803(8)(C).”), cert. denied, 464 U.S. 895, 104 S.Ct. 243, 78 L.Ed.2d 232 (1983).

The more difficult question is whether the alleged untrustworthiness of the Soviet report renders the ICAO Report inadmissible on those same grounds, at least those portions based in part on the Russian intercept line. Rule 803(8)(C) “assumes admissibility in the first instance but with ample provision for escape if sufficient negative factors are present.” Fed.R. Evid. 803 advisory committee note. “The burden is on the party disputing admissibility to prove the factual finding to be untrustworthy.” United States v. American Tel. & Tel. Co., 498 F.Supp. 353, 364 (D.D.C.1980); accord In re Aircrash in Bali, Indonesia, 871 F.2d 812, 816 (9th Cir.), cert. denied, — U.S. -, 110 S.Ct. 277, 107 L.Ed.2d 258 (1989). In Rainey, the Supreme Court observed that “a trial judge has the discretion, and indeed the obligation, to exclude an entire report or portions thereof ... that she determines to be untrustworthy.” 488 U.S. at 167, 109 S.Ct. at 448. . While the Soviet report may have *1479been particularly untrustworthy in its rendition of the actual intercept (indeed, in a recent interview with Izvestia, the Soviet pilot responsible for the shootdown disputed the sanitized official version, see Wash. Post, Jan. 25, 1991, at A16), the radar track itself was generally believed by several sources, including the United States government. At a news briefing on the day of the incident, the Secretary of State told reporters that KE007 “strayed into Soviet airspace over the Kamchatka Peninsula” and was tracked by the Soviets “for some 2lk hours.”

The district court clearly understood its duty to make a threshold trustworthiness finding and recognized its power to exclude any portions of the Report it felt were untrustworthy. KAL’s counsel effectively conceded that the factual section of the ICAO Report (contained in the first 35 pages) was admissible. The district court initially decided not to admit one section of the ICAO Report discussing the Russian intercept line, but then let even that part in when KAL’s counsel retracted his initial objection to the section standing alone. The district court decided that KAL’s trustworthiness objections were more properly addressed to the jury for purposes of evaluating the weight to be accorded the Secretary General’s conclusions: “[Y]ou might convince the jury that it is not worth the paper it is written on, but I am not going to throw the whole report out just because they might believe that in this particular case.” Not surprisingly, the transcript of the district court’s decision from the bench lacks the clarity of a written opinion. In hindsight, we would rather the court had made explicit preliminary findings, preferably in limine, as to the trustworthiness of each challenged portion of the ICAO Report. But mindful that the burden was and is on KAL, we are not convinced that the court failed to carry through on its duties under Rules 104(a) and 803(8) or abused its discretion when it admitted the ICAO Report in its entirety.

Furthermore, as explained previously, plaintiffs did not rely on the Secretary General’s report (or the Soviet intercept report) to establish the probable flight path. Instead, once the jury accepted plaintiffs’ extrapolation, the ICAO Report helped fill the causation gap by evaluating the likelihood that an INS programming error was at fault. Indeed, even the ANC Report that KAL introduced to criticize the ICAO Report confirmed that there was a “significant deviation from track” that could not be explained. The danger, of course, is that the jury was also presented with the ICAO Report’s working assumption based on the Russian line, and this may well have influenced their willingness to believe plaintiffs’ extrapolation. Even so, this is precisely the sort of judgment call by the trial judge that is entitled to deference on appeal and will not be disturbed absent clear error. See United States v. Payne, 805 F.2d 1062, 1066-67 (D.C.Cir.1986) (per curiam). The district court did not abuse its discretion under Rules 803(8)(C) and 403.

Finally, KAL criticizes the admission of testimony by plaintiffs’ experts, again because of the supposed lack of any basis, other than the ICAO and Soviet reports, for the assumed deviation from which all their opinions sprang. Appellees respond that both experts were qualified pilots with significant trans-Pacific experience, and that the ICAO Report was a legitimate source for them to rely on. Because we have decided that the ICAO Report was properly admitted, the expert testimony is also unobjectionable.

2. Evidence of Prior Incidents

Plaintiffs introduced evidence of pri- or KAL incidents to suggest that the crew knew they risked suspension if they returned to Anchorage for reprogramming, and that the crew was aware of the hazards of straying into Soviet airspace. KAL objects that evidence of these prior incidents was irrelevant and unduly prejudicial. Rule 404(b), Fed.R.Evid., allows the admission of evidence concerning “prior bad acts” in limited circumstances, to prove “motive, opportunity, intent, preparation, plan, knowledge, identity, or absence of mistake or accident.” It is of course true that a prior incident must be sufficiently *1480similar to be probative of intent or motive in relation to the incident at issue. See, e.g., Edwards v. Consolidated Rail Corp., 567 F.Supp. 1087, 1105-06 (D.D.C.1983), aff'd mem., 733 F.2d 966 (D.C.Cir.), cert. denied, 469 U.S. 883, 105 S.Ct. 252, 83 L.Ed.2d 189 (1984).

In this case, plaintiffs sought to demonstrate through circumstantial evidence what the crew of KE007 knew or should have known about the consequences of their conduct and also to rebut KAL’s suggestion that misprogramming was impossible or correctable en route. These are permissible ends that do not require the exacting degree of similarity that would be necessary when using prior bad acts to prove specific intent or motive. See Exum v. General Elec. Co., 819 F.2d 1158, 1162-63 (D.C.Cir.1987) (“How substantial the similarity must be is in part a function of the proponent’s theory of proof,” and a theory of notice “imposes a less rigorous requirement of similarity.”).

KAL asserts that it never claimed mis-programming was impossible or that crews did not know about the hazards of intruding Soviet airspace. Neither statement is entirely correct. While the evidence of these prior incidents was no doubt somewhat prejudicial, plaintiffs appear to have introduced it for permissible purposes. See, e.g., United States v. Payne, 805 F.2d at 1066-67. Knowledge (not specific intent or motive) was a relevant issue, notwithstanding KAL’s mild concession in opening argument that the crew would have known of the dangers of flying into Soviet airspace. If KAL wanted to remove the issue from trial so as to exclude evidence of the 1978 shootdown, a pre-trial stipulation might have done the trick. In fact, before trial KAL was arguing that the intercept was an unforeseeable, superseding cause of the disaster, a claim the court felt had to be left to the jury. See In re KAL Disaster, 704 F.Supp. at 1149 & n. 37. We conclude that the evidence of prior incidents was properly admitted at trial.

C. Punitive Damages

BUCKLEY, Circuit Judge:

These consolidated wrongful death actions are governed by the terms of the Warsaw Convention, a multilateral treaty to which the United States has adhered since 1934. See Convention for the Unification of Certain Rules Relating to International Transportation by Air, Oct. 12, 1929, Article 1(1), 49 Stat. 3000, T.S. No. 876 (1934), reprinted in 49 U.S.C.App. § 1502 note (1988). Relying on Floyd v. Eastern Airlines, 872 F.2d 1462, 1483-89 (11th Cir.1989) (“Floyd /”), rev’d on other grounds, — U.S. -, 111 S.Ct. 1489, 113 L.Ed.2d 569 (1991), KAL argues that the Convention disallows awards of punitive damages and that the district court therefore erred in submitting the issue to the jury. KAL’s position was recently adopted by the Second Circuit in In re Air Disaster at Lockerbie, Scotland, on Dec. 21, 1988, 928 F.2d 1267 (2d Cir.1991) (“Lockerbie”).

Plaintiffs maintain that punitive damages are permitted whenever they are available under applicable local law and, in any event, should be recoverable once it is shown that the carrier acted with willful misconduct. For the reasons that follow, we agree with KAL and our sister circuits; hence, we set aside the jury's award of punitive damages.

