Baker v. Exxon Mobile Corp.

PER CURIAM Opinion; Dissent by Judge BROWNING.

PER CURIAM.

I. INTRODUCTION

We look for the third time at the punitive damages imposed in this litigation as a result of the 1989 grounding of the oil tanker Exxon Valdez, and the resulting economic harm to many who earned their livelihood from the resources of that area. See Baker v. Hazelwood (In re the Exxon Valdez), 270 F.3d 1215 (9th Cir.2001)[hereinafter Punitive Damages Opinion I]; Sea Hawk Seafoods, Inc. v. Exxon Corp., No. 03-35166 (9th Cir., Aug. 18, 2003). We are precluded, as the jury was, from punishing Exxon for befouling the beautiful region where the oil was spilled, because that punishment has already been imposed in separate litigation that has been settled. See Punitive Damages Opinion I, 270 F.3d at 1242. As we explained in Punitive Damages Opinion I, the plaintiffs’ punitive damages case was saved from preemption and res judicata because the award “vindicates only private economic and quasi-economic interests, not the public interest in punishing harm to the environment.” Id. “The plaintiffs’ claims for punitive damages expressly excluded consideration of harm to the environment.” In re the Exxon Valdez, 296 F.Supp.2d 1071, 1090 (D.Alaska 2004).

The resolution of punitive damages has been delayed because the course of this litigation has paralleled the course followed by the Supreme Court when, in 1991, it embarked on a series of decisions outlining the relationship of punitive damages to the principles of due process embodied in our Constitution. See, e.g., Pac. Mut. Life Ins. Co. v. Haslip, 499 U.S. 1, 111 S.Ct. 1032, 113 L.Ed.2d 1 (1991); TXO Prod. Corp. v. Alliance Res. Corp., 509 U.S. 443, 113 S.Ct. 2711, 125 L.Ed.2d 366 (1993) (plurality); BMW of N. Am., Inc. v. Gore, 517 U.S. 559, 116 S.Ct. 1589, 134 L.Ed.2d 809 (1996); State Farm Mut. Auto. Ins. Co. v. Campbell, 538 U.S. 408, 123 S.Ct. 1513, 155 L.Ed.2d 585 (2003). Intervening Supreme Court decisions have caused us to remand the matter twice to the district court for reconsideration of punitives in light of evolving Supreme Court law. The district court’s opinion, after our last remand for it to consider the impact of the Supreme Court’s decision in State Farm, is published at In re the Exxon Valdez, 296 F.Supp.2d 1071 (D.Alaska 2004) [hereinafter District Court Opinion]. It is the subject of this appeal.

*602Now, with the guidance of the Supreme Court’s decisions, the district judge’s thoughtful consideration of the issues, and our own prior decisions in the litigation, we trust we are able to bring this phase of the litigation to an end. While we agree with much of the analysis of the district court, we are required to review de novo the district court’s legal analysis in applying the Supreme Court’s guideposts. See Cooper Indus., Inc. v. Leatherman Tool Group, Inc., 532 U.S. 424, 436, 121 S.Ct. 1678, 149 L.Ed.2d 674 (2001).

While the original punitive damages award was $5 billion and in accord with the jury’s verdict, the district court reduced it to $4 billion after our first remand. In re the Exxon Valdez, 236 F.Supp.2d 1043, 1068 (D.Alaska 2002), vacated by Sea Hawk, No. 03-35166. Then, after our second remand, it entered an award of $4.5 billion. District Court Opinion, 296 F.Supp.2d at 1110. For the reasons outlined further in the factual development and the analysis of this opinion, we conclude that the ratio of punitive damages to actual economic harm resulting from the spill, reflected in the district court’s award of $4.5 billion, exceeds by a material factor a ratio that would be appropriate under Punitive Damages Opinion I and the current controlling Supreme Court analysis. See State Farm, 538 U.S. at 425, 123 S.Ct. 1513. We order a remittitur of $2 billion, resulting in punitive damages of $2.5 billion. We do so because, in assessing the reprehensibility of Exxon’s misconduct, the most important guidepost according to the Supreme Court’s opinion in State Farm, there are several mitigating facts. See id. at 419, 123 S.Ct. 1513. These include prompt action taken by Exxon both to clean up the oil and to compensate the plaintiffs for economic losses. These mollify, at least to some material degree, the reprehensibility in economic terms of Exxon’s original misconduct. Punitive Damages Opinion I, 270 F.3d at 1242. In addition, in considering the relationship between the size of the award and the amount of harm, we concluded in our earlier punitive damages opinion that the substantial costs that Exxon had already borne in clean up and loss of cargo lessen the need for deterrence in the future. Id. at 1244. We disagree, however, with Exxon’s ultimate contention that, as a result of two sentences in Punitive Damages Opinion I, written five years ago and before the Supreme Court’s opinion in State Farm, Exxon is entitled to have punitive damages assessed at no higher than $25 million. See id.

Our dissenting colleague goes to the other extreme. Exxon’s misconduct was placing a relapsed alcoholic in charge of a supertanker. Punitive Damages Opinion I, 270 F.3d at 1234. Yet, the dissent claims that we should ignore our unanimous conclusion in Punitive Damages Opinion I, 270 F.3d at 1242, that Exxon’s conduct with respect to the spill was not intentional. The dissent effectively treats Exxon as though it calculatingly and maliciously steered the ship into disaster. Purporting to rely on the intervening Supreme Court decision in State Farm, the dissent also refuses to apply our earlier holding that Exxon’s mitigation efforts reduce the reprehensibility of its conduct. This amounts to a rejection of the bedrock principle of stare decisis.

State Farm was an insurance contract case. Nothing in it suggests that this court’s decision in Punitive Damages Opinion I was improper. The Supreme Court did not explicitly or implicitly hold that mitigation plays no role in determining the constitutionality of a punitive damages award. Such a lack of discussion in an insurance contract case cannot supplant our express holding in the toxic-tort arena that mitigation efforts are a factor in as*603sessing the punitive damages award in this case. Controlling authority should not be ignored or distorted. As Learned Hand famously once said, “a victory gained by sweeping the chess pieces off the table is not enduring.” Learned Hand, Mr. Justice Cardozo, 52 Harv. L. Rev. 361, 362 (1939).

We reiterate our previous holding that Exxon’s conduct was not willful. Accordingly, a punitive damages award that corresponds with the highest degree of reprehensibility does not comport with due process when Exxon’s conduct falls squarely in the middle of a fault continuum.

Because the history of this litigation tracks the recent jurisprudential history of punitive damages, our analysis is best made in light of a thorough understanding of that history. We therefore outline that history with what we hope is sufficient clarity and thoroughness.

II. LEGAL AND FACTUAL BACKGROUND

A. From the Time of the Accident through the First Punitive Damages Award and Denial of Motion for New Trial: The Common Law through the Supreme Court Decision in TXO.

The Exxon Valdez ran aground on Bligh Reef in Alaska’s Prince William Sound on March 24, 1989. Punitive damages at that time were governed by general common law principles. At common law, the jury determined the punitives, and the trial judge conducted a limited review to determine whether the jury’s verdict was the product of passion and prejudice, or whether the award was one that shocked the conscience. See Renee B. Lettow, New Trial for Verdict Against Law: Judge-Jury Relations in Early Nineteenth Century America, 71 Notre Dame L.Rev. 505, 542-51 (1996); Paul DeCamp, Beyond State Farm: Due Process Constraints on Noneconomic Compensatory Damages, 27 Harv. J.L. & Pub. Pol’y 231, 246-48 (2003); see also Browning-Ferris Indus. of Vt., Inc. v. Kelco Disposal, Inc., 492 U.S. 257, 278 n. 24, 109 S.Ct. 2909, 106 L.Ed.2d 219 (1989) (affirming district court’s application of Vermont’s “grossly and manifestly excessive” standard for judicial review); Honda Motor Co. v. Oberg, 512 U.S. 415, 432 n. 10, 114 S.Ct. 2331, 129 L.Ed.2d 336 (1994). Although there were cases dating from the Lochner era that had suggested that there may be a due process ceiling on punitive damages, at the time of this accident in 1989, the Supreme Court had never invalidated an award on grounds that the size of the award violated due process. See BMW v. Gore, 517 U.S. at 600-01, 116 S.Ct. 1589 (Scalia, J., dissenting) (discussing the history of due process review of punitive damages awards) (citing Seaboard Air Line Ry. v. Seegers, 207 U.S. 73, 78, 28 S.Ct. 28, 52 L.Ed. 108 (1907); Southwestern Tel. & Tel. Co. v. Danaher, 238 U.S. 482, 489-91, 35 S.Ct. 886, 59 L.Ed. 1419 (1915); Waters-Pierce Oil Co. v. Texas, 212 U.S. 86, 111-12, 29 S.Ct. 220, 53 L.Ed. 417 (1909); Standard Oil Co. of Ind. v. Missouri, 224 U.S. 270, 286, 290, 32 S.Ct. 406, 56 L.Ed. 760 (1912); St. Louis, I.M. & S.R. Co. v. Williams, 251 U.S. 63, 66-67, 40 S.Ct. 71, 64 L.Ed. 139 (1919)).

