FISHER, Circuit Judge:

This product liability case arises from the death of three-year-old Walter White, the son of plaintiffs Ginny and Jimmie White, who was killed when Mr. White's parked Ford F-350 pickup truck rolled over him in the family's driveway. The case is before us for the second time following a remand for a new trial on punitive damages. See White v. Ford Motor Co., 312 F.3d 998 (9th Cir. 2002) (" White I ") (affirming the first jury's award of $2,305,435 in compensatory damages but reversing as to punitive damages). Defendant Ford Motor Company appeals the district court's decision that a second jury's award of $52 million in punitive damages on remand did not violate the Due Process Clause of the Fourteenth Amendment. In addition, Ford argues that during the retrial, the district court committed multiple reversible errors in its pretrial and other jury instructions and evidentiary rulings. In light of the Supreme Court's intervening decision in Philip Morris USA v. Williams, ___ U.S. ___, 127 S.Ct. 1057, 166 L.Ed.2d 940 (2007), we reverse and remand for a new trial on punitive damages.

I. Background

A. The Defective Product

In 1991, Ford began to produce model year 1992 F-series pickup trucks using a self-adjusting parking brake designed and manufactured in the 1980s by the Orscheln Company. Although conventional brakes employ a cable that becomes loose over time and therefore must be adjusted periodically to maintain tension, Orscheln's self-adjusting brake kept the cable tight, even as the vehicle aged. The brake used a self-adjusting ratchet wheel and a "pawl," which is a "hinged or pivoted finger that sticks into a tooth of the ratchet wheel." White I, 312 F.3d at 1002-03. A driver would engage the brake by depressing the pedal, setting the mechanism in gear until the pawl tooth fit in between two of the ratchet teeth.

By 1990, Ford had preproduction reports of potential problems with the Orscheln brake, and by the time the F-series trucks were in production, the reports had increased. For example, some customers reported that sometimes the parking brake did not engage and instead pressed freely down to the floor, while other customers reported that their trucks rolled despite the parking brake being engaged. Id. at 1003.

Ford told Orscheln to identify and fix the problem, and in the fall of 1992, the company assigned Timothy Rakowicz, a young Ford engineer, to assist in the investigation. By November 1992, Ford and Orscheln discovered that sometimes the pawl tooth would skip over the tops of the ratchet wheel teeth instead of engaging in one of the gaps. Ford called this the "skip-through-on-apply" or "skip out" problem. Id. In February 1993, Orscheln testing showed that if the pawl tooth engaged a ratchet tooth at its tip, rather than firmly engaging between the two teeth, the driver would feel resistance when pressing the brake pedal even though the pawl tooth was in fact resting on a ratchet tooth tip. Testing also showed that an outside force on the vehicle could disturb the equilibrium and cause the brake to disengage, allowing the vehicle to roll. Ford referred to the tip-on-tip condition as "spontaneous disengagement," id. at 1007, or self-release, and the corresponding effect on the vehicle as "rollaway," id. at 1003.

Rakowicz included Orscheln's test results in a February 22, 1993 draft paper to Ford's Critical Product Problem Review Group (CPPRG), a committee whose job was to assist with Ford's investigation of potential safety problems. In the paper, titled "F-Series Parking Brake Control Self Releasing Field Campaign and Owner Notification Paper," Rakowicz wrote that "the parking brake control will intermittently self release after pedal apply causing a decrease in pressure to the rear brakes." Rakowicz defined the "root cause of the concern" as a "load carrying, ratchet tooth tip on tip condition," and referring to tests three days earlier, wrote that "[t]he condition has been duplicated during parking brake hill hold testing on February 19, 1993 at the supplier facility using a part removed from a problem vehicle." Rakowicz wrote that a customer with a brake assembly affected with this condition "would experience the following affects [sic]":

a. the parking brake pedal apply would feel normal.

b. the customer will leave the vehicle and in an arbitrary amount of time, the parking brake control will self release. A popping noise will be heard if the customer is within hearing distance. The parking brake pedal will remain in approximately the same position it was applied to.

c. If the vehicle is on an incline, the vehicle mil potentially roll down the incline (emphasis added).