The Warsaw Convention establishes the liability of international air carriers for harm to passengers, baggage, or goods; fixes limitations on such liability; and achieves a degree of uniformity in documentation and in the procedures and substantive law applicable to claims arising out of international air carriage. See Lowen-feld & Mendelsohn, The United States and the Warsaw Convention, 80 Harv.L.Rev. 497, 498-500 (1967). The central purpose of the Convention is to limit the liability of air carriers. The contracting states in 1929 believed that limitations on liability would promote the development of the fledgling commercial air industry by allowing the airlines to predict their exposure to monetary damages and thereby obtain needed capital and adequate insurance coverage. See Eastern Airlines v. Floyd, — U.S. -, -, 111 S.Ct. 1489, 1499-1500, 113 L.Ed.2d 569 (1991) (“Floyd II”); Lock-*1481erbie, 928 F.2d at 1270-71; Reed v. Wiser, 555 F.2d 1079, 1089 (2d Cir.), cert. denied, 434 U.S. 922, 98 S.Ct. 399, 54 L.Ed.2d 279 (1977); Lowenfeld & Mendelsohn, supra, 80 Harv.L.Rev. at 498-500.

Article 17 of the Convention defines a carrier’s liability for harm to passengers. See Air France v. Saks, 470 U.S. 392, 397, 105 S.Ct. 1338, 1341, 84 L.Ed.2d 289 (1985). Article 17 provides:

The carrier shall be liable for damage sustained in the event of the death or wounding of a passenger or any other bodily injury suffered by a passenger, if the accident which caused the damage so sustained took place on board the aircraft or in the course of any of the operations of embarking or disembarking.

Article 24 allows contracting states to decide the standing and “respective rights” of claimants who seek recovery under Article 17, provided that “any action for damages, however founded, can only be brought subject to the conditions and limits set out in this convention.” Under Article 25, a carrier “shall not be entitled to avail himself of the provisions of this convention which exclude or limit his liability” when it is shown that the damage suffered by the claimant was caused by the carrier’s “willful misconduct,” as defined by the law of the forum court.

Under the Convention as originally ratified, Article 22(1) limited the carrier’s Article 17 liability to approximately $8,300, and Article 20(1) allowed the carrier to avoid liability altogether by proving that it acted with due care. See Floyd I, 872 F.2d at 1467. With respect to flights to or from the United States, these provisions were superseded by the Montreal Agreement, an agreement among air carriers executed in 1966 and approved by the Civil Aeronautics Board pursuant to its authority under the Federal Aviation Act of 1958. In the Montreal Agreement, air carriers raised the per passenger liability limit to $75,000 and assumed virtual strict liability for death or injury of passengers by waiving their due care defenses under Article 20(1). See Order of Civil Aeronautics Board Approving Increases in Liability Limitations of Warsaw Convention and Hague Protocol, May 13, 1966 (approving Agreement CAB 18900), 31 Fed.Reg. 7302 (1966), reprinted in 49 U.S.C.App. § 1502 note. KAL has been a party to the Montreal Agreement since 1969. See Chan v. Korean Air Lines, 490 U.S. 122, 124, 109 S.Ct. 1676, 1678-79, 104 L.Ed.2d 113 (1989).

Whether punitive damages are allowable in actions governed by the Warsaw Convention must, in our opinion, turn on the nature of the liability contemplated by Article 17. The Convention sets forth an international liability scheme that was intended to be uniform. See Lockerbie, 928 F.2d at 1279-80; Floyd I, 872 F.2d at 1488. We must honor this uniformity by “givfing] the specific words of the treaty a meaning consistent with the shared expectations of the contracting parties.” Saks, 470 U.S. at 399, 105 S.Ct. at 1342. If a recovery is inconsistent with the shared expectations underlying Article 17, it is precluded. See Lockerbie, 928 F.2d at 1282; Floyd I, 872 F.2d at 1487. Cf. Floyd II, — U.S. at -, -, 111 S.Ct. at 1493-95, 1496-98 (term “bodily injury” in Article 17 excludes recovery for purely mental injuries); Saks, 470 U.S. at 405, 105 S.Ct. at 1345 (“accident” requirement precludes liability under Article 17 unless injury caused by unexpected or unusual event external to passenger).

Although the Convention does not address the issue of punitive damages, the Article 17 phrase “liable for damage sustained” strongly implies that the carrier’s responsibility is compensatory and extends only to the reparation of loss resulting from the death or injury of passengers. See Floyd I, 872 F.2d at 1483, 1486; In re Air Crash Disaster at Gander, Newfoundland, 684 F.Supp. 927, 931 (W.D.Ky.1987) (“Gander"). The words “damage sustained” do not refer to legal damages; they refer to actual harm experienced, whether physical injury to the passenger or, in the case of death, monetary or other loss to his survivors. See Lockerbie, 928 F.2d at 1281; see also Calkins, The Cause of Action Under the Warsaw Convention (pt. *14822), 26 J.Air L. & Com. 323, 335 (1959). Article 17 does not specify that the damage must be sustained “by the passenger” and thus allows for the possibility of traditional wrongful death actions. The requirement that the damage. “so sustained” be “caused” by an accident on board the aircraft, however, reenforces the conclusion that recovery is available only for actual loss; an accident cannot “cause” punitive damages. Lockerbie, 928 F.2d at 1281-82.

French is the controlling language of the Convention. See Article 36; Floyd II, — U.S. at -, 111 S.Ct. at 1493-95; Saks, 470 U.S. at 397, 105 S.Ct. at 1341. The relevant phrase in French is “dommage survenu.” The parties have not suggested that this phrase carries any special meaning in French law, and we are satisfied that “damage sustained” in the official American version of Article 17 — the version ratified by the Senate in 1934 — is an accurate translation of “dommage survenu.” See Lockerbie, 928 F.2d at 1280-81; Floyd I, 872 F.2d at 1486-87; Gander, 684 F.Supp. at 931.

The wording of Article 17 thus describes liability for compensatory or actual damages — that is, damages that “will compensate the injured party for the injury sustained ... [or] will simply make good or replace the loss caused by the wrong or injury.” Black’s Law Dictionary 352 (5th ed. 1979). In contrast, punitive damages are recognized by federal courts as retributive and deterrent in nature. See Lockerbie, 928 F.2d at 1284; Floyd I, 872 F.2d at 1487. They “are not compensation for injury,” but rather are awarded “to punish reprehensible conduct and to deter its future occurrence.” IBEW v. Foust, 442 U.S. 42, 48, 99 S.Ct. 2121, 2125, 60 L.Ed.2d 698 (1979) (quoting Gertz v. Robert Welch, Inc., 418 U.S. 323, 350, 94 S.Ct. 2997, 3012, 41 L.Ed.2d 789 (1974)). “Their role therefore ‘runs counter to the normal reparative function of tort and contract remedies.’ ” Browning-Ferris Indus. v. Kelco Disposal, Inc., 492 U.S. 257, 109 S.Ct. 2909, 2932, 106 L.Ed.2d 219 (1989) (O’Connor, J., concurring in part and dissenting in part) (quoting K. Redden, Punitive Damages § 2.1, at 24 (1980)). The purposes served by punitive damages are incompatible with the liability scheme set forth in Article 17. See Floyd I, 872 F.2d at 1487.

In Lockerbie, the Second Circuit pointed out that a minority of state courts in the United States view punitive or exemplary damages as serving, at least in part, a “compensatory” function. Lockerbie, 928 F.2d at 1284-85. The court concluded, however, that the Warsaw Convention preempts all state-law causes of action; otherwise, the uniform application of the Convention would be disrupted by the uneven patchwork of state punitive damages theories (a potential problem in diversity suits arising in states that adhere to the minority view). See id. at 1274-77. We need not reach this preemption question because plaintiffs here base their claim to punitive damages solely on federal maritime law. See Brief for Appellees at 29, 35-36.

