I, of course, agree with the majority opinion as far as it goes. I write separately to address the punitive damages issue in greater detail. In my view, it is not enough merely to reaffirm that the standards for reviewing a new trial order are deferential and then to dispose of this case on that ground. The Legislature has expressly authorized trial courts to grant new trials when damages are “excessive” (Code Civ. Proc., § 657 (section 657)), and the trial court here granted a new trial as to punitive damages on that ground. Our role as the final interpreter of state law obligates us to give meaning to the term “excessive” in relation to punitive *422damages. Otherwise, the deferential standards of review that we reaffirm today will inhibit us in providing guidance to trial courts as to how their broad discretion should be exercised. (See, e.g., Neal v. Farmers Ins. Exchange (1978) 21 Cal.3d 910, 932 [148 Cal.Rptr. 389, 582 P.2d 980] (Neal) [discussing factors relevant to assessing the size of a punitive damage award though the rule of deference would have provided a sufficient basis to affirm].) In other words, though we defer to trial court rulings, our deference does not free us from the obligation to inform those courts of the considerations we think relevant to their decisions. The point is particularly significant here because, without clear guidance, judicial oversight of punitive damage awards will not be uniform, and the resulting gross variability in awards will ultimately implicate the rights of defendants to due process and the equal protection of the laws under our state Constitution. (Cal. Const., art. I, § 7, subd. (a).)
Here, the trial court found excessive under section 657 a punitive damage award of $80 million. Plaintiffs argue Hughes Aircraft Company’s wealth justifies this large award. The wealth of a defendant is clearly relevant when determining the size of a punitive damage award, as is the amount of the compensatory damages and the reprehensibility of the defendant’s conduct. (Neal, supra, 21 Cal.3d at p. 928; see also Adams v. Murakami (1991) 54 Cal.3d 105, 110-111 [284 Cal.Rptr. 318, 813 P.2d 1348].) But merely to state what is “relevant” offers little in the way of concrete guidance. Without a beginning point of reference, punitive damage awards tend to vary widely, depending as much on the vicissitudes of jury decisionmaking as on the facts of the case. This variability can result in disparate and fundamentally unfair treatment of similarly situated defendants, a possibility we should not ignore.
Significantly, the variability in punitive damage awards does not flow so much from any inherent inability of different juries to agree on the wrongfulness of specific conduct. Rather, it results from the way courts ask juries to measure that wrongfulness. If the same evidence were presented to 10 separate juries and each were asked to decide a concrete question of historical fact, we would likely see a reassuring level of uniformity; each jury would be selecting the most plausible fact scenario from a fixed set of options based on specific evidence. Similarly, if 10 juries were asked to rate the egregiousness of a defendant’s conduct on a scale of 1 to 6, we again would likely see a reassuring level of uniformity; in that case, the range would be “bounded and anchored ... at both ends” and thus the valuations of all juries would be calibrated to the same scale. (Sunstein et al., Assessing Punitive Damages (with Notes on Cognition and Valuation in Law) (1998) 107 Yale L.J. 2071, 2106 (Assessing Punitives); see generally id. at pp. 2098-2100.) But if we were to ask the same juries to assess the appropriate *423dollar amount of punitive damages, we would likely see widely varying valuations because, in that case, the upper end of the scale would be “unbounded” while the lower end would remain “defined by a meaningful zero point.” (Id. at p. 2106; see generally id. at pp. 2103, 2105.) This phenomenon, which results from the use of a “magnitude scale” without a “modulus,” is widely recognized by psychological researchers, who structure their tests in order to account for its effect. (Id. at p. 2106 & fn. 138.)
The variability in punitive damage awards does not, therefore, reflect any inconsistency in jury factfinding, merely that awards are not calibrated to a common scale. (Assessing Punitives, supra, 107 Yale L.J. at pp. 2106-2107.) Nevertheless, jury determinations of punitive damages are likely to contain an element of arbitrariness as long as the awards remain uncalibrated. To assure basic fairness, courts must consider ways of calibrating punitive damage awards. (Id. at p. 2110 [“To obtain the virtues associated with the rule of law, ... the legal system should counteract the arbitrariness that comes from the unbounded scale of dollars.”].) We must ask what anchoring variable will make an award of punitive damages an appropriate measure of punishment rather than a test of a jury’s ability to imagine big numbers.