Rakowicz believed that the tip-on-tip condition warranted a recall. At retrial, he testified that he was personally aware of 22 reported rollaways when he wrote his report. However, more senior engineers at Ford disagreed with Rakowicz's draft and required him to tone it down, concluding instead that the Orscheln test was not valid. Id. Thus, the CPPRG did not then or ever refer the matter to Ford's Field Campaign Review Committee (FCRC), which made recall decisions. At the same time, Ford management worried about a recall and its potential cost. At a February 23, 1993 meeting of Ford and Orscheln personnel, a senior Ford engineer lamented that "this problem may cause serious financial ramifications for both companies with warrant recall." A March 30, 1993 memo similarly referred to a full recall as the "worst case scenario." The ultimate cost to recall 875,000 manual transmission vehicles was approximately $22 million.

In March 1993, Orscheln proposed a solution for the skip out problem. The fix was a small plastic wedge, costing 15 cents to manufacture, which could be installed over the pawl to make sure it pressed down between the teeth instead of skipping over them. However, the wedge also disabled the self-adjusting feature of the brake and increased Ford's cost. Id. at 1003-04.

Meanwhile, the evidence of problems with the F-series parking brake increased, and in the same month (March 1993), the National Highway Traffic Safety Administration (NHTSA) became involved in the investigation, having received reports of rollaways. In May 1993, Ford disclosed to NHTSA 65 reports of either skip-through or rollaway. The agency requested additional information in September 1993, including complaints, testing documents and accident reports, which Ford provided in February 1994. In addition, the agency collected and inspected 10 parking brake assemblies from complainants, bench tested five of these assemblies and tested three of them on a complaint vehicle. Although NHTSA was unable to detect a "spontaneous disengagement" malfunction in these specific assemblies, it did find that some of them experienced skip out.

By August or September 1993, the plastic wedge was approved for use, and enough were manufactured for Ford to install them in all of the trucks on the road. Id. at 1003. But rather than recall the vehicles, Ford issued a Technical Service Bulletin to dealers on November 10, 1993, making the wedge available for any vehicle owner who complained about a problem with skip out. Id. at 1004. The bulletin did not mention spontaneous disengagement or rollaway.

In mid-1994, NHTSA began to pressure Ford to conduct a recall and install the wedge as a precautionary measure. On August 30, 1994, and in response to NHTSA's urging, Ford informed the agency that it had decided to recall 884,400 manual transmission vehicles — which included 1992 through 1994 F-series pickup trucks — "rather than engag[e] in a protracted dispute with the agency." In its letter to the agency, Ford stated that "[a]lthough all of the vehicles are potentially susceptible to the condition arising, less than one percent" — or 8,440 vehicles — "are expected to exhibit it, and those which do . . . are expected to do so on an intermittent basis." In describing the problem, Ford noted that it "may occur occasionally when the parking brake control self-adjust pawl lines up in a tip-to-tip relationship with the self-adjust ratchet," explaining that "[i]f this occurs, the pawl may skip over one or more teeth in the ratchet during parking brake application." Ford further informed the agency of 44 reports of "unattended vehicle[s] roll[ing] and resum[ing] in property damage," one of which alleged that an injury had occurred. However, the Whites proved at retrial that Ford knew of at least eight injuries and that it had received reports of over 100 rollaways at the time it submitted its letter to NHTSA.