In 1991, however, the Supreme Court decided Pacific Mutual Life Insurance Co. v. Haslip, 499 U.S. 1, 111 S.Ct. 1032, 113 L.Ed.2d 1 (1991). There, for the first time in the modern era, the Court conducted a substantive review of an award of punitive damages. Haslip was an insurance fraud case, in which the agent pocketed the premiums and caused the plaintiffs insurance to lapse. Id. at 4-5, 111 S.Ct. 1032. The Court upheld a punitive damages award *604that amounted to four times the award of compensatory damages and 200 times the out-of-pocket costs of the defrauded insured. Id. at 23-24, 111 S.Ct. 1032. The Court noted that the ratios might be “close to the line,” but said the award had to be upheld because it “did not lack objective criteria.” Id. The Court therefore concluded that the punitive damages did not “cross the line into the area of constitutional impropriety.” Id. The Supreme Court did not, at that time, and has not since, defined any bright line of constitutional impropriety. It has, repeatedly, indicated that there is none. See, e.g., State Farm, 538 U.S. at 424-25, 123 S.Ct. 1513.

In 1993, two years after Haslip, the Court took on another major punitive damages case. In TXO Production Corp. v. Alliance Resources Corp., 509 U.S. 443, 113 S.Ct. 2711, 125 L.Ed.2d 366 (1993), the Court reviewed a jury award of $19,000 in compensatory damages and $10 million in punitive damages. Id. at 451, 113 S.Ct. 2711. That case arose out of an oil and gas development fraud scheme. Id. at 447-51, 113 S.Ct. 2711. The case produced no majority opinion. The plurality, reiterating that due process places some limit on punitive damages, said that the award was not so “grossly excessive” that it should be overturned, thus invoking the standard used in Haslip. Id. at 462, 113 S.Ct. 2711. The Court declined to provide any particular guidance in determining when an award would be “grossly excessive.” Id. The plurality chose instead to say that the dramatic disparity between the actual financial loss and the punitive award was not controlling. Id. The award was upheld. Id.

It was against this background that the jury in this case was instructed in 1994. The jury was told to take into account the reprehensibility of the misconduct, the amount of actual or potential harm arising from the misconduct, and, additionally, to take into account mitigating factors such as the clean up costs and fines already imposed as deterrents. District Court Opinion, 296 F.Supp.2d at 1081-82. The instructions were the product of mutual effort of the parties and the district court, and have not been seriously challenged. Id. They are not questioned here and were, in retrospect, quite forward looking.

On September 16, 1994, the jury returned a $5 billion punitive damages verdict, having some time earlier imposed a compensatory award of $287 million. The district court accepted the punitive award and entered judgment. Citing Haslip and TXO, the district court denied Exxon’s motion for a new trial in January of 1995.

B. The Appeal of the Damage Allocation Plan and Our Decisions in Baker and Icicle.

Prior to trial, several plaintiffs, many of the sea food processors, had entered into settlement agreements with Exxon. Icicle Seafoods, Inc. v. Baker (In re the Exxon Valdez), 229 F.3d 790, 792 (9th Cir.2000) [hereinafter Icicle]; Baker v. Exxon Corp. (In re the Exxon Valdez), 239 F.3d 985, 986 (9th Cir.2001) [hereinafter Baker], The agreements anticipated a sizable punitive damages award. See Icicle, 229 F.3d at 793; Baker, 239 F.3d at 986-87. In return for receiving substantial millions in payments from Exxon, the settling plaintiffs, in two separate agreements, agreed to allocate a portion of their punitive award to Exxon. One agreement was a so called “cede back agreement,” Icicle, 229 F.3d at 793, and the other was an assignment of the future award, Baker, 239 F.3d at 986-87.

The district court, however, did not know of the agreements during trial. Icicle, 229 F.3d at 793. When the court did learn of them, during consideration of the *605parties’ proposed damage allocation plan, and after the punitives had been imposed in accordance with the jury’s verdict, the district court frowned on the settlements. Id. at 794. In the district court’s view, Exxon should have told the jury about the agreements so that the jury would have known how much Exxon was actually going to have to pay in punitive damages. Id. The district court, therefore, refused to permit the settling plaintiffs to receive any of the punitive damages award, on the theory that Exxon should not benefit from the settlements. Id.; Baker, 239 F.3d at 987. Exxon pursued two appeals from the district court’s refusal to enforce the agreements: one involving the cede back agreement, Icicle, 229 F.3d at 793, and the other involving the assignment agreement, Baker, 239 F.3d at 987-88.

The two different forms of agreement were intended to have essentially the same effect: allowing Exxon to keep some portion of the eventual punitive award in exchange for settling compensatory damage claims. In Icicle, this panel considered the cede back agreement. In a thorough opinion, we held that the cede back agreement was valid and enforceable and that the jury quite properly was not told of its existence. Icicle, 229 F.3d at 800. We reasoned that had the jury been told of the agreement, it might well have compensated for the settlement by imposing more damages. Id. at 798. This, in turn, would have frustrated the efforts of parties to reach settlements. We pointed out that settlements should be encouraged, particularly in large class actions like this one. Id. “Far from being unethical, cede back agreements make it easier to administer mandatory class actions for the assessment of punitive damages and encourage settlement in mass tort cases. As a result, such agreements should typically be enforced.” Id.

The second appeal, Baker, considered an assignment agreement. Baker, 239 F.3d at 987-88. Following the Icicle reasoning, this panel reached the same conclusion. Id. at 988.

C. The Supreme Court’s Decision in BMW v. Gore.

As the parties were beginning their preparation for the first appeal of the $5 billion punitive damages award, the Supreme Court issued its first major due process/punitive damages decision after TXO. In 1996, it decided BMW of North America, Inc. v. Gore, 517 U.S. 559, 116 S.Ct. 1589, 134 L.Ed.2d 809 (1996). This was the Supreme Court’s first attempt to describe specific factors that a court should consider in reviewing a jury’s award of punitive damages. See id. at 575, 116 S.Ct. 1589. The Court invoked the traditional concepts of due process to describe the purpose of the review as an assurance of fair notice to the defendant of the consequences of its conduct. Id. at 574, 116 S.Ct. 1589.

The Court described three factors to be considered. Id. at 575, 116 S.Ct. 1589. The first was the reprehensibility of the conduct. Id. The Court explained that reprehensibility is “[p]erhaps the most important indicium of the reasonableness of a punitive damages award,” and said that an award should reflect “the enormity” of the offense. Id. (citations omitted).

The second factor was the disparity between the actual or potential harm to the plaintiffs flowing from that conduct, and the punitive damages assessed by the jury. The Court said that the disparity factor was the most commonly cited. Id. at 580, 116 S.Ct. 1589. The Court reasoned this factor is important because it “has a long pedigree” extending back to English statutes from 1275 to 1753 providing for double, treble or quadruple damages. Id. at *606580-81, 116 S.Ct. 1589. Thus the critical measure here is the ratio between the punitive award and the amount of harm inflicted on the plaintiff, or plaintiffs, before the court.

The third factor was the difference between the punitives and the civil and criminal penalties authorized by the state for that conduct. Id. at 583, 116 S.Ct. 1589. The Court indicated that reviewing courts should use this factor to “accord substantial deference to legislative judgments concerning appropriate sanctions for the conduct at issue.” Id. at 583, 116 S.Ct. 1589 (internal quotations omitted).

In BMW v. Gore, the defendant had engaged in a practice of repainting damaged cars and passing them off as never-damaged cars with their original paint. Id. at 563-64, 116 S.Ct. 1589. The plaintiff who had purchased one of these cars was awarded $4,000 in compensatory damages and $4 million in punitives. Id. at 565, 116 S.Ct. 1589. The Alabama Supreme Court reduced the punitives to $2 million, and the defendant petitioned for certiorari review. Id. at 567, 116 S.Ct. 1589. The Supreme Court held the punitives were excessive. Id. at 585, 116 S.Ct. 1589.

In examining the reprehensibility of the conduct, the Supreme Court in BMW v. Gore stressed that the only harm inflicted by the defendant was economic and not physical. Id. at 576, 116 S.Ct. 1589. The Court also emphasized that the conduct to be considered was only the conduct of the defendant towards the plaintiff in the Alabama case and not other conduct that might be a part of a nationwide practice. Id. at 572, 116 S.Ct. 1589. Justice Breyer’s concurring opinion noted the danger in subjecting a defendant to punishment multiple times for the same conduct. Id. at 593, 116 S.Ct. 1589 (Breyer, J., concurring).

Thus, in looking at the ratio between the punitives and the harm, and in stressing that the ratio must be a reasonable one, the Court was holding that the ratio must be measured by the ratio of punitive damages to the harm suffered by the plaintiff in that case, without regard to harm that might have been experienced by others and for which the defendant might also be responsible. Id. at 580, 116 S.Ct. 1589. It concluded that a ratio of 500 to 1 was grossly excessive. Id. at 583, 116 S.Ct. 1589. Such an excessive ratio resulted from the jury’s improperly measuring the punitives in relation to the damage inflicted on a nation of potential plaintiffs rather than the damage to the plaintiff before that jury. Id. at 573, 116 S.Ct. 1589.

With respect to the third factor, the relationship between the punitive damages and the comparable penalties under state law, BMW v. Gore looked to the Court’s federalism jurisprudence. The Court’s opinion stressed that reviewing courts should be mindful of the need to pay due deference to the legislative judgments of states in assessing the reprehensibility of conduct. Id. at 583, 116 S.Ct. 1589 (“[A] reviewing court engaged in determining whether an award of punitive damages is excessive should ‘accord “substantial deference” to legislative judgments concerning appropriate sanctions for the conduct at issue.’ ”) (quoting Browning-Ferris, 492 U.S. at 301, 109 S.Ct. 2909 (O’Connor, J., concurring in part, dissenting in part)).