Our construction of the Convention does not end with a word-by-word parsing of Article 17. We may properly look to the larger context of the Convention and its history, including the negotiations of its drafters and the interpretations given it by contracting states. See Floyd II, — U.S. at -, 111 S.Ct. at 1493-95; Saks, 470 U.S. at 396-97, 105 S.Ct. at 1340-41. In doing so, we must interpret and apply the terms of Article 17 in accordance with their French legal meaning because the Convention was drafted in French by civil law jurists. Floyd II, — U.S. at -, 111 S.Ct. at 1493-95; Saks, 470 U.S. at 399, 105 S.Ct. at 1342; see Block v. Compagnie Nationale Air France, 386 F.2d 323, 330 (5th Cir.1967), cert. denied, 392 U.S. 905, 88 S.Ct. 2053, 20 L.Ed.2d 1363 (1968).

Nothing in the minutes and notes of the Convention’s drafters indicates that the contracting parties ever considered the concept of punitive or deterrent damages for passengers’ deaths. See Lockerbie, 928 F.2d at 1280-81; Floyd I, 872 F.2d at 1487. The drafting history does indicate, however, that the drafters intended the carrier’s liability to be contractual in nature. *1483See Second International Conference on Private Aeronautical Law: Minutes (Warsaw 1929), at 22 (R. Horner & D. Legrez trans. 1975) (“Warsaw- Minutes”) (“The system of liability of the Convention is based only upon the document of carriage.”) (statement of Henri De Vos, reporter).

The kinds of damages available in French civil actions founded on contract parallel the compensatory forms of monetary recovery available in American tort suits. See Lockerbie, 928 F.2d at 1279-80, 1282. One commentator has observed, “There is nothing in French law prohibiting compensation for any particular kind of damage, be it mental injury, suffering due to the death of a member of the family, or pain and suffering due to a physical injury. Provided the damage is certain and direct, all forms of damage can be compensated to their full extent.” G. Miller, Liability in International Air Transport 112 (1977) (emphasis in original) (footnote omitted). The French concept of dommage materiel provides compensation for pecuniary loss resulting from death or injury; dommage moral is roughly equivalent to compensatory damages for nonpecuniary loss, such as pain and suffering. See R. Mankiewicz, The Liability Regime of the International Air Carrier 157 (1981); G. Miller, supra, at 112-14.

These notions of recoverable damages are compensatory and not punitive in character. The same can be said of the kinds of recovery afforded by other civil law contracting states, such as Germany. See Lockerbie, 928 F.2d at 1282. The Warsaw Convention practice of the principal common law contracting party, Great Britain, is consistent with this understanding. Although English courts in tort contexts do permit punitive damages in “unique and rare circumstances,” id. at 1282 (citing Rookes v. Barnard, [1964] 1 All E.R. 367, 410 (H.L.) (Lord Devlin)), they have construed “dommage” in Article 17 to encompass only “monetary loss.” See id. at 1281 (citing Fothergill v. Monarch Airlines Ltd., [1981] App.Cas. 251, [1980] 2 All E.R. 696, [1980] 2 Lloyd’s Rep. 295, 299 (H.L.) (Lord Wilberforce)). We conclude, therefore, that a monetary liability that is punitive and does not compensate for loss suffered would be contrary to the expectations of the jurists who drafted Article 17 and the contracting states that adopted it.

This conclusion comports with our obligation to construe the Convention in a manner that will promote uniformity. See Floyd II, — U.S. at -, 111 S.Ct. at 1500-02. The uniform application of the treaty would be threatened if the United States, alone among contracting states, imposed a form of liability wholly outside the compensatory scheme of Article 17. See Lockerbie, 928 F.2d at 1287-88; Floyd I, 872 F.2d at 1487-88. Undoubtedly, punitive liability for international carriers “would be controversial for most signatories,” and we should construe the Convention to avoid such a “potential source of divergence.” Floyd II, — U.S. at -, 111 S.Ct. at 1500-02.

In challenging our reading of “damage sustained,” plaintiffs rely on Smith v. Wade, 461 U.S. 30, 36 n. 5, 103 S.Ct. 1625, 1629 n. 5, 75 L.Ed.2d 632 (1983), in which the Supreme Court dismissed the argument that the phrase “for redress” in 42 U.S.C. § 1983 precludes recovery of punitive damages. Smith only confirms our analysis. In addition to noting that the tort cause of action created by section 1983 is expansive, 461 U.S. at 36 n. 5, 103 S.Ct. at 1629 n. 5, the Court relied on the common law context in which Congress enacted section 1983, a context in which punitive damages were a well-established tort remedy. See id. at 35, 37 n. 5, 103 S.Ct. at 1629, 1630 n. 5. Cf., Silkwood v. Kerr-McGee Corp., 464 U.S. 238, 255, 104 S.Ct. 615, 625, 78 L.Ed.2d 443 (1984) (federal regulation of nuclear plant safety does not preempt punitive damages award because “Congress assumed that traditional principles of state tort law would apply with full force unless ... expressly supplanted”). Similarly, we may infer from its civil law context that Article 17 does not contemplate the award of punitive damages.

Plaintiffs argue that Article 24’s references to actions for damages “however founded” and to claimants’ “respective *1484rights” make the availability of punitive damages a matter of local law. But Article 24 requires that any action for damages within the scope of the Convention satisfy the Convention’s “conditions and limits.” This requirement ensures the uniform application of the Convention. See N. Matte, Treatise on Air-Aeronautical Law 419 (1981). The most natural reading of the treaty is that the terms of liability established by Article 17 are among the “conditions” enforced by Article 24. See Calkins, supra, 26 J. Air L. & Com. at 327 (“conditions and limits” language in Article 24 designed “to protect the Conventional cause of action,” i.e., Article 17); see also G. Miller, supra, at 125 (“What Article 17 does is to determine the conditions under which an air carrier shall be liable.”); Floyd II, — U.S. at -, 111 S.Ct. at 1492-93. Thus, Article 24 provides a clear textual basis for concluding that the Convention overrides liability awards that are inconsistent with Article 17.

Some courts and commentators have concluded that Article 24 preserves independent domestic causes of action based on tort law. See, e.g., In re Aircrash in Bali, Indonesia, on Apr. 22, 1974, 684 F.2d 1301, 1311 n. 8 (9th Cir.1982); Calkins, supra, at 327. Others disagree, taking the view that the Convention, when applicable, provides the exclusive cause of action. See, e.g., Boehringer-Mannheim Diagnostics, Inc. v. Pan Am. World Airways, 737 F.2d 456, 458 (5th Cir.1984), cert. denied, 469 U.S. 1186, 105 S.Ct. 951, 83 L.Ed.2d 959 (1985). See also Floyd I, 872 F.2d at 1482 n. 33 (discussing cases). In deciding only that punitive damages are contrary to the Convention, we need not take sides in the exclusivity debate.

Regardless of their views on independent causes of action, most authorities agree that pursuant to Article 24 the proper “measure” of damages recoverable under Article 17 is left to the domestic law of the contracting states. See Harris v. Polskie Linie Lotnicze, 820 F.2d 1000, 1002 (9th Cir.1987); Mertens v. Flying Tiger Line, 341 F.2d 851, 858 (2d Cir.), cert. denied, 382 U.S. 816, 86 S.Ct. 38, 15 L.Ed.2d 64 (1965); R. Mankiewicz, supra, at 189; H. Drion, Limitation of Liabilities in International Air Law 125-26 (1954). This fact, however, begs the question whether such damages may be more than compensatory.

We think it evident they may not. In reporting the preliminary draft of the proposed treaty to the national delegates at Warsaw, the International Technical Committee of Air Law Experts (“CITEJA”) stated:

The question was asked ... if one could determine who the persons upon whom the action devolves in the case of death are, and what are the damages subject to reparation. It was not possible to find a satisfactory solution to this double problem, and the CITEJA esteemed that this question of private international law should be regulated inde-pendantly [sic] from the present Convention.

Warsaw Minutes at 255 (emphasis added). The reporter’s use of the word “reparation” bolsters our reading of Article 17 and “tends to exclude the concept of punitive damages.” Lockerbie, 928 F.2d at 1284.