There is no question that juries have broad discretion to determine damages. “The measure of damages suffered is a factual question and as such is a subject particularly within the province of the trier of fact.” (Bertero v. National General Corp. (1974) 13 Cal.3d 43, 65, fn. 12 [118 Cal.Rptr. 184, 529 P.2d 608, 65 A.L.R.3d 878].) But discretion must be distinguished from whim and serendipity. In the case of large awards, punitive damages should rarely exceed compensatory damages by more than a factor of three, and then only in the most egregious circumstances clearly evident in the record. In arguing for this standard, I do not mean to suggest that three times compensatory damages is a benchmark measure of punitive damages. (Cf. Assessing Punitives, supra, 107 Yale L.J. at p. 2127 & fn. 191.) Far from it. The standard is an uppermost limit, and most punitive damage awards should fall well below that limit. (See Eisenberg & Wells, Punitive Awards after BMW, a New Capping System, and the Reported Opinion Bias (1998) Wis. L.Rev. 387, 420, 422 (Punitive Awards) [charting punitive damage awards for all states in 1992 and for California from 1960-1984].)
Commentators have criticized reforms of the kind suggested here, of course. The chief criticism is that, though some sort of calibration of awards is necessary to ensure fairness to defendants, use of compensatory damages as the calibrating factor is arbitrary. (See, e.g., Assessing Punitives, supra, 107 Yale L.J. at p. 2127; Polinsky & Shavell, Punitive Damages: An Economic Analysis (1998) 111 Harv. L.Rev. 869, 900 (Economic Analysis).) In other words, if the purposes of punitive damages are deterrence and *424retribution, then the amount of compensatory damages—the amount of harm the defendant caused the plaintiff—is irrelevant. For example, if a defendant can easily conceal its conduct or otherwise escape responsibility, a large punitive damage award may be necessary to achieve the appropriate level of deterrence, even if the amount of harm to the plaintiff is relatively small. Similarly, if a defendant does something particularly reprehensible, such as firing a gun into a crowd, a large punitive damage award may be appropriate, even if the bullet providentially lodges harmlessly in a tree. (See, e.g., 107 Yale L.J. at p. 2127.)
However, while deterrence and retribution are part of the equation, fundamental notions of justice require some correlation between punishment and harm. Getting too little good is suffering injustice, and getting too much is doing injustice. (Aristotle, Nicomachean Ethics (Hackett Pub. Co. 1985) p. 135.) We expect the punishment to fit the crime. Thus, the use of compensatory damages as a calibrating factor, even if imperfect, is not wholly arbitrary because the degree of harm to the plaintiff bears a substantial relationship to the appropriate punishment. (Neal, supra, 21 Cal.3d at p. 928.) The use of compensatory damages as a calibrating factor has a long history, dating back at least 700 years. (BMW of North America, Inc. v. Gore (1996) 517 U.S. 559, 581 [116 S.Ct. 1589, 1601, 134 L.Ed.2d 809].) Juries naturally select compensatory damages as a calibrating factor when determining punitive damages. In fact, statistical studies establish that “[a]bout half the variance in punitive awards is explained by the level of compensatory awards.” (Punitive Awards, supra, (1998) Wis. L.Rev. at p. 393, fn. omitted.) This correlation may indicate the extent to which the adage “no harm, no foul” embodies widely shared notions of basic fairness in our society.
Moreover, large compensatory damage awards not based on a defendant’s ill-gotten gains have a strong deterrent and punitive effect in themselves. The magnitude of such awards should be considered in deciding whether and to what extent punitive damages should be imposed. We ourselves have pointed the way. “Punitive damages can be justified only as a deterrent measure or as retribution.” (Mirkin v. Wasserman (1993) 5 Cal.4th 1082, 1106 [23 Cal.Rptr.2d 101, 858 P.2d 568].) But “the additional deterrent value of punitive damages” might not be needed if the “actual damages, alone, represent a potentially crushing liability . . . .” (Ibid.) We should follow the lead of other courts in implementing this observation. (See, e.g., Beliz v. W.H. McLeod & Sons Packing Co. (5th Cir. 1985) 765 F.2d 1317, 1332 [“Deterrent effect may be achieved without awarding exemplary damages. Whether an award accomplishes this purpose, in addition to affording compensation, is determined by considering not only the amount allowed to each plaintiff . . . but also the total amount of the award . . . .”]; Maiorino *425v. Schering-Plough Corp. (1997) 302 N.J.Super. 323 [695 A.2d 353] [Punitive damage award not necessary where “the large compensatory damage award ... by itself provided significant deterrence even to an employer as large as [the defendant]”].)
This dual effect of compensatory damages is especially strong when there is a large award for noneconomic harms. This point is well made in the Sunstein law review article; “[Although pain-and-suffering awards are essentially compensatory, there can be little doubt that such awards sometimes reflect jury judgments about the egregiousness of the defendant’s behavior. Hence such judgments are likely to have a punitive component.” (Assessing Punitives, supra, 107 Yale L.J. at p. 2133.) The trial court should instruct the jury that in determining whether to impose a punitive damage award and, if so, its magnitude, the jury should consider the amount and punitive or deterrent effect of its compensatory damage award. Only if the jury finds the compensatory damage award insufficient to punish or deter should an additional punitive award be imposed. And if imposed, the punitive award must be limited to an amount needed to deter and punish when added to the sum already imposed as compensatory damages. Courts, too, should consider this same factor in deciding whether an award is excessive and, perhaps, in deciding whether to present the question of punitive damages to the jury in the first place.