In November 1994, Ford provided NHTSA with its proposed recall notice, which stated: "During rapid pedal application of the parking brake, the pedal may go to the floor with little or no effort. Should this occur, the parking brake system may not achieve full tension, potentially resulting in parking brake ineffectiveness, or diminished effectiveness." NHTSA acknowledged receipt of the recall notice and described the defect as: "The parking brake control self-adjust pawl does not line up properly with the self-adjust ratchet allowing the pawl to skip over one or more teeth in the ratchet. The parking brake will not hold the vehicle allowing the vehicle to roll freely." The agency, however, did not ask Ford to make any changes to its notice. On December 16, 1994, Ford issued its recall notice to vehicle owners.

B. The Accident, First Trial and First Appeal

On October 9, 1994, two months before Ford issued its recall, Jimmie White parked his company's 1993 Ford F-350 pickup truck facing downhill in his sloped driveway. Mr. White put the truck into first gear, set the parking brake by stepping on the brake pedal and went inside the home. The Whites' three-year-old son, Walter, was playing unsupervised in the front yard, with Mrs. White occasionally checking on him through the window. Walter climbed into the vehicle's cab, and according to the Whites, knocked the gear-shift into neutral. When the parking brake did not hold, the truck started to roll and Walter fell or climbed out of it. Tragically, the truck's rear dual wheels rolled over and crushed the boy's body. Id. at 1002.

The truck that killed Walter was built in April 1993 and sold to Jimmie White's employer on September 22, 1993. By the time of the sale of that vehicle, it was clear to Ford that there was a potential rollaway problem. Id. at 1004. Although Ford agreed to conduct a recall before the date of the accident, Ford did not begin issuing the recall notice to owners until December 1994, two months after Walter's death. Moreover, Mr. White's employer did not receive a recall notice until March 1995, 25 months after Rakowicz had recommended a recall and seven months after Ford told NHTSA it would do so. Thus, as we stated in White I, "the accident happened after the brake problem was discovered and figured out, after the technical bulletin to dealers had gone out, and after Ford had decided to recall the trucks to install the fix, but before the recall notices or any warnings to ultimate consumers were sent out." Id.

The Whites brought a products liability action under Nevada law against Ford and Orscheln, alleging strict product liability (defective design), negligence, failure to warn, intentional misrepresentation and negligent infliction of emotional distress. Their theory of the case was that the parking brake failed to engage, allowing the truck to roll, and that Ford knew the brake was prone to failure but refused to recall it or warn consumers of the danger. Id. at 1002.

During the 1998 trial, Orscheln settled with the Whites and the case went to verdict against Ford. Id. The jury found that the brake was "defective in design," but that this design defect did not proximately cause the accident. Rather, the jury found that the brake was "defective for Ford's failure to warn," that Ford was negligent with respect to the brake and that Ford had thereby proximately caused Walter's death. The jury also found the Whites 40 percent contributorily negligent. Finally, the jury found Ford liable for negligent infliction of emotional distress and intentional misrepresentation of the safety of the vehicle, and found by clear and convincing evidence that "`Ford acted with oppression or malice in the conduct'" that caused Walter's death. Id. at 1012 (quoting the jury questionnaire). The jury returned a verdict against Ford for $2,305,435 in compensatory damages — $1,150,000 to each parent for emotional distress and $5,434.57 to Walter's estate for funeral expenses — and $150,884,400 in punitive damages. The district court found that the punitive damages award was constitutionally excessive and therefore remitted it to $69,163,037.10-30 times the compensatory award — on the theory that the largest punitive damages award the Nevada courts had approved was 30 times the compensatory damages. Id. at 1002; see also id. at 1029 (Graber, J., dissenting). The parties then cross-appealed.