Again refusing to draw any kind of mathematical bright line between acceptable and unacceptable ratios, the Court described the 500 to 1 ratio in BMW v. Gore as “breathtaking.” Id. It remanded for further, not inconsistent, proceedings, because, unlike Haslip, where the Court affirmed a questionable award, the Court in BMW was “fully convinced” that this *607award was “grossly excessive.” Id. at 585-86, 116 S.Ct. 1589.

D. The First Punitive Damages Appeal.

It was against this background that briefing in the first appeal of the original $5 billion punitive damages award in this case went forward. Exxon contended the amount of the award violated due process principles, as described in BMW v. Gore. Punitive Damages Opinion I, 270 F.3d at 1241. The district court had not had an opportunity to review BMW v. Gore before its original judgment became final and ap-pealable upon denial of Exxon’s motion for a new trial. Id.

In its appeal from the $5 billion award, Exxon, in addition to challenging the amount of the punitive damages, challenged the sufficiency of the evidence supporting punitive damages; the jury instructions; the allowability of any punitive damages as a matter of public policy, maritime law and res judicata; and the preemption of punitive damages by other federal law. Needless to say, briefing was extensive. After appellate proceedings were stayed from January 1998 to September 1998 for the parties to pursue a limited remand, this panel heard argument in May of 1999.

While the case was under submission, the Supreme Court granted certiorari in another Ninth Circuit case, and in May 2001, decided Cooper v. Leatherman Tool Group. The Court there held our review of punitive damages was to be de novo. Cooper, 532 U.S. at 436, 121 S.Ct. 1678. This did not ease our task.

E. Punitive Damages Opinion I.

We issued our first opinion on punitives damages in November, 2001. Our opinion went in detail through the facts of the disaster and the conduct of Exxon, and of Captain Hazelwood, because they bore so heavily on the consideration of the issues on appeal. Punitive Damages Opinion I, 270 F.3d at 1221-24. In an opinion of more than 40 pages, we rejected Captain Hazelwood’s separate appeal, and dealt at some length with all of the issues raised by Exxon. We ultimately rejected all of them except the challenge to the amount of punitive damages. Id. at 1254.

Referring to the “unique body of law” that governs punitive damages, we focused on the two Supreme Court opinions that had been decided after the district court’s decision in the case, and we termed them “critical.” Id. at 1239. These were BMW v. Gore and Cooper v. Leatherman Tool Group. We said:

In BMW, the Supreme Court held that a punitive damage award violated the Due Process Clause of the Fourteenth Amendment because it was so grossly excessive that the defendant lacked fair notice that it would be imposed. Dr. Gore’s car was damaged in transit, and BMW repainted it but did not tell Dr. Gore about the repainting when it sold him the car. The jury found that to be fraudulent, and awarded $4,000 in compensatory damages for reduced value of the car and $4 million in punitive damages. The Alabama Supreme Court cut the award to $2 million, but the Court held that it was still so high as to deny BMW due process of law for lack of notice, because the award exceeded the amounts justified under the three “guideposts.” The BMW guideposts are: (1) the degree of reprehensibility of the person’s conduct; (2) the disparity between the harm or potential harm suffered by the victim and his punitive damage award; and (3) the difference between the punitive damage award and the civil penalties authorized or imposed in comparable cases. We apply these *608three guideposts to evaluate whether a defendant lacked fair notice of the severity of a punitive damages award, and to stabilize the law by assuring the uniform treatment of similarly situated persons.

Id. at 1240-41 (internal quotations omitted). We noted that in Cooper v. Leatherman Tool Group the Supreme Court decided that “considerations of institutional competence” require de novo review of punitive damages awards. Id. at 1240 (quoting Cooper, 532 U.S. at 440, 121 S.Ct. 1678).

We went on to observe that the district court had not reviewed the award under the standards announced in those cases because neither case had been decided by the time the jury returned its verdict, and Exxon had never challenged the amount of the award on constitutional grounds until after the jury’s verdict. Id. at 1241. In view of the need for de novo review and the intervening decisions of BMW v. Gore and Cooper v. Leatherman Tool Group, we remanded for reconsideration of punitive damages. Id. We also provided some observations on possible alternative analyses of punitive damages under the BMW v. Gore factors. Id. at 1241-46.

These observations began with the factor of reprehensibility, quoting the Supreme Court’s admonition in BMW v. Gore that it is “[pjerhaps the most important indicum of the reasonableness of a punitive damage award.” Id. at 1241. We pointed to the Court’s analogy to criminal cases, and its statement that nonviolent crimes are less reprehensible than violent ones. Id. We drew an analogy to the facts of this case, where Exxon’s conduct was reckless, but there was no intentional spilling of oil “as in a midnight dumping case.” Id. at 1242. We agreed with the plaintiffs that Exxon’s conduct was reprehensible in that it knew of the risk of an oil spill in transporting huge quantities of oil through the Sound, and it knew Hazelwood was a relapsed alcoholic. Id. at 1242. We observed, however, that such reprehensibility went more to justify punitive damages than to justify such a high amount. Id. We noted some mitigating factors, including prompt ameliorative action and the millions spent in clean up. Id.

We then turned to the ratio of actual harm caused by the misconduct to punitive damages awarded. Id. at 1243. Again analyzing BMW v. Gore, we said that it was difficult to determine what we called the “numerator,” that is, the value of the harm caused by the spill. Id. We used the jury award of $287 million in compensatory damages as one possible numerator and also, as alternative numerators, the district court’s estimates of harm, which at that time ranged from $290 million to $418 million. Id. We noted that if compensatory liability were used, any amounts Exxon had voluntarily paid in settlements should not be taken into account. We said that

[t]he amount that a defendant voluntarily pays before judgment should generally not be used as part of the numerator, because that would deter settlements prior to judgment. “[T]he general policy of federal courts to promote settlement before trial is even stronger in the context of large scale class actions.”

Id. at 1244 (citing Icicle, 229 F.3d at 795; Baker, 239 F.3d at 988).

As a final observation on the relationship between the punitive damages award and the harm, we pointed out that the substantial clean up costs and other losses to Exxon from the oil spill had already had considerable deterrent effect. We indicated such deterrence should, depending on the circumstances, call for a lower, rather than a higher ratio. Id.

Turning to the third BMW v. Gore factor, we observed that the nature of criminal fines, which are potential state and *609federal penalties, might be useful in reviewing punitives. Id. at 1245. We observed that “[cjriminal fines are particularly informative because punitive damages are quasicriminal.” Id. We then looked to the general federal statutory measure for fines and discussed a number of alternative guideposts. Id. We noted the federal fines could range from $200,000 to $1.08 billion. Id. We looked as well at the ceiling of civil liability under the Trans-Alaska Pipeline Act and noted it was $100 million in strict liability for anyone who spills oil from the pipeline. Id.

In addition to those possible penalties, we looked at the actual penal evaluation made in the case by the Attorneys General of the United States and of the state of Alaska. Id. at 1245-46. Agreeing with the district court that they did not establish a limit, we noted that they did represent an adversarial judgment, by executive officers, of an appropriate level of punishment. Id. at 1246. Finally, without necessarily exhausting available analogies in the penalty field, we noted that Congress had subsequently amended the statute to increase the amount of civil penalties for grossly negligent conduct, and that the maximum penalty here under the new federal statute would be a maximum of $786 million. Id. The federal penalties are based upon the number of barrels of oil spilled. 33 U.S.C. § 1321(b)(7).

In suggesting various possible guidelines to assess whether the $5 billion was “grossly excessive” we did not imply that any single guidepost would be controlling. Concluding that the $5 billion was too high to withstand the review we were required to give it under BMW v. Gore and Cooper v. Leatherman Tool Group, and noting that those cases came down after the district court had ruled, we remanded for it to apply the due process analysis required under those decisions, with what we hoped would be helpful guidance from our opinion. Id. at 1241. No district court analysis of BMW v. Gore was before us and we thus could not have decided any specific issue arising from any such analysis arising from its guideposts. Id. We offered only guidance culled from what was then controlling Supreme Court precedent and general principles applicable to the calculation of damage liability. Id.

F. The District Court Opinion on our First Remand.

The district court again did an extensive analysis of the relative reprehensibility of Exxon’s misconduct and of the harm it caused. In re the Exxon Valdez, 236 F.Supp.2d at 1054-60. Though noting that an accurate assessment of the full extent of the plaintiffs’ actual harm was impossible, the district court attempted to reconstruct that harm by adding together the jury’s compensatory damages verdict of $287 million, judgments in related cases, as well as payments and settlements made to plaintiffs before and during the punitive damages litigation. Id. at 1058-60. The district court concluded that the actual harm was just over $500 million. Id. at 1060. The district court also concluded that the circumstances of this case justified a ratio of punitive damages to harm of 10 to 1. Id. at 1065. This calculation would have supported the original $5 billion award. Id. The district court nevertheless reduced the punitive damages to $4 billion, to conform to what it viewed as our mandate. Id. at 1068.

G. The Second Appeal, the Supreme Court’s Opinion in State Farm, and our Second Remand.

Not surprisingly, Exxon appealed again. And, not surprisingly, the Supreme Court issued an opinion in still another punitive damages case while the appeal was pend*610ing. State Farm Mut. Auto. Ins. Co. v. Campbell, 538 U.S. 408, 123 S.Ct. 1513, 155 L.Ed.2d 585 (2003).