Moreover, the drafting history shows that the CITEJA’s inability to formulate a single rule for calculating damages resulted from the difficulty of choosing among widely varying national laws of descent and distribution governing who could sue on a decedent’s behalf in wrongful death actions and who could sue only for personal loss. See id. at 1283-84; see also Calkins, supra, at 327-28 (discussing excerpts of drafting record). This inability does not suggest that the drafters ever contemplated the possibility of imposing liability that goes beyond compensation for loss, however determined. Lockerbie, 928 F.2d at 1284.

Nor does a finding of willful misconduct under Article 25 create a right to recover punitive damages. Article 25 bars the carrier from relying on those provisions in the Convention that “exclude or limit” liability. Article 17 is not among these. It is settled that willful misconduct negates *1485the due care exclusion from liability contained in Article 20 and the monetary limitations contained in Article 22. See Lockerbie, 928 F.2d at 1286; Grey v. American Airlines, 227 F.2d 282, 285 (2d Cir.1955), cert. denied, 350 U.S. 989, 76 S.Ct. 476, 100 L.Ed. 855 (1956). Disagreement about whether other provisions are affected, see H. Drion, supra, at 80 n. 5 (citing authorities), does not obscure the fact that certain key articles in the Convention continue to apply in cases of willful misconduct, and no authority suggests that the basic liability terms of Article 17 (or any of the other “conditions” preserved by Article 24) were to be displaced. See Lockerbie, 928 F.2d at 1286. Given the drafting context discussed above, we agree with the Second Circuit that the drafters would not have considered Article 17 a “limitation” on liability. See id. at 1286. In sum, Article 25 is “most reasonably interpreted as [an] exception ] to the limitations on the recovery of compensatory damages within the Convention, not as authority for the recovery of punitive damages.” Gander, 684 F.Supp. at 932 (emphasis in original); accord Floyd I, 872 F.2d at 1484.

The Hague Protocol of 1955 and the Montreal Protocol No. 4 of 1975 provide further confirmation. These Protocols, among other things, clarified Article 25 to make it explicit that the limits on liability lifted in the event of willful misconduct are only the monetary limits contained in Article 22. See Floyd I, 872 F.2d at 1483-84 & n. 35. The United States has not ratified the Hague or Montreal Protocols and they are not binding on us, but we may look to them for clarification of the Convention’s terms. Id. at 1484 n. 35; see Saks, 470 U.S. at 403-04, 105 S.Ct. at 1344-45 (“subsequent interpretations of the signatories help[ ] clarify the meaning” of terms (referring to history of unratified Protocols)); see also Floyd II, — U.S. at -, 111 S.Ct. at 1498-1500. As the dissent points out, the Hague Protocol did more than clarify Article 25; it also “narrow[ed] the waiver of limitations in cases of willful misconduct.” Dissenting Op. 1494. It did so by adopting a new treaty definition of willful misconduct that supplanted Article 25’s reference to forum law. See Lowen-feld & Mendelsohn, supra, 80 Harv.L.Rev. at 503-06. But that substantive amendment does not belie the view that the Protocol’s specific reference to the monetary limits of Article 22 was understood only to be a clarification. See Floyd I, 872 F.2d at 1483.

The policies underlying the Convention support our conclusions. The Convention represented a bargain between air carriers on the one hand and passengers and shippers on the other. Carriers obviously gained from the limitations on liability agreed to by the contracting states, but passengers and shippers also benefited, primarily from the Convention’s clear presumption of liability, which eliminated the difficult task of proving fault on the part of the carrier. Furthermore, under the law of contract as it existed prior to the Convention, the carrier could essentially avoid liability for harm resulting from the often hazardous business of air transportation:

[I]n reality, this Convention creates against the air carrier an exceptional system, because in the majority of the countries of the world, contracts of carriage are concluded under a system of free contract. The carrier is free to insert in the contract clauses which exclude or reduce his liability, as much for goods as for travelers. You are of course aware that they have lost no opportunity to do so, and presently many air carriers operate under this unregulated contractual system, and, in practice, in fact, they are not liable.

Warsaw Minutes at 47 (statement of Georges Ripert of France).

The essential bargain struck by the contracting parties in Warsaw was presented to the United States Senate as a principal reason for ratifying the Convention:

It is believed that the principle of limitation of liability will not only be beneficial to passengers and shippers as affording a more definite basis of recovery and as tending to lessen litigation, but that it will prove to be an aid in the development of international air transportation, as such limitation will afford the carrier *1486a more definite and equitable basis on which to obtain insurance rates, with the probable result that there would eventually be a reduction of operating expenses for the carrier and advantages to travelers and shippers in the way of reduced transportation charges.

Senate Comm, on Foreign Relations, Message from the President of the United States Transmitting a Convention for the Unification of Certain Rules Relating to International Transportation by Air, Sen. Exec.Doc. G, 73d Cong., 2d Sess. 3-4 (1934) (Report of Secretary of State Cordell Hull). The award of punitive damages would be contrary to this essential bargain because it would increase the amount of litigation, the cost of insurance, and ultimately the price of air transportation. See Lockerbie, 928 F.2d at 1287-88; Floyd I, 872 F.2d at 1487-88.

Our dissenting colleague suggests that our reading of the Convention will leave undeterred the diabolical carrier who might deliberately sabotage its own aircraft in the hope of reaping a fraudulent insurance recovery. This scenario ignores, of course, the important deterrence provided by the criminal law. Moreover, because of the “accident” requirement in Article 17, it is by no means certain that the protections of the Convention would be available to a carrier that destroyed its own aircraft. Cf. Floyd II, — U.S. at -, 111 S.Ct. at 1493-95 (“an air carrier is liable for passenger injury [under the Convention] only when ... there has been an accident”); Saks, 470 U.S. at 406-07, 105 S.Ct. at 1345-46, (same). As the dissent’s scenario is not before us, we express no opinion on any speculative questions it might or might not raise.

What we can and do say, however, is that whatever evil one might be willing to impute to the corporate mind, the possibility of so wanton an act does not license us to disregard the policy choices made by the Convention’s contracting parties. See Floyd II, — U.S. at -, 111 S.Ct. at 1498-1500 (“Whatever may be the current view among Convention signatories, in 1929 the parties were more concerned with protecting air carriers and fostering a new industry than providing full recovery to injured passengers, and we read [the terms of Article 17] in a way that respects that legislative choice.”). Article 17 of the Warsaw Convention “sets the parameters” of plaintiffs’ right of recovery at compensatory damages only. Gander, 684 F.Supp. at 932. These parameters are not expanded by Articles 24 or 25. Accordingly, we vacate the jury’s award of punitive damages.

KAL argues that the district court committed a second error by failing to undertake a choice-of-law analysis prior to instructing the jury on punitive damages. As we conclude that such damages are not recoverable even if plaintiffs are correct about choice of law, this issue is not relevant to our disposition of the case. We therefore express no view on this aspect of the dissenting opinion.

III. Conclusion

The district court properly left the willful misconduct decision to the jury and did not abuse its discretion in admitting plaintiffs’ evidence. The district court erred, however, in allowing plaintiffs to pursue their punitive damage claims, and therefore we vacate the jury’s punitive damage award in this case.

So ordered.

Dissenting Opinion

MIKVA, Chief Judge,

dissenting from Part 11(C):

I do not share my colleagues’ conclusion that the Warsaw Convention bars recovery of punitive damages. Instead, I would remand with instructions that the district court engage in a proper choice of law analysis.

A. Warsaw Convention

The Warsaw Convention, officially denominated the Convention for the Unification of Certain Rules Relating to International Transportation by Air, Oct. 12, 1929, was adhered to by the United States in 1934. See 49 Stat. 3000, T.S. No. 876 (1934), reprinted in 49 U.S.C.App. § 1502, note. The Convention was negotiated with hopes of facilitating the growth of the in*1487fant airline industry, and the drafters sought to create a scheme that would presume carrier liability for air disasters but cap monetary liability at a very low level. Article 17 provides that “[t]he carrier shall be liable for damage sustained in the event of the death or wounding of a passenger....” As the majority explains, this liability appears to cover only compensatory damages. Article 22 originally capped monetary damages at approximately $8,300 per passenger. That figure was raised to $75,000 by the 1966 Montreal Agreement for flights to and from the United States (undermining the goal of uniformity for the sake of preventing denunciation by the United States government).