In White I, this court unanimously affirmed the first jury's compensatory and punitive liability findings. With respect to punitive damages, it held that the evidence was sufficient to support the following conclusions, which in turn were sufficient under Nevada law to support a punitive damages award:

(1) "that Ford knew the parking brake could [malfunction] and let a pickup truck roll away, didn't fix it, didn't recall it, and didn't warn drivers of the trucks, all prior to the White accident";

(2) "that Ford knew a parking brake [malfunction] could result in a rollaway, and failed to warn people driving pickup trucks about the brake in conscious disregard of their safety (that is, knowing that someone could be injured or killed and deliberately failing to warn)";

(3) "that when the engineers had figured out the problem and the fix for it, Ford covered it up with a euphemistic notice to dealers, rather than consumers, instead of a plain warning and immediate recall";

(4) "that `skip-through,' where a person pressed the parking brake with his foot and felt no resistance, and rollaway, where the parking brake seemed to engage but didn't prevent the truck from rolling away. . . . were both the same thing, failure of the pawl to drop properly between the teeth of the ratchet wheel"; and

(5) "once Ford knew that pickup trucks could be rolling down hills without drivers behind the wheel, it was obvious that someone was likely to be killed if Ford didn't do something about it."

Id. at 1010 (majority opinion).

However, a majority of the panel ultimately reversed the punitive damages award because the district court's instructions "unconstitutionally allowed a Nevada jury to punish Ford for out-of-state conduct." Id. at 1020. The majority held that "the jury must be limited to punitive damages reasonably required to vindicate Nevada's legitimate interests in punishment and deterrence, if any, and prohibited from imposing punitive damages to protect people or punish harm outside of Nevada." Id. It therefore remanded for a new trial on punitive damages without addressing whether the original or remitted amount of punitive damages was unconstitutionally excessive. See id. at 1016, 1020.

C. The Second Trial

In 2004, District Judge Hagen, who presided over the first trial, conducted a retrial limited to the amount of punitive damages before a new jury. The district court did not inform the second jury of the amount of the White's compensatory damages award, but rather told the jury that "[t]he original jury found for the [Whites] and awarded damages to compensate them for their injuries," and that the Whites had "already been compensated for the injuries they suffered." It also explained that Ford's "liability for punitive damages has already been established" and that "[t]he only remaining question for the jury . . . to decide is the amount of punitive damages."

After an eight-day trial, the jury awarded the Whites $52 million. The district court initially reduced the award to $41,497,821 to account for the Whites' comparative fault, but subsequently reinstated the full award, concluding that under Nevada law the Whites' fault should not be considered. The district court also concluded that the full $52 million award, representing a 22.6 to 1 ratio of punitive to compensatory damages, was constitutionally permissible.

Defendant Ford Motor Company appealed the second jury's award as a violation of the Due Process Clause of the Fourteenth Amendment, and urged us to reverse for instructional and evidentiary errors. After we heard argument, the Supreme Court decided Philip Morris USA v. Williams, ___ U.S. ___, 127 S.Ct. 1057, 166 L.Ed.2d 940 (2007), holding that states must provide assurances that juries do not award punitive damages based in part on a desire to punish defendants for harming nonparties. We reheard argument to consider the implications of Williams.

II. Discussion

A. Harm to Nonparties Instruction

The Due Process Clause "forbids a State to use a punitive damages award to punish a defendant for injury that it inflicts upon nonparties or those whom they directly represent, i.e., injury that it inflicts upon those who are, essentially, strangers to the litigation." Williams, 127 S.Ct. at 1063. A jury may consider evidence of actual harm to nonparties as part of its reprehensibility determination, but may not "use a punitive damages verdict to punish a defendant directly." Id. at 1064. Where there is a significant risk that jurors will misapprehend the distinction, the court must upon request protect against that risk by "avoid[ing] procedure that unnecessarily deprives juries of proper legal guidance." Id.

During closing argument, counsel for the Whites told the jury that by January 1999, Ford knew that 54 people had been injured by "rollaways." Counsel argued that Ford "decided to do everything possible to avoid telling people the truth" about the rollaway problem because "[t]hey know there are not less than 54 people as of January of 99 who were injured from a rollaway . . . [a]nd they don't ever want to have to be accountable for that conduct."