The plaintiffs in State Farm, the Camp-bells, were involved in a head-on collision and sued their automobile insurer, State Farm, for bad faith. Id. at 413, 123 S.Ct. 1513. The claim was based on State Farm’s rejection of an offer to settle the Campbells’ claims at the policy limit, State Farm’s assurances to them that they had no liability for the accident, State Farm’s resulting decision to take the case to court despite the substantial likelihood of an excess judgment, and its subsequent refusal to pay an adverse judgment over three times the policy limits. Id. at 413-14, 123 S.Ct. 1513. The case was similar to BMW v. Gore in that there were only two plaintiffs before the jury. Id. Nevertheless, as in BMW v. Gore, the jury was allowed to consider the effects of similar but unrelated misconduct on many potential plaintiffs who were not before the court. Id. at 415, 123 S.Ct. 1513. Final judgment after appeal to the Utah Supreme Court was for $1 million in compensatory and $145 million in punitive damages. Id. at 412, 123 S.Ct. 1513. The United States Supreme Court remanded for the Utah courts to reduce the award. Id. at 429, 123 S.Ct. 1513.

The Supreme Court in State Farm once again emphasized that the “most important indicium” of a punitive damages award’s reasonableness is the relative reprehensibility of the defendant’s conduct. Id. at 419, 123 S.Ct. 1513; see also BMW v. Gore, 517 U.S. at 575, 116 S.Ct. 1589. Yet State Farm significantly refined the reprehensibility analysis by instructing courts to weigh five specific considerations: (1) whether the harm caused was physical as opposed to economic; (2) whether the conduct causing the plaintiffs harm showed “indifference to or a reckless disregard of the health or safety of others;” (3) whether the “target of the conduct” was financially vulnerable; (4) whether the defendant’s conduct involved repeated actions as opposed to an isolated incident; and (5) whether the harm caused was the result of “intentional malice, trickery, or deceit, or mere accident.” 538 U.S. at 419, 123 S.Ct. 1513. The Court did not rank these factors. It did explain, however, that only one factor weighing in a plaintiffs favor may not be sufficient to support a punitive damages award, and the absence of all factors makes any such award “suspect.” Id.

As to BMW v. Gore’s second guidepost, the ratio between harm or potential harm to the plaintiff and the punitive damages award, the Court “decline[d] again to impose a brightline ratio which a punitive damages award cannot exceed.” Id. at 425, 123 S.Ct. 1513. But it provided some sharper guidance than it had in previous cases.

First, it indicated that ratios in excess of single-digits would raise serious constitutional questions, and that single-digit ratios were “more likely to comport with due process.” Id. fact, despite the Court’s disclaimer that “there are no rigid benchmarks that a punitive damages award may not surpass,” the Court strongly indicated the proportion of punitive damages to harm could generally not exceed a ratio of 9 to 1. Id. at 425, 123 S.Ct. 1513 (“[F]ew awards exceeding a single-digit ratio between punitive and compensatory damages, to a significant degree, will satisfy due process.”).

Second, the Court discussed particular combinations of factors that would justify relatively higher or lower ratios. For example, where a “particularly egregious act has resulted in only a small amount of economic damages” or where “the injury is hard to detect or the monetary value of *611the noneconomic harm might have been difficult to determine,” ratios in the high single-digits and perhaps even higher might be warranted. Id. (quoting BMW v. Gore, 517 U.S. at 582, 116 S.Ct. 1589). Conversely, “[w]hen compensatory damages are substantial, then a lesser ratio, perhaps only equal to compensatory damages, can reach the outermost limit of the due process guarantee.” Id.

Finally, the Court minimized the relevance of criminal penalties as a guide, saying that they were not particularly helpful in determining fair notice. Id. at 428, 123 S.Ct. 1513. Indeed, the Court did not analyze State Farm’s potential criminal penalty at all, characterizing it as a “remote possibility.” Id. As to civil penalties, the Court noted only that the $145 million punitive damages award “dwarfed” the $10,000 maximum applicable fine. Id.

The Supreme Court’s opinion in State Farm was filed in 2003, after the district court, on our first remand, had already reviewed the punitive damages award. Because the district court performed its review without the benefit of the more focused guidance provided by the Court in State Farm, we remanded the second appeal summarily for the district court to reconsider the punitive damages award in light of State Farm. Sea Hawk, No. 03-39166.

H. The District Court Opinion on our Third Remand and this Appeal.

On remand for the third time, the district court, in an assessment similar to that in its opinion after our first remand, calculated plaintiffs’ harm at $513.1 million. District Court Opinion, 296 F.Supp.2d at 1103. Interpreting State Farm as holding that “single-digit multipliers pass constitutional muster for highly reprehensible conduct,” and citing our decision in Zhang v. American Gem Seafoods, Inc., 339 F.3d 1020 (9th Cir.2003), the district court decided to increase punitives from $4 billion to $4.5 billion. 296 F.Supp.2d at 1110. The final punitive damages award represented a ratio of just under 9 to 1. Id.

Once again, Exxon appealed. The plaintiffs also appealed, seeking to reinstate the jury’s full $5 billion punitive damages verdict.

In this appeal, Exxon has focused intensively on the sentences in our earlier opinion where we noted that prejudgment payments generally should not be part of the “numerator” to avoid deterring pre-judgment settlements. Punitive Damages Opinion I, 270 F.3d at 1242. Exxon has argued strenuously in the district court and to us that all of its settlement and other pre-judgment compensatory payments to plaintiffs must be subtracted from the over $500 million amount of actual harm in the ratio of punitive damages we use to review the award pursuant to the BMW v. Gore/State Farm factors. This would reduce the harm to the relatively paltry figure of $20.3 million.

We recognized in Punitive Damages Opinion I that Exxon, soon after the spill, instituted a claims payment system that almost fully compensated plaintiffs for their economic losses and did so promptly. Id. We also recognized that Exxon’s prompt payment of compensatory damages should be a substantial mitigating factor in our review of punitives. Id.

In Exxon’s appeal, major issues therefore relate to how, after State Farm, to assess the reprehensibility of Exxon’s conduct and the effect of the mitigating factors. An important subsidiary issue is the extent to which we are bound to give literal effect to the sentences in our earlier opinion concerning subtracting the prejudgment payments from actual harm, even though State Farm suggests the miti*612gating factors should be taken into account differently. For the reasons more fully explained in this opinion, we do not accept the minimal bottom line figure urged by Exxon and properly rejected by the district court. We do, however, conclude there is merit to Exxon’s contention that punitives should be reduced.

In their cross appeal, plaintiffs seek a reinstatement of the original $5 billion punitive award. We do not fully adopt their position either because doing so would peg the ratio of punitive damages to harm at a level State Farm reserves only for the most egregious misconduct. There was no intentional infliction of harm in this case. In addition, because Exxon’s mitigating efforts after the accident diminish the relative reprehensibility of its original misconduct for purposes of reviewing punitive damages, such a high ratio is not warranted in this case.

III. ANALYSIS

A. Lessons From History.

The history of the experience of the Supreme Court with punitive damages over the last decade-and-a-half reflects an evolutionary, not a revolutionary, course. In its first opinion in Haslip, the Court suggested that there might be a bright line of demarcation between punitive damages that comport with constitutional protections, and punitive damages that do not. Haslip, 499 U.S. at 23, 111 S.Ct. 1032. Although it did not say what “the line” would be, it termed ratios of punitive damages to compensatory damages of 4 to 1, and to out-of-pocket costs of 200 to 1, to be close to it. Id.

In subsequent cases, however, the Court expressly avoided a rigid mathematical formula or limit, while refining its ratio analysis, concluding in State Farm that a ratio of punitive damages to actual harm of less than 10 to 1 was more likely to comport with due process than an award with a higher ratio. State Farm, 538 U.S. at 425, 123 S.Ct. 1513. Along the way, the Court’s experience reflects efforts to comport with the tried and true concepts inherent in due process, i.e., those of notice and fairness. See, e.g., Mullane v. Cent. Hanover Bank & Trust Co., 339 U.S. 306, 70 S.Ct. 652, 94 L.Ed. 865 (1950); Int’l Shoe Co. v. Washington, 326 U.S. 310, 66 S.Ct. 154, 90 L.Ed. 95 (1945).

In State Farm, the Court expressly noted its concern that the jury had been allowed to take into account the effect of conduct that may have taken place nationwide on thousands of potential plaintiffs. State Farm, 538 U.S. at 422, 123 S.Ct. 1513. The unfairness of a defendant being hit with punitive damages many times for the same conduct was central to the Court’s analysis in remanding. Id. The Court explained, “[pjunishment on these bases creates the possibility of multiple punitive damages awards for the same conduct; for in the usual case non-parties are not bound by the judgment some other plaintiff obtains.” Id. at 423, 123 S.Ct. 1513.

Indeed, in State Farm, the Court stressed that the most important factor is the reprehensibility of the particular conduct in the case. State Farm, 538 U.S. at 419, 123 S.Ct. 1513. This is because, in assessing the foreseeability of the possible effects of the defendant’s conduct as it might bear on punitive damages, the reviewing court is in reality dealing with the traditional concept of the need for fair notice of the possible legal consequences of one’s misconduct. Id. at 417, 123 S.Ct. 1513.

Perhaps because such traditional elements of due process are flexible, the Supreme Court has not often taken on the task of reviewing the amount of punitive *613damages and has, in fact, overturned only-two punitive awards because of their size. Each of them exceeded by a multiple of more than 100 the amount of compensatory payments necessary to compensate a plaintiff for the actual harm caused by the defendant’s misconduct. BMW v. Gore, 517 U.S. at 582, 116 S.Ct. 1589 (striking down a 500:1 ratio); State Farm, 538 U.S. at 429, 123 S.Ct. 1513 (striking down a 145:1 ratio).