Two other provisions of the Convention are of particular relevance in the instant appeal. Article 24 leaves procedural questions to local law provided that “any action for damages, however founded, can only be brought subject to the conditions and limits set out in this convention.” This language contemplates the availability of causes of action that are separate from Article 17, though governed by the limitations of the Convention. Finally, Article 25 provides that “[t]he carrier shall not be entitled to avail himself of the provisions of this convention which exclude or limit his liability, if the damage is caused by his willful mis-conduct_” The dispute before us re-

quires that we carefully sort out the interplay of these various sections.

The majority holds that the Warsaw Convention was intended to provide compensatory relief only, and that therefore punitive damages are not available even in causes of action not based on Article 17. Nowhere are punitive damages explicitly prohibited. In this country, “legislative silence with respect to punitive damages do[es] not preclude such a recovery.” Racich v. Celotex Corp,, 887 F.2d 393, 396 (2d Cir.1989); see Silkwood v. Kerr-McGee Corp., 464 U.S. 238, 255, 104 S.Ct. 615, 625, 78 L.Ed.2d 443 (1984). Even so, I am persuaded by the majority’s conclusion that Article 17 cannot be read to authorize the recovery of punitive damages for claims based on the Convention. If any of the plaintiffs in these cases premised their complaints solely on a Warsaw Convention cause of action, I would concur in holding that they were not entitled to recover punitive damages. But few if any of the complaints were so limited in pleading only the Convention as a cause of action.

The plaintiffs in these cases brought their damages claims against KAL under a potpourri of legal theories including the Warsaw Convention. In denying KAL’s motion to strike plaintiffs’ jury demand, the district court untangled the sources underlying each of the plaintiff’s complaints and found that most alleged a combination of state and federal claims (both common law and statutory) along with claims based on the Convention. See In re Korean Air Lines Disaster of September 1, 1983, 704 F.Supp. 1135, 1151-55 (D.D.C.1988). Almost all of the complaints alleged something other than merely the Convention as a basis for recovery and several failed to plead the applicability of the Convention at all. See id. at 1151 n. 40, 1154— 55. The district court correctly emphasized that, whatever the cause of action, the claims were all subject to the limitations contained in the Convention.

The majority decides that complaints sounding in other causes of action such as federal maritime law are also not entitled to punitive damage awards. I find this step in the court’s logic somewhat difficult to fathom. There appear to be two basic rationales underlying such a conclusion: (1) the Convention provides the exclusive cause of action (thereby entirely preempting other possible causes of action that might separately allow punitive damages), or (2) Article 17 creates an implicit limitation on liability governing recoveries premised on separate causes of action (a.k.a. partial preemption). The majority appears to employ the latter rationale, but then dismisses the force of Article 25 in part on the strength of decisions premised on the exclusivity rationale. I think neither one is persuasive and therefore respectfully dissent.

1. Exclusivity

The majority correctly eschews reliance on the first approach, choosing “not [to] *1488take sides in the exclusivity debate.” Op. at 1488. The Convention is not the exclusive remedy for passengers injured on international flights; in fact, Article 24’s reference to “any action for damages, however founded” clearly contemplates actions arising under separate sources of law but places some limits on recovery. Courts and commentators have rejected suggestions that the Convention provides the exclusive cause of action for international air disasters. See, e.g., In re Air Crash in Bali, Indonesia, 684 F.2d 1301, 1311 n. 8 (9th Cir.1982) (“[T]he Convention has never been read to limit plaintiffs to a cause of action arising thereunder, but rather to limit the recovery in suits for injury.”); Tokio Marine & Fire Insur. Co. v. McDonnell Douglas Corp., 617 F.2d 936, 942 (2d Cir.1980) (“[T]he Convention draftsmen ... did not intend that cause of action to be exclusive.”); Calkins, The Cause of Action under the Warsaw Convention, 26 J. Air L. & Com. 323, 327-28 (1959).

The Second Circuit just recently decided that the Warsaw Convention prohibits the recovery of punitive damages. See In re Air Disaster at Lockerbie, Scotland on Dec. 21, 1988, 928 F.2d 1267 (2d Cir.1991). The court took exclusivity as its working premise, see id. at 1273-75, 1282-83, somewhat astonishing in light of the Second Circuit’s own precedents. For a long time, that court had held that no cause of action is established by the Convention. See Benjamins v. British European Airways, 572 F.2d 913, 919 (2d Cir.1978) (finally overruling proposition that had been accepted for twenty years), cert. denied, 439 U.S. 1114, 99 S.Ct. 1016, 59 L.Ed.2d 72 (1979). In subsequent decisions, the court ruled that this cause of action was not exclusive. See Tokio Marine, 617 F.2d at 942. Now, by dismissing the statement in Tokio Marine as mere dicta, the Second Circuit has gone full circle from believing that the Convention provides no cause of action to deciding that it provides the sole cause of action for plaintiffs.

Apart from its apparent failure to abide by circuit precedent, the Lockerbie court’s support for the exclusivity holding is hardly overwhelming. For instance, the court emphasized (at 1274) that other countries have held that the Convention is the exclusive cause of action. These were not interpretations of the Convention, however: the common law countries cited in the opinion all had to enact national legislation making Article 17 the exclusive cause of action. The United States has, of course, never done so. Moreover, its description of decisions from other circuits is also somewhat perplexing. For instance, in one case the Ninth Circuit observed that “the delegates did not intend that the cause of action created by the Convention to be exclusive.” In re Mexico City Aircrash of Oct. 31, 1979, 708 F.2d 400, 414 n. 25 (9th Cir.1983). Yet the Lockerbie court asserts that the decision “rebutted the idea that a cause of action may be founded on some law other than the Convention.” 928 F.2d at 1273-74.

In fact, the Ninth Circuit has concluded on more than one occasion that the cause of action is not exclusive. See, e.g., In re Air Crash in Bali, Indonesia, 684 F.2d at 1311 n. 8. All told, the decisions from other circuits appear to be evenly divided on the issue, notwithstanding the court’s claim in Lockerbie, at 1282-83 that the weight of authority is fairly one-sided. See Floyd v. Eastern Airlines, Inc., 872 F.2d 1462, 1482 n. 33 (11th Cir.1989) (collecting cases), rev’d on other grounds, — U.S. -, 111 S.Ct. 1489, 113 L.Ed.2d 569 (1991). See also Alvarez v. Aerovias Nacionales de Colombia S.A., 756 F.Supp. 550 (S.D.Fla.1991) (distinguishing exclusive remedy of the Convention from the suggestion that it also provides the exclusive cause of action). The Supreme Court has twice now declined to address the exclusivity question. See Eastern Airlines, Inc. v. Floyd, — U.S. at -, 111 S.Ct. at 1502; Air France v. Saks, 470 U.S. 392, 408, 105 S.Ct. 1338, 1346-47, 84 L.Ed.2d 289 (1985). The fundamental error of the Lockerbie court’s decision is the premise that the cause of action provided by the Warsaw Convention is exclusive, and my colleagues are wise not to enter this fray unnecessarily-

*14892. Implicit Limitation in Article 17

If the court does not premise its holding on an exclusivity rationale, it must mean that Article 17 acts as a limitation on the liability that an air carrier would face when sued under separate causes of action. But then Article 25 would come into play and waive any such limitation in cases of willful misconduct: “The carrier shall not be entitled to avail himself of the provisions of this convention which exclude or limit his liability, if the damage is caused by his wilful misconduct_” Art. 25(1). The majority relies heavily on Lockerbie for the proposition that Article 25 does not trump Article. 17, but there the court had taken exclusivity as its starting point. It is one thing to say that the Convention prohibits punitive damages because it is the sole available cause of action and does not provide for them; it is quite another to conclude that Article 17 acts as a limitation on damages restricting recovery under other causes of action that allow punitive damages.