Concerned that the jury would punish Ford for the harm suffered by other rollaway victims, counsel for Ford had objected to the district court's proposed jury instructions, and requested an instruction that would prevent the jury from punishing "[Ford] in this case not just for the harm to these plaintiffs, but for harm to other plaintiffs, whether in state or out of state." Ford conceded that evidence of harm to other people could be considered by the jury in assessing reprehensibility, but argued that the jury could punish only "for the harm to this plaintiff." The district court refused such an instruction, deciding that it was not required by existing precedent.

Although we are sympathetic to the difficult task that both district court judges and counsel face in maneuvering through this ever-shifting area of the law, we are compelled to reverse here and remand for a new trial in light of the intervening precedent of Williams. As in Williams, there is a significant risk that the jury, in arriving at its punitive damage award, punished Ford for harm to nonparties. Absent a proper limiting instruction, the jury could have mistakenly understood the Whites' argument that Ford's conduct injured 54 other people to justify not just a finding of reprehensibility, but also to consider those other injuries in calculating the amount of damages warranted to punish Ford's reprehensible conduct. Given Williams' guidance, we must conclude that the court's failure to give a harm to non-parties instruction violated due process.

The Whites argue that any constitutional error was harmless because the district court gave an extraterritoriality instruction, restricting jurors from "add[ing] damages to protect people or to punish harm to people outside of Nevada" (emphasis added). We agree this limiting instruction mitigated the problem. It did not render the error "more probably than not harmless," Swinton v. Potomac Corp., 270 F.3d 794, 805 (9th Cir. 2001), however, because the jury may have interpreted the extraterritoriality instruction as allowing damages for harm to people residing inside Nevada.

There was no evidence linking any of the 54 nonparty victims to Nevada, but neither did the evidence establish that none of the victims were Nevada residents. This gap in the evidence was critical because the jury heard plaintiffs' counsel assert that Nevada residents purchased Ford truck models affected by the rollaway problem in disproportionate numbers. Thus, jury members may have assumed that at least some of the 54 nonparty victims lived in Nevada.

To remedy the due process violation, we may either remand for a new trial on punitive damages or "change . . . the level of the punitive damages award." Williams, 127 S.Ct. at 1065. We conclude that a new trial on punitive damages is the proper remedy. See Werbungs Und Commerz Union Austalt v. Collectors' Guild, Ltd., 930 F.2d 1021, 1027 (2d Cir. 1991) (holding that remittitur is not designed to compensate for excessive verdicts in cases where jury is improperly instructed). On remand, the district court must explain to the jury that although evidence of harm to nonparties may bear on Ford's reprehensibility, any award of punitive damages cannot be used "to punish [Ford] directly for harms to . . . non-parties." Williams, 127 S.Ct. at 1064.

B. Other Alleged Instructional and Evidentiary Errors

The remainder of Ford's challenges can be divided into three categories. First, Ford argues the district court erroneously refused to provide a "reasonable relationship" instruction and further failed to tell the jury the amount of compensatory damages the first jury awarded to the Whites. Second, Ford faults the district court's opening jury instructions in several respects. Third, Ford argues the district court erred by admitting into evidence a redacted, one-page version of Ford's December 21, 2003 consolidated balance sheet, which Ford claims misrepresented the company's financial condition.

We review the district court's formulation of jury instructions and its evidentiary rulings for abuse of discretion. See Beachy v. Boise Cascade Corp., 191 F.3d 1010, 1012 (9th Cir. 1999). We conclude that the district court abused its discretion by not telling the jury: (1) the amount awarded to the Whites in compensatory damages; (2) that the first jury found Ford's brake defect did not proximately cause the accident; and (3) that the Whites were 40 percent responsible for the accident. We decline to decide whether the district court abused its discretion by excluding Ford's 2002-03 Financial Statement and 2002 Annual Report, and we reject the remainder of Ford's arguments.