B. BMW v. Gore/State Farm Guideposts.

BMW v. Gore identified three guideposts for reviewing punitive damages, and State Farm added important refinements. The guideposts are (1) the reprehensibility of the defendant’s misconduct, (2) the ratio of punitives to harm, and (3) comparable statutory penalties. They need not be rigidly or exclusively applied, for we agree with our sister circuit that “[t]hese guideposts should not be taken as an analytical straight jacket.” Zimmerman v. Direct Federal Credit Union, 262 F.3d 70, 81 (1st Cir.2001). We must, nevertheless, examine them in the context of this case.

1. Reprehensibility.

The most important guidepost is the reprehensibility of Exxon’s misconduct. State Farm, 538 U.S. at 419, 123 S.Ct. 1513 (quoting BMW v. Gore, 517 U.S. at 575, 116 S.Ct. 1589). In our prior opinion, we defined the relevant misconduct supporting punitive damages as Exxon’s keeping Hazelwood in command with knowledge of Hazelwood’s relapse into alcoholism. We said that “Exxon knew Ha-zelwood was an alcoholic, knew that he had failed to maintain his treatment regimen and had resumed drinking, knew that he was going on board to command its supertankers after drinking, yet let him continue to command the Exxon Valdez through the icy and treacherous waters of Prince William Sound.” Punitive Damages Opinion I, 270 F.3d at 1237-38. We see no need to reconsider this issue, despite Exxon’s invitation to do so.

To evaluate the reprehensibility of the misconduct, State Farm refers to five sub-factors: (1) the type of harm, (2) whether there was reckless disregard for health and safety of others, (3) whether there were financially vulnerable targets, (4) whether there was repeated misconduct and (5) whether it involved intentional malice, trickery, or deceit, rather than mere accident. State Farm, 538 U.S. at 419, 123 S.Ct. 1513.

We must also consider mitigating factors. In Punitive Damages Opinion I, in the context of this particular case, we looked to Exxon’s response to the catastrophe, including its prompt cleanup and compensatory payments. We held they were factors mitigating the reprehensibility of the original misconduct. Punitive Damages Opinion I, 270 F.3d at 1242. “Reprehensibility should be discounted if defendants act promptly and comprehensively to ameliorate any harm they cause in order to encourage such socially beneficial behavior.” Id.

The dissent takes issue with two components of our BMW v. Gore analysis. Its reasons, however, are surprising, because they contradict our unanimous holding in Punitive Damages Opinion I, 270 F.3d at 1242, that the spill was not intentional nor Exxon’s conduct malicious. See Dissent at 633 (characterizing Exxon’s conduct as “malicious”). Then, the dissent misapplies the Supreme Court’s mandate that we must perform an exacting appellate review to ensure that “an award of punitive damages is based upon an ‘application of law, rather than a decisionmaker’s caprice.” State Farm, 538 U.S. at 418, 123 S.Ct. *6141513 (citing BMW v. Gore, 517 U.S. at 587, 116 S.Ct. 1589).

First, the dissent maintains that the value of defendant’s pre-litigation mitigation efforts should not affect punitive damages because the Supreme Court did not explicitly provide for such a calculus in State Farm. Dissent at 628. Thus, the dissent would reject the principle of stare decisis and the law of the case and overturn our holding in Punitive Damages Opinion I, 270 F.3d at 1242, that Exxon’s voluntary compensation to the plaintiffs effectuated good public policy in making an injured party whole as quickly as possible. We are not prepared to question the soundness of our unanimous conclusion in Punitive Damages Opinion I merely because intervening Supreme Court jurisprudence in the insurance context did not address the issue. See State Farm, 538 U.S. 408, 123 S.Ct. 1513, 155 L.Ed.2d 585. By contrast here, we have already held that mitigation is both relevant and conscientious in the toxic-tort setting. It would be unwise in reviewing punitive damages to ignore the prompt steps of a defendant to take curative action in a mass tort case.

The dissent also claims that we improperly treat BMW’s fifth factor, the fault analysis, as a dichotomy with two mutually exclusive options: finding Exxon’s conduct intentional and thus grossly reprehensible, or finding it accidental and thus to a large degree excusable. Dissent at 630. This is not our analysis. We acknowledge that Exxon’s conduct was not intended to cause an oil spill, but neither was allowing a relapsed alcoholic to command a supertanker “mere accident.” Majority at 617. Exxon’s reckless malfeasance falls in the middle of a continuum between accidental and intentional conduct. Accordingly, the fifth subfactor of the reprehensibility analysis supports neither high nor low reprehensibility on the part of Exxon.

The Supreme Court has reserved the upper echelons of constitutional punitive damages (a 9 to 1 ratio) for conduct done with the most vile of intentions. Thus, an affirmance of the district court’s application of such a ratio in this case, where the defendant’s conduct was reckless but not intentional, would transgress the requisite constitutional boundaries as the Supreme Court has explained them to date.

We turn now to the specific State Farm reprehensibility subfactors. These demonstrate that a 5 to 1 ratio more appropriately comports with due process.

a. Type of Harm — Physical versus Economic.

To evaluate the type of harm, State Farm instructs us to consider whether “the harm was physical as opposed to economic,” because conduct producing physical harm is more reprehensible. State Farm, 538 U.S. at 419, 123 S.Ct. 1513. In this case the district court found that Exxon’s conduct caused no actual physical harm to people, but caused more than mere economic harm to them, because the economic effects of its misconduct produced severe emotional harm as well. We agree with the district court’s explanation that “the spilling of 11 million gallons of crude oil into Prince William Sound and Lower Cook Inlet disrupted the lives (and livelihood) of thousands of claimants for years.” District Court Opinion, 296 F.Supp.2d at 1094.

The Supreme Court has recognized conduct causing emotional as well as economic harm can be more reprehensible than conduct causing mere economic harm. See BMW v. Gore, 517 U.S. at 576 n. 24, 116 S.Ct. 1589. There it cited Blanchard v. Morris, 15 Ill. 35, 36 (1853), a case affirming a $700 punitive award against individuals who caused no physical harm and only *615$13 of economic harm, but used mental torture to extort it.

In Bains LLC v. Arco Products Co., 405 F.3d 764, 775 (9th Cir.2005), we held that “intentional, repeated ethnic harassment” increased the level of reprehensibility beyond the merely economic. See also Swinton v. Potomac Corp., 270 F.3d 794, 818 (9th Cir.2001). The gratuitous, intentional mental oppression of the victims made it “highly reprehensible conduct, though not threatening to life or limb.” Id. at 777. In Planned Parenthood v. American Coalition of Life Activists, 422 F.3d 949, 958 (9th Cir.2005), we held that a “true threat” increased reprehensibility even though it was not carried out, because the threat was intended to intimidate, and the economic component went beyond reducing the victim’s wealth or income to trying to drive the victims away from their practices of medicine. Our Planned Parenthood decision was consistent with BMW’s citation with approval of older decisions upholding awards based on the “mental fear, torture, and agony of mind” caused by the threat of violence. BMW, 517 U.S. at 575-76, n. 24, 116 S.Ct. 1589.

The district court concluded that the mental distress caused by the oil spill to the fishermen and property owners who were harmed economically justified a higher level of reprehensibility, and Exxon urges that emotional distress damages were not before the jury. Because our review must be de novo under Cooper Indus., Inc. v. Leatherman Tool Group, Inc., 532 U.S. 424, 436, 121 S.Ct. 1678, 149 L.Ed.2d 674 (2001), we are not bound by the district court’s rationale. The cases discussed above show that punitive damages can — and traditionally do — consider the effects of the tortfeasor’s conduct on the victim’s mentality, not just his pocketbook. On the other hand, they may not go so far, and we need not, as to justify punitive damages for accidentally causing mental distress. State Farm states that compensatory damages for mental distress generally include a punitive element, so including mental distress in punitive damages may be duplicative. 538 U.S. at 426, 123 S.Ct. 1513.

What comes to something near the same result in this case, though it would not in most cases, is the entirely foreseeable disruption to the way tens of thousands people live their lives if a giant oil tanker were to run aground and spill its cargo. When tens of thousands of people have to change the way they make their living, their mental distress is not comparable to a BMW owner, or even a large number of BMW owners, being distressed because their cars were scratched or dented during shipment and repaired without their knowledge. Anyone setting an oil tanker loose on the seas under command of a relapsed alcoholic has to know that he is imposing this massive risk. Though spilling the oil is an accident, putting the relapsed alcoholic in charge of the tanker is a deliberate act. The massive disruption of lives is entirely predictable when a giant oil tanker goes astray. Thus, Exxon’s reprehensibility goes considerably beyond the mere careless imposition of economic harm.

b. Reckless Disregard for Health and Safety of Others.

The second subfactor we consider in assessing reprehensibility is whether Exxon displayed a reckless disregard for the health and safety of others. State Farm, 538 U.S. at 418, 123 S.Ct. 1513. We conclude this subfactor also militates toward greater reprehensibility. When Exxon trusted an officer it knew was incompetent to command the Exxon Valdez through the treacherous waters of Prince William Sound, Exxon acted with reckless disre*616gard for the health and safety of all those in the vicinity.

The Exxon Valdez grounding created a grave risk of physical harm for the crew and those who had to come to its rescue. The district court found that something as simple as an electro-static discharge could have ignited the crude oil and incinerated everyone in the vicinity. District Court Opinion, 296 F.Supp.2d at 1095. We therefore agree with the district court that Exxon acted with reckless disregard of the health and safety of others when it put in command a person not competent to perform that role.