The Eleventh Circuit has also decided not to take sides in the debate over exclusivity. In resolving the punitive damages question in Floyd, the court first decided that the Convention is not itself a basis for seeking punitive damages. The court concluded that Article 17 was entirely compensatory in tone and rejected plaintiffs’ argument that Article 25’s reference to willful misconduct implied an exception. See 872 F.2d at 1483-85. Next the court rejected the plaintiffs’ effort to seek punitive damages under state common law, holding that state law was preempted to the extent that it was inconsistent with the compensatory scheme of Article 17. See id. at 1485-87.

Preemption analysis will not work in this case, however, because we are confronted with two legal schemes that stand in rough equipoise under the Supremacy Clause: the Warsaw Convention and federal maritime law. See In re Air Crash Disaster Near New Orleans, 821 F.2d 1147, 1161 n. 19 (5th Cir.1987) (“As a ratified treaty of the United States the Warsaw Convention is equal in stature and force as any other domestic federal law.”); Committee of U.S. Citizens in Nicaragua v. Reagan, 859 F.2d 929, 936-37 (D.C.Cir.1988). The Convention cannot, therefore, preempt federal law to the extent of any inconsistency with its compensatory tone; Article 17 can only prevent recovery of punitive damages otherwise available under federal maritime law if it is construed as a limit or condition governing damages under Article 24, but then it becomes subject to waiver in cases where Article 25 applies.

The court in Floyd also relied on subsequent proposals which would have amended Article 25 so that it would only trump the monetary cap in Article 22. See 872 F.2d at 1483-84 (discussing the Hague and Montreal Protocols). But the original language of Article 25 does not suggest that only Article 22 was covered, and the United States never ratified these amendments. See id. at 1468-69. Moreover, the United States responded to the proposed changes by threatening to denounce the entire Convention unless the monetary limits were raised substantially. See Lowenfeld & Mendelsohn, The United States and the Warsaw Convention, 80 Harv.L.Rev. 497, 509-16, 532-46 (1967). In Floyd, the Eleventh Circuit candidly admitted that this country never ratified the Protocols, but found “their clarification of the operation of Article 25 to be instructive.” 872 F.2d at 1484 n. 35. This reasoning is difficult to square with the Supreme Court’s recent instruction that “where the text is clear ... we have no power to insert an amendment.” Chan v. Korean Air Lines, Ltd., 490 U.S. 122, 134, 109 S.Ct. 1676, 1683, 104 L.Ed.2d 113 (1989) (“We must thus be governed by the text — solemnly adopted by the governments of many separate nations—whatever conclusions might be drawn from the intricate drafting history_”).

Indeed, in reversing the decision that compensatory damages for purely mental injuries were available under Article 17, the Supreme Court criticized the lower court’s interpretation of subsequent drafting history surrounding the Protocols. See Eastern Airlines, Inc. v. Floyd, — U.S. at - -, 111 S.Ct. at 1498-1502. To the extent that the 1955 negotiating history cited by the Eleventh Circuit in Floyd reflects the *1490Hague drafters’ understandings about what their predecessors may have had in mind a generation earlier, it is entitled to only marginal weight. See McKelvey v. Turnage, 792 F.2d 194, 200 (D.C.Cir.1986) (“As the Supreme Court and this court have repeatedly remarked, the views of later Congresses as to the meaning of enactments by their predecessors are of little, if any, significance.”), aff'd, 485 U.S. 535, 108 S.Ct. 1372, 99 L.Ed.2d 618 (1988).

At best, the subsequent proposals and accompanying drafting history demonstrate only that the original text was ambiguous. At worst, the proposals imply that the negotiators wanted to adjust the bargain that had been struck originally. Indeed, the amended version of Article 25 proposed in the unratified Hague Protocol would not have simply clarified an existing understanding of Article 25 but represented a revised quid pro quo: in exchange for a higher ceiling on damages, carriers sought to narrow the waiver of limitations in cases of willful misconduct. See Lowenfeld & Mendelsohn, at 503-06. See also Comment, Role of Choice of Law in Determining Damages for International Aviation Accidents, 51 J. Air L. & Com. 953, 989 (1986) (discussing Guatemala Protocol which would have eliminated Article 25 altogether in exchange for a higher monetary cap of $100,000). Thus, the history as well as the language of Article 25 supports the conclusion that punitive damages are not prohibited by the Warsaw Convention in cases of willful misconduct even if a cause of action based solely on the Convention would not authorize such damages.

The majority concludes that allowing punitive damage windfalls would contravene the Convention’s primary purpose and understanding of the contracting parties that the treaty was intended to limit the liability of air carriers. The scheme of the Convention closely resembles workmen’s compensation statutes which allow workers to recover prescribed compensatory damages for accidental job-related injuries. These laws reflect a quid pro quo between employees, who are relieved of the burden of proof, and employers, who benefit from the cap on damages. See 2A A. Larsen, Workmen’s Compensation Law § 65.11 (1990), at 12-9. Similarly, the Warsaw Convention balances strict carrier liability for injuries sustained by passengers (Articles 17 and 20) against a cap on monetary damages (Article 22). See Block v. Compagnie Nationale Air France, 386 F.2d 323, 327 (5th Cir.1967), cert. denied, 392 U.S. 905, 88 S.Ct. 2053, 20 L.Ed.2d 1363 (1968).

However, workmen’s compensation laws do not limit the availability of separate damage actions in cases of intentional misconduct by the employer. See, e.g., Pratt v. National Distillers & Chem. Corp., 853 F.2d 1329, 1336-39 (6th Cir.1988) (punitive damages are available for intentional injuries), cert. denied, 489 U.S. 1012, 109 S.Ct. 1121, 103 L.Ed.2d 184 (1989); Larsen §§ 68.00-.14. Similarly, the narrow exception in Article 25 of the Convention does not undermine the basic purpose of limiting carrier liability; it simply recognizes that in cases of serious misconduct carriers should not enjoy the benefits of these limitations. Imagine that a carrier decided to deliberately sabotage one of its own flights in hopes of receiving a large insurance settlement on the aircraft. If such egregious conduct went undetected, a financially strapped airline could invest a paltry amount in anticipated payments to decedents’ estates (on a typical flight of the Concorde, for instance, that might create maximum liabilities of $500,000 to the passengers) in exchange for a multi-million dollar insurance pay-off. Even if such unthinkable conduct came to light, the maximum compensatory judgment to the plaintiffs might still be less than the potential insurance recovery discounted by the risk that the fraud might be discovered. The majority’s decision would completely foreclose the availability of punitive damages even in such an egregious case. While the facts of the present case are not as stark as those hypothesized, the point remains the same: if plaintiffs can demonstrate willful misconduct, the limitations of the Convention (and the goals that those limitations serve) should have no application.

*1491B. Choice of Law

KAL’s pre-trial motion to dismiss plaintiffs’ punitive damage claims was premised exclusively on its argument that the Warsaw Convention precluded such an award. Near the end of the trial, the district court denied KAL’s motion from the bench. Apparently for the first time in the proceedings, KAL’s counsel then requested a choice of law analysis, but in a very vague and non-committal way.

The Court: What law should I pick?

Mr. Barry: I don’t know.

The Court: Where is this case being tried on the issue of liability?

Mr. Barry: Here in the District of Columbia.

The Court: With respect to damages, this aspect of damages which would ordinarily tail along will be a part of an action on liability and damages, except for the construct of the treaty; I mean, what law would apply?

Mr. Barry: Well, Your Honor, I think you have to do a choice of law analysis, just as Judge Green did in the Air Florida litigation here on this very same issue, and she held that laws of differing jurisdictions were going to apply to different issues_ I don’t know the answer to the question.... I raise only a problem. I wasn’t anticipating this, to tell you the truth.