1. Reasonable Relationship Instruction and Compensatory Damages

Similarly, Williams' holding is inapposite in the context of the reasonable relationship inquiry. Williams mandates that juries receive proper instruction on harm to nonparties, an instruction that is essential if the jury is to calculate the proper amount of punitive damages. But the reasonable relationship inquiry is markedly different from the jury's determination of a specific amount of punitive damages; its purpose is to aid in ascertaining the constitutional ceiling. Unlike the initial damage calculation, determining the constitutional ceiling on a punitive damage award is a question of law, properly reserved for the court. See Bains LLC v. Arco Prods. Co., 405 F.3d 764, 777 (9th Cir. 2005) (holding that "[t]he [constitutional] level of punitive damages is not a finding of `fact' that must be determined by the jury; it may be determined de novo by the court"). Although states are certainly free to incorporate the reasonable relationship concept into jury instructions, see Morgan v. Woessner, 997 F.2d 1244, 1256-57 n. 13 (9th Cir. 1993), it is also constitutionally permissible for a district court to delay the reasonable relationship inquiry until the judge's post-verdict review. Nevada has taken the latter course. See Bongiovi v. Sullivan, 122 Nev. 556, 138 P.3d 433, 452 (2006) (adopting State Farm guideposts for post-verdict judicial review for excessiveness under state law). Compare Nev. J.I. 10.20, with Cal. BAJI 14.71 (including Nevada's identically worded reprehensibility and deterrence factors but adding a reasonable relationship factor).

Compensatory damages. Although we hold that Ford was not entitled to a reasonable relationship instruction under either the Due Process Clause or Nevada law, we nevertheless conclude that the amount of compensatory damages is relevant to the amount of punitive damages. In accordance with Nevada law, the jury was properly instructed to consider two factors in determining the amount of punitive damages: "the reprehensibility of the conduct of the defendant" and "the amount of punitive damages which will have a deterrent effect on the defendant in light of the defendant's financial condition." Nev. J.I. 10.20. In a typical case, the same jury would award both compensatory and punitive damages. Here, because of this case's unique procedural history, the jury empaneled to award punitive damages was unfamiliar with the original jury's verdict and the amount of compensatory damages it awarded. Without knowing the amount of those damages, the punitive damages jury could not have come to a reasoned conclusion as to the amount of additional damages necessary to deter Ford from similar conduct in the future. The district court's withholding of that information was, therefore, an abuse of discretion.

On remand the district court must instruct the jury that the Whites have been awarded $2,305,435 in compensation for the death of their son. Consonant with Nevada law, the district court in formulating its instruction should ensure that the jury is not left with the erroneous impression that its punitive damages award must be proportional to the compensatory award. With that caveat, we leave to the district court's discretion the precise wording of the instruction, as well as the decision of what information, if any, other than the bare award amount must be provided to the jury.

2. Opening Jury Instructions Facts found by first jury. We reject Ford's challenge to statements made by the district court during its opening jury instructions. The court did not abuse its discretion by telling the jury to assume as true that: (1) Mr. White had put the truck in gear and set the parking brake; (2) the brake failed to hold; and (3) that Ford knew the parking brake was prone to failure, but continued to sell it without recalling it or warning customers. In White I, this court "assume[d] that the jury believed Mr. White's recollection" of the facts, including that he engaged the parking brake and that it was prone to failure but that Ford continued to use it. See White I, 312 F.3d at 1002; see also id. at 1008 ("If Mr. White's testimony was correct, and we must assume that the jury so found. . . ."). In a typical civil trial, a summary of facts would be unnecessary because the same jury would decide both compensatory and punitive damages. The district court's decision to familiarize the second jury with facts previously relied on, and implicitly affirmed, by this court was not an abuse of discretion.

Similarly, the district court did not abuse its discretion by telling the jury that this court "upheld" the first jury's liability findings. White I affirmed the first jury's compensatory and punitive liability determinations and remanded only on the amount of punitive damages. See id. at 1006, 1012, 1020. The district court did not abuse its discretion by reciting the law of the case.