Exxon argues that State Farm requires us to ignore Exxon’s disregard of the potential harm to the crew and rescuers because they are not plaintiffs to this litigation. Exxon misreads State Farm. State Farm disapproved punishing defendants for conduct in other states in which it might be lawful. 538 U.S. at 421, 123 S.Ct. 1513. Likewise, we had held in White v. Ford Motor Company, before State Farm came down, that a jury’s punitive damages award based on extraterritorial conduct (plaintiffs lawyer had made a “send them a message” argument addressing nationwide conduct) violated principles of federalism established in BMW v. Gore. 312 F.3d 998, 1013-14 (9th Cir.2002). These cases do not prohibit consideration of the potential harm to individuals merely because they are not plaintiffs. See 538 U.S. at 420-22, 123 S.Ct. 1513. The lesson is that the award in the other litigation “should have been analyzed in the context of the reprehensibility guidepost only.” Id.; BMW v. Gore, 517 U.S. at 574 n. 21, 116 S.Ct. 1589. State Farm therefore holds it is appropriate to look at the risk to others in analyzing reprehensibility. State Farm, 538 U.S. at 427, 123 S.Ct. 1513.

State Farm does warn against considering dissimilar acts of the defendant, or what is described as acts “independent from the acts upon which liability was premised.” Id. at 422, 123 S.Ct. 1513. The Court explained this is because “[a] defendant should be punished for the conduct that harmed the plaintiff, not for being an unsavory individual or business.” Id. at 423, 123 S.Ct. 1513. Here, however, the conduct that threatened the safety of the crew and rescuers is the same conduct that harmed the plaintiffs, and is the conduct that underlies this punitive damages litigation: Exxon’s knowingly placing a relapsed alcoholic in charge of the Exxon Valdez. The prohibition in State Farm against considering dissimilar acts does not apply here because taking into account the potential harm to the crew and rescuers punishes Exxon for the same conduct that harmed the plaintiffs. We have made this point before. See, for example, Hangarter v. Provident Life and Accident Insurance Co., 373 F.3d 998, 1015 n. 11 (9th Cir.2004), where we analyzed company-wide policies in a single-plaintiff lawsuit and distinguished State Farm’s warning against considering dissimilar acts. We said “unlike in State Farm, a legally sufficient nexus existed between Defendant’s allegedly widespread corporate policies and the termination of [the plaintiffs] benefits.” Id.

Accordingly, where the same conduct risked harm to all, the risk to all can be considered as a factor in assessing reprehensibility. The district court did not err in recognizing that Exxon recklessly disregarded the physical safety of the crew and rescuers, and thereby increased the reprehensibility of its conduct in putting Hazel-wood in command.

c. Financially Vulnerable Targets.

The district court found Exxon’s conduct harmed financially vulnerable subsistence fishermen. District Court Opinion, 296 *617F.Supp.2d at 1095. Exxon does not dispute that subsistence fishermen were financially vulnerable or that its reckless actions harmed them. It does contend that this factor applies only in fraud cases when a defendant intentionally defrauds financially vulnerable targets, such as the sick or elderly. While we do not believe the subfactor is so limited, we agree there must be some kind of intentional aiming or targeting of the vulnerable that did not occur here.

The purpose of reprehensibility analysis is to determine “the enormity” of the offense, which “reflects the accepted view that some wrongs are more blameworthy than others.” BMW v. Gore, 517 U.S. at 575, 116 S.Ct. 1589. The notion of “targeting” connotes some element of intent to harm particular individuals or categories of individuals. See Planned Parenthood, 422 F.3d at 958-59 (holding plaintiffs were financially vulnerable because the defendants’ threats attempted to scare the plaintiffs into quitting the jobs on which the plaintiffs’ livelihoods depended). Exxon did not intentionally target subsistence fishermen.

We conclude in this case that this consideration does not materially affect our assessment of the reprehensibility of Exxon’s conduct.

d. Repeated Action.

The district court found that the conduct was repetitive because Exxon repeatedly allowed Hazelwood to command its supertankers for three years after it knew he had resumed drinking. District Court Opinion, 296 F.Supp.2d at 1096. As the district court observed, Exxon did so, even though Exxon was fully aware of the tremendous risk of harm that it entailed. Id. “Over and over again, Exxon did nothing to prevent Captain Hazelwood [from sailing] into and out of Prince William Sound with a full load of crude oil.” Id.

Exxon argues that the relevant conduct is the grounding, not the knowledge of Hazelwood’s incapacity to command. That is not consistent with our description of the relevant misconduct in Punitive Damages Opinion I as putting (and leaving) Captain Hazelwood in command. Punitive Damages Opinion I, 270 F.3d at 1237-38. The district court’s finding of repetitive misconduct was not clearly erroneous. Planned Parenthood, 422 F.3d at 954. It militates in favor of increased reprehensibility.

e. Intentional Malice or Mere Accident.

Putting Captain Hazelwood in command of the supertanker was knowing and reckless misconduct. We agree with the district court that this misconduct was not “mere accident.” District Court Opinion, 296 F.Supp.2d at 1096.

Exxon points out that relieving Hazel-wood of command would have denied Ha-zelwood an employment opportunity on the basis of alcoholism and theoretically subjected Exxon to a disability discrimination lawsuit. While Exxon’s concerns may have been appropriate considerations in its evaluation of the risk, they do not justify the dangers its decision created to the livelihoods of tens of thousands of individuals. Spilling the oil was an accident, but putting a relapsed alcoholic in charge of a supertanker was not. And anyone doing so would know they were imposing a tremendous risk on a tremendous number of people who could not do anything about it. Exxon’s knowing disregard of the interests of commercial fishermen, subsistence fishermen, fish processors, cannery workers, tenders, seafood brokers and others dependent on Prince William Sound for their *618livelihoods, cannot be regarded as merely accidental.

At the same time, we must acknowledge that Exxon acted with no intentional malice towards the plaintiffs. We have consistently treated intentional conduct as more reprehensible than other forms of conduct subject to punitive damages. See Zhang, 339 F.3d at 1043; Bains LLC v. Arco Products Co., 405 F.3d 764, 775 (9th Cir.2005); Southern Union Co. v. Southwest Gas Corp., 415 F.3d 1001, 1011 (9th Cir.2005). In this case, however, as we have already recognized, “as bad as the oil spill was, Exxon did not spill the oil on purpose.” Punitive Damages Opinion I, 270 F.3d at 1242-43. While the reprehensibility of Exxon’s conduct that produced economic harm to thousands of individuals is high, the conduct did not result in intentional damage to anyone. This subfactor thus militates against viewing Exxon’s misconduct as highly reprehensible. Id.

f. Mitigation of Reprehensibility.

In assessing reprehensibility, we must not only take into account the reprehensibility of the original misconduct, but we have held that we must also take into account what has been done to mitigate the harm that the misconduct caused. Punitive Damages Opinion I, 270 F.3d at 1242; see also Swinton, 270 F.3d at 814-15 (discussing weight and relevance of post-tort mitigation evidence). As we said in Punitive Damages Opinion I, mitigation is to be considered “in order to encourage such socially beneficial behavior.” Punitive Damges Opinion I, 270 F.3d at 1242. Here, Exxon instituted a system of voluntary payments to plaintiffs and it undertook prompt cleanup efforts. We agree with what we said before: “Exxon spent millions of dollars to compensate many people after the oil spill, thereby mitigating the harm to them and the reprehensibility of its conduct.” Id.

g. Evaluation of Reprehensibility.

Placing a relapsed alcoholic in control of a supertanker was highly reprehensible conduct. As a result, Exxon disrupted the lives of thousands of people who depend on Prince William Sound for their livelihoods, and endangered its own crew and their rescuers. Over the span of three years, Exxon could and should have relieved Captain Hazelwood of command of supertankers, but it did not do so. At the same time, however, Exxon did not act with malice toward plaintiffs or anyone else; Exxon did not intend to damage plaintiffs’ livelihoods or cause them the emotional grief that went with the economic loss.

Thus, Exxon’s conduct is in the higher realm of reprehensibility, but not in the highest realm. In addition Exxon’s post-grounding efforts to mitigate the harm serve materially to reduce the reprehensibility of the original misconduct. They reduce the reprehensibility for purposes of our review to, at most, a mid range.

2. Ratio of Harm to Punitives.

The second BMW guidepost, as reiterated and refined by State Farm, is the “ratio between harm, or potential harm, to the plaintiff and the punitive damages award.” State Farm, 538 U.S. at 424, 123 S.Ct. 1513. The goal of our review at this guidepost is to “ensure that the measure of punishment is both reasonable and proportionate to the amount of harm to the plaintiff and to the general damages recovered.” Id. at 426, 123 S.Ct. 1513.

a. Calculating The Harm.

In this case, the figure the district court used to represent the harm to plaintiffs was $513.1 million. District Court Opinion, 296 F.Supp.2d at 1103. Calculating *619the total harm to plaintiffs proved to be difficult because, in addition to considerable economic losses, the spill caused other undeniable, if not easily quantifiable, harms. See id. at 1094. The district court eventually calculated the harm figure by adding the compensatory damages verdict from the second phase of the trial to the actual judgments, settlements, and other recoveries various plaintiffs obtained as a result of the spill. Id. at 1099-1101.

Exxon does not dispute that the district court’s finding of $513.1 million in harm is fundamentally a valid measure of the actual harm caused by the spill. However, it disagrees that it should be the figure we ultimately use as part of the ratio of punitive damages to harm that we review as the second guidepost.