The decision alluded to by KAL counsel was In re Air Crash Disaster at Washington, D.C. on January 13, 1982, 559 F.Supp. 333 (D.D.C.1983) (hereinafter “In re Air Florida Crash ”). The district court’s questions reflect a couple of assumptions: that forum law would automatically govern liability issues and that damages would presumptively be governed by the same jurisdiction’s laws. Both premises are questionable: choice of law principles followed by the District of Columbia courts seek to apply the substantive law of the jurisdiction with the closest connection to and greatest interest in a claim, see Hercules & Co. v. Shama Restaurant Corp., 566 A.2d 31, 40-41 (D.C.App.1989), and separate choice of law analysis of different substantive issues (called “depe-cage”) is now routine. See id.; In re Air Florida Crash, 559 F.Supp. at 341.

The threshold question is whether the choice of law basis for KAL’s motion to dismiss the punitive damages claims was raised too late. Unlike jurisdictional issues, courts need not address choice of law questions sua sponte. It is not clear whether the District of Columbia would follow a default rule that presumes local law controls unless choice of law objections are raised in a timely manner, but that seems to be the norm. See Cavic v. Grand Bahama Devel. Co., 701 F.2d 879, 882-83 (11th Cir.1983); Kramer, Interest Analysis and the Presumption of Forum Law, 56 U.Chi.L.Rev. 1301 (1989). Although “[t]he choice of law question regarding punitive damages should be resolved as soon as possible,” In re Air Crash Disaster at Sioux City, Iowa, on July 19, 1989, 734 F.Supp. 1425, 1429 (N.D.Ill.1990) (hereinafter “In re Iowa Crash”), Rule 12(h)(2), Fed.R.Civ.P., provides that “[a] defense for failure to state a claim upon which relief can be granted ... may be made ... [as late as] the trial on the merits.” If the short verbal exchange cited above was enough to place the issue before the judge, then the district court should have engaged in a choice of law analysis or at least asked for supplemental briefing at that point. See Smith v. Atlas Off-Shore Boat Service, Inc., 653 F.2d 1057, 1059-60 n. 1 (5th Cir.1981) (failure to state a claim could be pressed on appeal where defendant had “sufficiently alerted” the court of the defense during trial). KAL preserved the argument, though just barely. Cf. In re Air Florida Crash, 559 F.Supp. at 336-37 (rejecting contention that Air Florida “waived” its right to present a choice of law argument for the first time on reconsideration of the district court’s ruling, noting that its “interest in the question arose only when this Court ruled that divergent punitive damages laws would govern these parties’ liability”).

The next question is whether KAL properly pled that Korean law was applicable. Rule 44.1, Fed.R.Civ.P., does not require that a choice of law argument based on *1492foreign law appear in the pleadings. Instead, as the Advisory Committee explained, “[t]he new rule does not attempt to set any definite limit on the party’s time for giving the notice of an issue of foreign law; in some cases the issue may not become apparent until the trial and notice then given may still be reasonable.” In a related vein, appellees contend that KAL failed to satisfy the proof requirements in Rule 44.1, but the explanation in the Advisory Committee notes suggests otherwise. See also Pollack, Proof of Foreign Law, 26 Am.J. Comp.L. 470 (1978). Along with its memoranda of law in support of its JNOY motion, KAL filed a translation and explanatory letter concerning the relevant provisions of the Korean civil code, and added that “KAL preserved its argument that Korean law applied to the issue of punitive damages by including in each Answer filed in this litigation a ‘Notice of Applicability of Foreign Law’ pursuant to Fed.R.Civ.P. 44.1.” The district court never explicitly rejected KAL’s choice of law argument on grounds of being unreasonably late, and I believe it should have been addressed.

The district court entirely failed to engage in the required choice of law analysis. KAL contends that the district court improperly applied D.C. law on the punitive damages question because a proper choice of law analysis would have required reference to Korean civil law. KAL argues that, in multi-district airplane crash litigation, involving plaintiffs from numerous other jurisdictions (both foreign and domestic), the trial court could not just unthinkingly apply forum law to the punitive damages question. Indeed, several other courts have engaged in sometimes agonizing choice of law analyses in similar cases. See In re Air Crash Disaster Near Chicago, Illinois on May 25, 1979, 644 F.2d 594 (7th Cir.), cert. denied, 454 U.S. 878, 102 S.Ct. 358, 70 L.Ed.2d 187 (1981) (hereinafter “In re Chicago Crash ”); In re Iowa Crash, 734 F.Supp. 1425; In re Air Florida Crash, 559 F.Supp. 333. See generally Lowenfeld, Mass Torts and the Conflict of Laws: The Airline Disaster, 1989 U.Ill.L. Rev. 157.

For state law claims brought in federal court under diversity jurisdiction, the district court must apply the choice of law principles of the state in which it sits. See Klaxon Co. v. Stentor Elec. Mfg. Co., 313 U.S. 487, 496, 61 S.Ct. 1020, 1021-22, 85 L.Ed. 1477 (1941). When a case is transferred pursuant to 28 U.S.C. § 1407(a) by the Panel on Multi-District Litigation, the transferee court must apply the choice of law rules of the states where the transfer- or courts sit. See In re Chicago, 644 F.2d at 610 (citing Van Dusen v. Barrack, 376 U.S. 612, 84 S.Ct. 805, 11 L.Ed.2d 945 (1964)). No such analysis was undertaken in this case. Cf. Day & Zimmermann, Inc. v. Challoner, 423 U.S. 3, 4, 96 S.Ct. 167, 168, 46 L.Ed.2d 3 (1975) (per curiam) (reversing federal court’s decision to ignore forum state’s choice of law rule that would have required application of Cambodian law to wrongful death action). The district court appears to assume that forum law automatically controls and that dépecage is inappropriate, but even application of District of Columbia choice of law principles would dictate otherwise, see Hercules & Co., 566 A.2d at 40-41, and we have no idea what would happen if the district court applied the choice of law principles of each transferor court. The district court erred in failing to undertake a choice of law analysis in this case, a task best left in the first instance to the district court on remand. Even so, it is necessary to explain at some length below why I am unpersuaded by appellees’ arguments.

Appellees suggest that their punitive damage claims were not derived from state law but arose from general federal maritime law. In similar circumstances, the Third Circuit held that

the maritime character of the tort brings the controversy under the governance of federal law and it is immaterial whether admiralty or diversity jurisdiction is relied upon as justification for suing in the federal forum. Obviously, a court thus undertaking to apply federal substantive law would have no occasion to defer to or apply state choice of law rules.

Scott v. Eastern Air Lines, Inc., 399 F.2d 14, 17 (3d Cir.1967) (plane crash in Boston *1493Harbor), cert. denied, 393 U.S. 979, 89 S.Ct. 446, 21 L.Ed.2d 439 (1968). The district court never explained its refusal to engage in a Klaxon choice of law analysis in these terms, though its treatment of KAL’s motion to strike plaintiffs’ jury demand suggests this reasoning. See In re KAL Disaster, 704 F.Supp. at 1151-57. There the district court held that all of the plaintiffs had properly pleaded (or could amend their complaints to so plead) federal maritime claims and the Warsaw Convention as causes of action, see id. at 1154, 1155-56 (but recognizing that some plaintiffs had only pled state law claims brought pursuant to diversity jurisdiction).

Even if that was the district court’s undisclosed reason for not engaging in a Klaxon analysis for each transferor court, it would not relieve the court of the need to engage in some choice of law analysis under federal law. See Lauritzen v. Larsen, 345 U.S. 571, 575-77, 73 S.Ct. 921, 924-26, 97 L.Ed. 1254 (1953) (finding a conflict between application of Danish and U.S. maritime law). In other words, if application of federal maritime law would authorize punitive damages but Korean law would not, a conflict may exist that needs to be resolved. This is precisely what happened in Harris v. Polskie Linie Lotnicze, 820 F.2d 1000 (9th Cir.1987), a case where the court had to decide as a matter of federal choice of law principles whether Polish law or California law provided the measure of recovery in a wrongful death claim arising from a crash in Poland and governed by the Foreign Sovereign Immunities Act.