Omissions from opening jury instructions. Ford also challenges certain omissions from the court's opening instructions. We agree with Ford that the district court should have mentioned the previous jury's finding that the brake defect did not proximately cause the accident. The punitive damages award was supposed to be based only on Ford's failure to warn, given the first jury's finding that the defect was not a proximate cause of the accident. Of course, the jury needed to know about the brake defect to understand why Ford's failure to warn caused the accident. As we explained in White I, Jimmie White "testified that, had he known that the brake could let go despite being set, he wouldn't have parked the truck on a slope." Id. at 1006. But by failing to inform the jury that the defect was not the proximate cause of the accident, the jury may not have been focused properly on the conduct actually found culpable by the first jury. Accordingly, on remand, the district court should inform the jury of the first jury's finding and instruct that punitive damages, if awarded, are to punish the "reprehensibility of [Ford's]" failure to warn. See Nev. J.I. 10.20.

The district court also should have disclosed that the Whites were found to be 40 percent responsible for the accident. To be sure, the overwhelming weight of authority suggests that Ford cannot offer evidence of the Whites' comparative negligence to offset the punitive damages award. See Clark v. Cantrell, 332 S.C. 433, 504 S.E.2d 605, 610 n. 5 (1998) (reviewing 16 reported opinions and identifying only one that allowed punitive damages to be reduced in accordance with comparative fault). Nonetheless, Nevada law required the jury to consider — in arriving at a punitive damages award — "the reprehensibility of [Ford's] conduct," Nev. J.I. 10.20, and reprehensibility is judged in relation to the conduct and actions of others, not merely by looking at Ford's conduct in the abstract. We assess blame for tortious conduct in relation to other contributory causes. Likewise, criminal acts may result in lesser punishments when the actor is not wholly responsible for the harm. This is true even for extremely blameworthy cases, such as capital defendants being permitted to show mitigating circumstances caused by others in order to reduce their punishments. See, e.g., Summerlin v. Schriro, 427 F.3d 623, 642 (9th Cir. 2005) (en banc). Thus, in determining punitive damages, the jury should be able to consider Ford's level of responsibility, as it bears directly on "the reprehensibility of [its] conduct." Nev. J.I. 10.20. The district court should therefore instruct the jury that Ford was only 60 percent responsible for the accident, and plaintiffs bore the remaining 40 percent of the blame. 3. Evidence of Ford's Financial Condition

To help the jury in imposing an appropriate punitive award, the district court admitted into evidence a redacted, one-page version of Ford's 2003 consolidated balance sheet, which revealed Ford's "[t]otal stockholders' equity," or net worth, to be $11,651,000,000. Ford raises a variety of objections to the district court's admission of this financial document.

First, Ford argues that the district court erred by admitting the balance sheet at all because such evidence is irrelevant to the setting of a punitive damages award. See Fed.R.Evid. 402. But under Nevada law, a defendant's wealth is relevant to a jury's determination of punitive damages. In Dillard Department Stores, Inc. v. Beckwith, 115 Nev. 372, 989 P.2d 882 (1999), the Nevada Supreme Court noted its consistent recognition that "[t]he wealth of a defendant is directly relevant to the size of [a punitive damages] award, which is meant to deter the defendant from repeating his misconduct as well as punish him for his past behavior." Id. at 887 (quoting Ainsworth v. Combined Ins. Co., 104 Nev. 587, 763 P.2d 673, 677 (1988)); see also United Fire Ins. Co. v. McClelland, 105 Nev. 504, 780 P.2d 193, 199 (1989) (recognizing "the financial position of the defendant" as a factor in assessing a punitive damages award). To that end, the Nevada pattern jury instructions expressly require that "[i]n arriving at any award of punitive damages," the jury consider "the amount of punitive damages which will have a deterrent effect on the defendant in light of the defendant's financial condition." Nev. J.I. 10.20 (emphasis added).