Exxon’s principal contention is that, before establishing the harm figure in the ratio, we must first deduct millions of dollars of payments and costs from the figure representing the total actual harm caused by the spill. Exxon would have us subtract a sum of about $493 million representing amounts paid to plaintiffs through Exxon’s voluntary claims program and other settlements. Exxon would then have us use that reduced figure to represent the total harm in assessing the ratio of punitives to harm.

This brings us to the central argument Exxon makes in this appeal. Exxon focuses on the language of our prior opinion in Punitive Damages Opinion I where we said, in a lengthy discussion of formulating possible ratios pursuant to BMW v. Gore, “[t]he amount that a defendant voluntarily pays before judgment should generally not be used as part of the numerator, because that would generally deter settlements prior to judgment.” 270 F.3d at 1244. Exxon contends this now means that in assessing the ratio of harm to punitives after State Fatm, we should ignore the total harm in favor of a figure that in fact more closely approximates Exxon’s remaining post-judgment liability for compensatory damages.

If we were to adopt Exxon’s interpretation of that sentence as binding us now, the measure of harm would be a meager $20.3 million. Applying the ratio of close to 1 to 1 that Exxon asserts is appropriate, Exxon contends we should cap punitive damages at $25 million. Under Exxon’s theory, even using a ratio of 9 to 1, which approaches the highest allowable under State Farm, punitive damages would be capped at $182.7 million. This would be the limit, even though Exxon’s recklessness led to more than $500 million in harm. We said, in discussing the nature of the relationship between punitive damages and harm:

The “reasonable relationship” is intrinsically somewhat indeterminate. The numerator is “the harm likely to result from the defendant’s conduct.” [BMW v. Gore, 517 U.S. at 581, 116 S.Ct. 1589]. The denominator is the amount of punitive damages. Because the numerator is ordinarily arguable, applying a mathematical bright line as though that were an objective measure of how high the punitive damages can go would give a false suggestion of precision. That is one reason why the Supreme Court has emphasized that it is not possible to “draw a mathematical bright line between the constitutionally acceptable and the constitutionally unacceptable that would fit every case.” [BMW v. Gore, 517 U.S. at 576, 116 S.Ct. 1589]. . . .

Although it is difficult to determine the value of the harm from the oil spill in the case at bar, the jury awarded $287 million in compensatory damages, and the ratio of $5 billion punitive damages to $287 million in compensatory *620damages is 17.42 to 1. The district court determined that “total harm could range from $287 million to $418.7 million,” which produces a ratio between 12 to 1 and 17 to 1. This ratio greatly exceeds the 4 to 1 ratio that the Supreme Court called “close to the line” in Pacific Mutual Life Ins. Co. v. Haslip [, 499 U.S. at 23, 111 S.Ct. 1032].

The amount that a defendant voluntarily pays before judgment should generally not be used as part of the numerator, because that would deter settlements prior to judgment. “[T]he general policy of federal courts to promote settlement before trial is even stronger in the context of large scale class actions,” such as this one. [Cf. Icicle, 229 F.3d at 795; Baker, 239 F.3d at 988].

Punitive Damages Opinion I, 270 F.3d at 1243-44.

The district court rejected the proposition that voluntary payments before judgment should not generally be used as part of the calculation of harm. But our prior decision did not constrain the ratio analysis so firmly as Exxon contends. We did not say that voluntary payments before judgment could not be considered in calculating the numerator for purposes of comparing the numerator with the amount of the award; we said that they “generally” could not. Considerations of settlement, critical to our analysis in Icicle, 229 F.3d 790, bear on the due process concerns at the heart of BMW’s discussion. Whenever a defendant governed by a board is sued for conduct egregious enough to create a genuine risk of punitive damages, those making its litigation decisions have to try to predict what may happen in court. Some may recommend obdurate resistance, and some may recommend settlement, or prejudgment payments even without settlement, each making arguments based on predictions. Those recommending payment can reasonably predict that the entity will not be hammered as hard as if it obstinately resisted acceptance of any responsibility. And their prediction would be reasonable. Criminal penalties have always been somewhat more lenient for those who accepted responsibility prior to judgment, see United States v. Gonzalez, 897 F.2d 1018, 1021 (9th Cir.1990) (upholding the constitutionality of U.S.S.G. § 3E1.1), and punitive damages are but a civil version of punishment for wrongdoing. It makes no practical sense to disarm all those in the future who want their boards to accept some responsibility by cutting out all the benefit their firms would get.

There is a limit, however, to how far acceptance of responsibility goes in both contexts. No criminal defendant guilty of a serious wrong ordinarily resulting in lengthy imprisonment could reasonably assume that he would receive no imprisonment at all if he promptly pleaded guilty. And no defendant’s board could reasonably predict that the defendant could escape all punishment by paying predicted compensatory damages before judgment. While “generally” prepayments should not be used as part of the calculation of harm, Punitive Damages Opinion I, 270 F.3d at 1244, that is not a mechanical arithmetic limit, just as the nine to one limit is not a mechanical arithmetic limit. See State Farm, 538 U.S. at 425, 123 S.Ct. 1513; Planned Parenthood, 422 F.3d at 962; Bains, 405 F.3d at 776-77. Due process considerations limit punitive damages to what the wrongdoer could reasonably foresee, and that works both ways.

Therefore, Exxon’s argument goes too far. It would produce, in Exxon’s analysis, a $25 million limit on punitive damages where the harm was $513 million but $493 million was paid before judgment. For *621purposes of notice to a tortfeasor of its liability risk, $25 million for causing a half billion loss would obviously be too good to be true. A defendant cannot buy full immunity from punitive damages by paying the likely amount of compensatory damages before judgment.

There is also a limit on the law of the case doctrine. One exception to this doctrine exists for an intervening change of law. See United States v. Bad Marriage, 439 F.3d 534, 538 (9th Cir.2006). In this case, there is such a change. Subsequent to our decision in Punitive Damages Opinion I, the Supreme Court decided State Farm. In that case, the fact that State Farm “paid the excess verdict before the complaint was filed,” State Farm, 538 U.S. at 426, 123 S.Ct. 1513, was a mitigating factor reducing the ratio. The Supreme Court did not use it to reduce the amount of total harm. The Court in State Farm itself took into account, in its consideration of whether the ratio was proper, the sub-stantiality and completeness of the compensatory award, the essentially economic nature of the harm, the likelihood that the punitive award duplicated the compensatory, and the defendant’s prompt settlement of compensatory damages. Id. All these mitigating factors were used to assess whether the ratio itself was likely to comply with due process. State Farm did not use such mitigating factors to reduce the harm. State Farm makes untenable the idea that a defendant’s voluntary, prejudgment payment of compensatory damages may not generally be used as part of the calculation of harm. Punitive Damages Opinion I, 270 F.3d at 1244.

There are also some secondary issues relating to calculating harm. One concerns payments made by Alyeska Pipe Lines Service Corporation. Exxon asks us to set off $98 million that its original co-defendant Alyeska Pipe Lines Service Corporation paid in settlement of plaintiffs’ claims. A consortium of oil companies, including Exxon, had contracted with Alyeska to respond to any oil spill in the area. After the Exxon Valdez disaster, plaintiffs sued Alyeska for negligence in its response to the spill, and eventually settled all claims against Alyeska, including punitive damages, for $98 million. Exxon’s argument here is that this $98 million payment represents harm attributable to Alyeska’s negligence, not Exxon’s recklessness, and therefore should not be used to calculate damages designed to punish and deter Exxon’s own harmful conduct.

There are two major reasons why Exxon’s position is not correct. First, the harm caused by the oil spill is attributable to Exxon under tort law principles. Exxon knowingly placed a relapsed alcoholic in control of a supertanker loaded with millions of gallons of oil. When it did so, Exxon accepted the foreseeable risk from its choice of captain that the tanker would have an accident causing an oil spill, and that Alyeska might further aggravate the harm. See Restatement (Second) of Torts §§ 433(a) cmt. c, 447(c),1 cmt. e.2 In fact, *622William Stevens, the President of Exxon, testified before Congress that Exxon knew Alyeska was not prepared to contain a spill of the size caused by the Exxon Valdez. Because Exxon could be held liable for this foreseeable risk, the district court properly included the harm caused by Alyeska’s response as the natural consequence of the harm caused by Exxon.

Second, the situation Exxon now complains of is strictly of its own making. In 1994, the Supreme Court held that the proportional fault rule governs calculation of non-settling defendant’s liability for compensatory damages in maritime torts. See McDermott, Inc. v. AmClyde, 511 U.S. 202, 114 S.Ct. 1461, 128 L.Ed.2d 148 (1994). Instead of following McDermott, Exxon agreed with plaintiffs to proceed as if a pro tanto rule with respect to co-defendants’ settlements still governed.3 Exxon apparently thought it more advantageous at the time to have the $98 million deducted from the final compensatory damage award after the fact, rather than have the jury make a proportionate fault finding. Since Exxon has already agreed that the $98 million does not represent harm attributable to Alyeska, Exxon is not warranted in asserting that this is what it represents now.

Exxon also contends that some $34 million included in the district court’s harm finding should not properly be considered harm at all. This figure represents an apparent $9 million overpayment by the Trans-Alaska Pipeline Liability Fund, $13.4 million from the Phase IV settlement Exxon claims is already accounted for elsewhere in the district court’s calculations, and $11.5 million paid to Native corporations and municipalities for environmental clean up.

We conclude that the $9 million overpayment, inadvertently included in the district court’s findings, should be subtracted from the total harm. Because Exxon does not specify where the $13.4 million in double-counting is reflected in other parts of the district court’s calculation, however, we are unable to determine from our own review of the record where they might be included. Therefore, Exxon has failed to convince us that this figure should be reduced from the harm.