In undertaking the federal choice of law analysis, the court in Harris looked to the Restatement (Second) of Conflict of Laws (1969) as “an appropriate starting point.” See id. at 1003-04. The Restatement is, however, a dubious choice of law methodology for a federal court to pick. Choice of law is a very messy field, with different approaches endorsed by competing camps. See In re Paris Air Crash of March 3, 1974, 399 F.Supp. 732, 739 (C.D.Cal.1975) (describing it as “a veritable jungle”). In the past, the lex loci delicti rule was widely used in tort eases, directing courts to use the substantive law of the jurisdiction where the injury occurred, and vestiges of this much criticized approach remain in the Restatement. See Restatement (Second) of Conflict of Laws (hereinafter “Rest.”) §§ 145-146, 175 (presumption of applying the local law where the injury occurred unless another jurisdiction has a more significant relationship to the incident or the parties involved). In Harris, the court decided to apply Polish law of damages, largely because that was the site of the accident and no other jurisdiction had a demonstrably greater interest in the case. See 820 F.2d at 1004 & n. 5. In fact, the lex loci starting point used by the Restatement might be difficult to apply in this case: it is not clear whether KE007 was shot down in Soviet airspace, over Japanese territory or in international waters. Cf. In re Air Florida Crash, 559 F.Supp. at 348 n. 21 (describing earlier case involving a DC-4 mid-air collision where half of the wreckage fell into the Potomac River (deemed part of the District of Columbia) and the other half fell on the Virginia shore); In re Air Crash Disaster Near Bombay, India, 531 F.Supp. 1175, 1182 (W.D.Wash.1982) (parties disputed whether aircraft crashed in international waters or within India’s territorial waters). Indeed, largely because of such uncertainties, the Supreme Court has declined to place much emphasis on the lex loci rule in maritime torts. See Romero v. International Terminal Oper. Co., 358 U.S. 354, 384, 79 S.Ct. 468, 486-87, 3 L.Ed.2d 368 (1959); Lauritzen, 345 U.S. at 583-84, 73 S.Ct. at 928-29.

The Restatement also endorses an amalgam of the “most significant relationship” (or “center of gravity”) test and governmental interest methodologies, see Rest. § 6, reflecting an uneasy compromise struck among the drafters. See von Meh-ren, Recent Trends in Choice-of-Law Methodology, 60 CORNELL L. Rev. 927, 963-64 (1975). The governmental interest approach seeks to identify which jurisdictions may have an actual interest in having their substantive law apply to a particular controversy, but the resolution of true conflicts is achieved through different means *1494depending on the variant. See generally G. Smith, Choice of Law in the United States, 38 Hastings L.J. 1041, 1043-50 (1987). Whatever the relative merits of the various approaches, the court in Harris went straight to the Restatement as a good source for general choice of law rules to be used as federal common law in the area.

In counting up the contacts, KAL emphasizes that South Korea is its place of incorporation, its principal place of business, and the place where its crews are trained. Ap-pellees respond that many of the passengers came from, the flight originated in, and all of the tickets were purchased in the United States. They cite a decision that suggests using a “center of gravity” approach like the Restatement in maritime cases. See Hellenic Lines Ltd. v. Rhoditis, 398 U.S. 306, 90 S.Ct. 1731, 26 L.Ed.2d 252 (1970). In that case, the Supreme Court applied the Jones Act to a claim by a Greek seaman injured on a Greek ship while it was in American territorial waters, emphasizing that the Greek company that owned the ship had its largest offices in the U.S., was owned by a Greek citizen domiciled in the U.S., and transported cargo to and from the United States. See id. at 307-10, 90 S.Ct. at 1733-35. But Hellenic Lines does not necessarily support appel-lees’ analysis of the relevant contacts in this case because the site of the accident and KAL’s principal place of business were not in the United States. See Romero, 358 U.S. at 383-84, 79 S.Ct. at 486-87 (applying Spanish law to maritime tort claim brought by a Spanish sailor injured on board a Spanish ship while in American waters); Lauritzen, 345 U.S. at 592, 73 S.Ct. at 933 (applying Danish law to maritime tort that occurred in Cuban waters because all parties and the ship were Danish, even though the injured sailor embarked on the ship in New York).

Lauritzen employed something akin to the general, multi-factor approach described in section 6 of the Restatement, though without the initial presumption for applying the lex loci rule:

Maritime law, like our municipal law, has attempted to avoid or resolve conflicts between competing laws by ascertaining and valuing points of contact between the transaction and the states or governments whose competing laws are involved. The criteria, in general, appear to be arrived at from weighing of the significance of one or more connecting factors between the shipping transaction regulated and the national interest served by the assertion of authority.

Id. at 582, 73 S.Ct. at 928 (the relevant contacts included the place of the wrongful act, the law of the flag, the allegiance or domicile of the injured, the allegiance of the shipowner, the place of the contract, the inaccessibility of a foreign forum, and the law of the forum); accord Romero, 358 U.S. at 383, 79 S.Ct. at 486. In fact, Lau-ritzen gave the “law of the flag” the same presumptive weight that the Restatement gives to the place of injury. See 345 U.S. at 585-86, 73 S.Ct. at 929-30 (because the ship was registered in Denmark, Danish law “must prevail unless some heavy counterweight appears”).

At least two federal courts have used the contacts discussed in Lauritzen to resolve choice of law problems in air disasters on the high seas. In Noel v. Airponents, Inc., 169 F.Supp. 348 (D.N.J.1958), a case where a Venezuelan airliner crashed in the Atlantic Ocean, the district court held that federal maritime law would govern to impose liability on the U.S. corporation that had serviced the aircraft before departure. The court emphasized that “[t]he parties to the present litigation are American nationals, and neither the foreign carrier nor the Republic of Venezuela appear to have any interest in the outcome.” Id. at 350-51. In Air Crash Near Bombay, 531 F.Supp. 1175, a case where an Air India Boeing 747 crashed into the Arabian Sea shortly after takeoff, the district court held that Indian law would govern product liability claims against the American manufacturers of allegedly defective aircraft components. The court focused on “the essentially Indian character of this aircraft accident in Indian territorial waters on a flight originating in India involving an Indian air carrier and predominantly Indian victims.” Id. at 1191. This case presents a much sharper *1495conflict, however, because the contacts appear to be more evenly divided and do not point overwhelmingly to one jurisdiction or the other.

As KAL correctly points out, the question of punitive damages is of primary interest to jurisdictions having some link to the defendant. See In re Chicago Crash, 644 F.2d at 612-13 (decedents’ home jurisdictions only have an interest in recovery of compensatory damages); In re Air Florida Crash, 559 F.Supp. at 352-55. Assuming this is not a Klaxon thicket, the choice is between recovery of punitive damages under general federal maritime law versus non-recovery under Korean civil law. See In re Air Florida Crash, 559 F.Supp. at 347 & n. 19, 351. The balancing of competing interests is no easy task, and even trying to decipher what those interests are is an undertaking fraught with difficulties. Even assuming that Korea made a conscious decision to protect home businesses from excessive judgments, an interest it deemed more important than the added increment of deterrence from punitive damages, such a trade-off for purposes of internal Korean law does not mean that other interested jurisdictions cannot choose to apply punitive damages if they have a countervailing interest in punishment and deterrence that is involved in the controversy. In this case, the United States has an interest in both punishing egregious conduct that begins in this country and deterring its future recurrence. The United States’ interest in deterring willful misconduct on international flights linked to its borders might well outweigh Korea’s more parochial interest in protecting home businesses. Cf. Bernhard v. Harrah’s Club, 16 Cal.3d 313, 128 Cal.Rptr. 215, 546 P.2d 719 (applying California dram shop law to accidents caused by Nevada taverns even though Nevada had intentionally immunized its taverns from such liability), cert. denied, 429 U.S. 859, 97 S.Ct. 159, 50 L.Ed.2d 136 (1976).

I would therefore remand the punitive damages question to the district court with instructions to engage in a proper choice of law analysis.

*1496[[Image here]]