Nevada law on this matter is consistent with the Supreme Court's and this circuit's longstanding recognition of the admissibility of net worth evidence. See TXO Prod. Corp. v. Alliance Res. Corp., 509 U.S. 443, 462 n. 28, 113 S.Ct. 2711, 125 L.Ed.2d 366 (1993) (noting that it is "well-settled" that a defendant's "net worth" is a factor that is "typically considered in assessing punitive damages"); Pacific Mut. Life Ins. Co. v. Haslip, 499 U.S. 1, 21-22, 111 S.Ct. 1032, 113 L.Ed.2d 1 (1991) (approving, among other factors used to determine a punitive damage award, the "financial position" of the defendant); Bains, 405 F.3d at 777 ("A punitive damages award is supposed to sting so as to deter a defendant's reprehensible conduct, and juries have traditionally been permitted to consider a defendant's assets in determining an award that will carry the right degree of sting."). Thus, Ford's argument that evidence of a defendant's financial condition is irrelevant to the setting of a punitive damages award is meritless.

Second, Ford argues that even if evidence of financial condition is relevant, the district court violated Federal Rule of Evidence 403 because the consolidated balance sheet's probative value was substantially outweighed by the danger of unfair prejudice. Reviewing the district court's decision to admit the balance sheet for an abuse of discretion, see United States v. Allen, 341 F.3d 870, 886 (9th Cir. 2003), we conclude that admission of the 2003 balance sheet was proper because the evidence was not unfairly prejudicial. Unfair prejudice means "an undue tendency to suggest decision on an improper basis, commonly, though not necessarily, an emotional one." Fed.R.Evid. 403, advisory committee's note. Because evidence of Ford's financial condition was probative and a proper factor in the jury's analysis, any "prejudice" experienced by Ford was not "unfair" as that term is used in Rule 403. See United States v. Bailleaux, 685 F.2d 1105, 1111 n. 2 (9th Cir. 1982) ("`[U]nfair prejudice' means that the evidence not only has a significant impact on the defendant's case . . . but that its admission results in some unfairness to the defendant because of its non-probative aspect.").

Third, Ford argues that even if the district court properly admitted the consolidated balance sheet, it should also have allowed the jury to consider the company's full 2002-03 Financial Statement and 2002 Annual Report. Ford also challenges the court's denial of Ford's request to argue to the jury its interpretation of the figures within those documents. Ford suggests that its financial statement and annual report had to be admitted also, so that it could argue to the jury that its net worth was due to factors unrelated to F-series pickup trucks. Apparently, Ford believes that a defendant's net worth is relevant only insofar as it is derived from the defendant's wrongdoing. This "disgorgement" theory of punitive damages misunderstands the relevance of net worth. Determining what amount of punitive damages will "sting" the defendant requires consideration of its total wealth, not merely wealth derived from wrongdoing. Nonetheless, because we are reversing and remanding for a new trial on punitive damages, we decline to decide whether the district court abused its discretion by excluding the financial statement and annual report. The district court should have another opportunity to determine on remand, and within the context of the new trial as it develops, whether to admit — subject to normal evidentiary rules such as relevance, hearsay and cumulativeness — additional documents that might shed light on Ford's financial condition.

III. Conclusion

We reverse for a new trial on punitive damages so that the district court can provide a proper harm to nonparties jury instruction, and instruct the jury that the Whites received $2,305,435 in compensatory damages, that Ford's brake defect did not proximately cause the accident and that the Whites were found to be 40 percent responsible for the accident. Accordingly, we do not address whether the jury's $52 million award was constitutionally excessive. See White I, 312 F.3d at 1020.

The district court's judgment is REVERSED and REMANDED for a new trial on punitive damages. Each party shall bear its own costs on appeal.

REVERSED and REMANDED.