Finally, the $11.5 million Exxon paid to the plaintiffs for clean up, like its early settlement of plaintiffs’ prospective commercial losses, is a mitigating factor relevant to our judgment about whether this punitive damages award is appropriate. Like the earlier settlements the proper place for its influence is as a mitigating circumstance to be considered in our overall determination of the ratio’s reasonable*623ness. It does, however, represent a part of the total harm for which Exxon is accountable.

In sum, the district court’s attempt to approximate the actual harm by adding together the various judgments, settlements, and liabilities that Exxon had already acknowledged was sound. Subtracting the $9 million Trans-Alaska Pipeline Liability Fund overpayment that the district court inadvertently overlooked, we conclude this record supports a total harm component of $504.1 million for purposes of analyzing the ratio of harm to punitives.

b. Evaluating the Reasonableness of the Ratio of Harm to Punitives.

After our second remand, the district court reduced the original punitive damages award of $5 billion to $4.5 billion. This yielded a punitive damages to harm ratio of 8.77 to 1. After our $9 million adjustment to the harm figure, that ratio now stands at 8.93 to 1-a proportion bordering on the presumption of constitutional questionability. See State Farm, 538 U.S. at 425, 123 S.Ct. 1513.

In State Farm, the Supreme Court explained that “few awards exceeding a single-digit ratio between punitive and compensatory damages, to a significant degree, will satisfy due process.” Id. at 425, 123 S.Ct. 1513. Relatively high single-digit ratios and perhaps even double-digit ratios may comply with due process where “a particularly egregious act has resulted in only a small amount of economic damages” or where “the injury is hard to detect or the monetary value of the noneconomic harm might have been difficult to determine.” Id. (quoting BMW v. Gore, 517 U.S. at 582, 116 S.Ct. 1589). Conversely, lower single-digit ratios, even as low as 1 to 1, might mark the outer limits of due process where compensatory damages are substantial. Id. This strongly suggests the ratio here is too high.

Our own decisions are also helpful. In Planned Parenthood, we used this guidance from State Farm to construct a “rough framework” for determining the appropriate ratio of punitive damages to harm. See 422 F.3d at 962. We held that in cases where there are “significant economic damages” but behavior is not “particularly egregious,” a ratio of up to 4 to 1 “serves as a good proxy for the limits of constitutionality.” Id. (citing State Farm, 538 U.S. at 425, 123 S.Ct. 1513). In eases with significant economic damages and “more egregious behavior,” however, a single-digit ratio higher than 4 to 1 “might be constitutional.” Id. (citing Zhang, 339 F.3d at 1043-44; Bains, 405 F.3d at 776-77). Finally, in cases where there are “insignificant” economic damages and the behavior is “particularly egregious,” we said that “the single-digit ratio may not be a good proxy for constitutionality.” Id.

The circumstances of this case fit into the second class of cases in the Planned Parenthood framework. Exxon’s reckless decision to risk the livelihood of thousands by placing a relapsed alcoholic in command of a supertanker, while mollified by its prompt settlement and clean up policies, was “particularly egregious.” Moreover, the $500 million of loss is well within the range of “significant” economic damages. Thus, under Planned Parenthood, an appropriate ratio would be above 4 to 1.

Our review of the reprehensibility and mitigation under the first guidepost of reprehensibility, however, compels us to conclude the award should be toward the lower end of that range. Our cases have generally reserved high single-digit ratios for the most egregious forms of intentional misconduct, such as threats of violence and intentional racial discrimination. See Zhang, 339 F.3d at 1044 (upholding a ratio *624of 7:1 for intentional racism); Bains, 405 F.3d at 776-77 (remanding for district court to set a ratio between 6:1 and 9:1 for intentional racism); Planned Parenthood 422 F.3d at 952, 963 (remitting to a 9:1 ratio for threats of violence). Exxon’s conduct in this case, while inexcusable, did not involve any intentional conduct that would normally be required to support a punitive damages award with a high single-digit ratio.

Here mitigating factors also come into play. Exxon instituted prompt efforts to clean up the spill and to compensate the plaintiffs for their economic harm. As we earlier observed, if a defendant acts promptly to ameliorate harm for which it is responsible, the size of a punitive damages award should be reduced to encourage socially beneficial behavior. Punitive Damages Opinion I, 270 F.3d at 1242. Moreover, the costs that Exxon incurred in compensating the plaintiffs and cleaning up the oil spill have already substantially served the purposes of deterrence, lessening the need for a high punitive damages award. Id. at 1244.

Thus, Exxon’s conduct was particularly egregious and involved significant economic damages. Nevertheless, its conduct was not intentional and it promptly took steps to ameliorate the harm it caused. With these considerations in mind, we conclude that a punitive damages to harm ratio of more than 5 to 1 would violate due process standards under current controlling Supreme Court and Ninth Circuit authority.

3. Comparable Penalties.

The third BMW v. Gore/State Farm guidepost is comparable legislative penalties. Given the emphasis on this factor in BMW v. Gore, we went to some lengths in Punitive Damages Opinion I to extrapolate the comparable penalties that would be imposed under state and federal law for the spill, the highest being approximately $1.03 billion dollars.

In State Farm, however, the Supreme Court stated that “need not dwell long on this guidepost.” State Farm, 538 U.S. at 428, 123 S.Ct. 1513. In that case, the comparable penalties were not particularly informative: the comparable civil penalty was easily “dwarfed” by the punitive award, and as to criminal penalties, the Court explained that although their existence “does have bearing on the seriousness with which a State views the wrongful action,” they had “less utility” “[w]hen used to determine the dollar amount of the award.” Id.

In our own circuit’s more recent post-BMW v. Gore and State Farm cases, we have generally not attempted to quantify legislative penalties. We have looked only to whether or not the misconduct was dealt with seriously under state civil or criminal laws. See, e.g., Planned Parenthood, 422 F.3d at 963. In several recent decisions we have not discussed the factor at all. See Southern Union Co., 415 F.3d at 1009-11 (9th Cir.2005); Hangarter, 373 F.3d at 1014-15. This may be because legislative judgments, unlike jury verdicts, do not represent an individualized assessment of reprehensibility.

Here, the matter of spilling oil in navigable water has clearly been taken quite seriously by legislatures, with Congress enacting a specific statute after the spill, and state and federal law having already authorized substantial penalties. See Punitive Damages Opinion I, 270 F.3d at 1245-46. Thus, the third BMW v. Gore/ State Farm factor, substantial legislative penalties, supports our conclusion that Exxon’s reckless conduct merits substantial punitive damages.

IV. CONCLUSION

For the foregoing reasons, Exxon’s reckless misconduct in placing a known *625relapsed alcoholic in command of a supertanker, loaded with millions of barrels of oil, to navigate the pristine and resource abundant waters of Prince William Sound was reckless and warrants severe sanctions. The misconduct did not, however, warrant sanctions at the highest range allowable under the due process analysis, as explained in the Supreme Court’s most recent opinion in State Farm.

The district court’s imposition of punitive damages of $4.5 billion, entered after our remand to reconsider due process in light of State Farm, represents damages at the very highest range, and is not warranted. It is not consistent with the Supreme Court’s opinion in State Farm or with the most important tenets of our pri- or opinion in Punitive Damages Opinion I relating to Exxon’s mitigation of reprehensibility. Although a one to one ratio marked the upper limit in State Farm, the conduct here was far more egregious and justifies a considerably higher ratio. An award of damages representing a ratio of punitives to harm of 5 to 1 is consistent with both.

The judgment of the district court is VACATED, and the matter is remanded with instructions that the district court further reduce the punitive damages award to the amount of $2.5 billion. We have decided pursuant to the de novo standard of review imposed by Leatherman, 532 U.S. at 436, 121 S.Ct. 1678, that this is the appropriate limit on punitive damages in this case under the prevailing legal precedent. Thus, we do not remand for further consideration of what the limit may be. It is time for this protracted litigation to end.

VACATED AND REMANDED.

. "The fact that an intervening act of a third person is negligent in itself or is done in a negligent manner does not make it a superseding cause of harm to another which the actor’s negligent conduct is a substantial factor in bringing about, if ... (c) the intervening act is a normal consequence of a situation created by the actor’s conduct and the manner in which it is done is not extraordinarily negligent.”

. "The words 'extraordinarily negligent’ denote the fact that men of ordinary experience and reasonable judgment, looking at the matter after the event and taking into account the prevalence of that 'occasional negligence,’ which is one of the incidents of human life,’ would not regard it as extraordinary that the third person’s intervening act should have been done in the negligent manner in which it was done. Since the third person's action is *622a product of the actor's negligent conduct, there is good reason for holding him responsible for its effects, even though it be done in a negligent manner, unless the nature of the negligence is altogether unusual.”

. The stipulation between the parties reads in relevant part:

"[N]otwithstanding the rule of proportionate shares set out in McDermott, Inc. v. AmClyde, credit for the Aleyska settlement ... shall be deducted from the sum that would, in the absence of this stipulation, be the aggregate amount of any judgment or judgment in favor of plaintiffs ... and the liability of Exxon and Shipping for compensatory damages to any and all plaintiffs herein shall be reduced by the aggregate sum of $98 million.... The parties expressly recognize and agree that the sum of $98 million is not necessarily a fair measure of what would be Alyeska's proportionate share of liability to plaintiffs!,] but the parties are entering into this Stipulation in order to avoid the alteration of their trial preparation that would result from a last-minute overturning of the parties’ assumption that[the pro tanto approach] would govern at trial and from requiring litigation of Alyeska's proportionate share.”