We already ask juries to consider the reprehensibility of the defendant’s conduct. Some commentators argue juries should also consider the defendant’s ability to escape responsibility. (Economic Analysis, supra, 111 Harv. L.Rev. at pp. 898-899; but see Assessing Punitives, supra, 107 Yale L.J. at pp. 2111-2112.) Though these considerations might provide important guidance, they are not anchors that can serve to calibrate a jury’s award to a common scale in the way compensatory damages can.
There is a related benchmark trial courts should consider in administering section 657. In more than 30 instances, the Legislature has provided for double or treble damages as a punishment for wrongful acts. (See, e.g., Bus. & Prof. Code, §§ 17537.4 [unlawful advertising], 21140.4 [violations of regulations governing fuel franchises]; Civ. Code, §§ 1812.9 [willful violation of laws governing retail installment sales], 1947.10 [evictions based.on fraudulent intent to occupy], 3345 [unfair or deceptive practices against senior citizens or disabled persons]; Lab. Code, § 206 [failure to pay certain wages].) Indeed, we are unable to find any context in which it has mandated a greater multiplier than three, notwithstanding the egregiousness of the wrong. Although none of these statutes are directly applicable here, “a legislative policy [may be] a source of law” (Van Beech v. Sabine Towing Co. (1937) 300 U.S. 342, 351 [57 S.Ct. 452, 456, 81 L.Ed. 685]), and statutes “are to be regarded as starting points of reasoning.” (Pope v. Atlantic Coast *426Line R. Co. (1953) 345 U.S. 379, 390 [73 S.Ct. 749, 755, 97 L.Ed. 1094] (dis. opn. of Frankfurter, J.).) Thus, “[c]ourts that fully appreciate their . . . obligations with respect to decisional law will not be deterred from a desirable result by inference from legislative silence, but only by . some indication of legislative intent to preclude the proposed result.” (Williams, Statutes as Sources of Law Beyond Their Terms in Common-Law Cases (1982) 50 Geo. Wash. L.Rev. 554, 567, fns. omitted; Moragne v. States Marine Lines, Inc. (1970) 398 U.S. 375, 392 [90 S.Ct. 1772, 1783, 26 L.Ed.2d 339] [“In many cases the scope of a statute may reflect nothing more than the dimensions of the particular problem that came to the attention of the legislature, inviting the conclusion that the legislative policy is equally applicable to other situations in which the mischief is identical.”].)
The statutory web anchors two observations. First, the Legislature has selected compensatory damages as the best, albeit imperfect, metric for calibrating the punitive component of a damages award. Second, the Legislature has determined that double or treble damages are, in most circumstances, sufficiently punitive. (See also Civ. Code, § 3426.3, subd. (c) [limiting punitive damages under the Uniform Trade Secrets Act to two times other damages].) Because punitive damages are “in addition to . . . actual damages” (Civ. Code, § 3294), an award of treble damages corresponds to a punitive damage award that is two times compensatory damages. Thus, the conclusion that large punitive damage awards should rarely exceed three times compensatory damages gives significantly greater leeway to juries than the Legislature has seen fit to give in these analogous contexts.
The statutes of other jurisdictions are also relevant. (See In re Waltreus (1965) 62 Cal.2d 218, 224 [42 Cal.Rptr. 9, 397 P.2d 1001] [borrowing from federal law in order to develop our common law rules of criminal procedure].) It is significant that our sister states have most frequently selected two or three times compensatory damages as the appropriate limitation on punitive damages. These jurisdictions include Connecticut, Delaware, Florida, Illinois, Indiana, Nevada, North Dakota, Oklahoma and Texas. (See BMW of North America, Inc. v. Gore, supra, 517 U.S. at pp. 615-616 [116 S.Ct. at pp. 1618-1619] (appen. to dis. opn. of Ginsburg, J.).) Moreover, several jurisdictions have imposed stricter dollar limits (ibid.), even in the case of intentional and malicious discrimination. (See 42 U.S.C. § 1981a(b)(3) [upper limit of $300,000].) Thus, the limitation I would urge the courts to consider is a relatively modest one, allowing ample flexibility by authorizing courts to approve awards that exceed the limit where appropriate.
At trial, plaintiffs stressed that a defendant’s wealth is relevant to determining punitive damages, and argued the extent of that wealth provides an *427adequate upper limit on a punitive damage award. Wealth is relevant, but it cannot provide an adequate upper limit. Many of the wealthiest defendants are corporations, and the size of a corporate defendant is not an additional evil that in itself warrants an enhanced penalty. (Economic Analysis, supra, 111 Harv. L.Rev. at pp. 910-914.) In order to remain competitive, large corporations count small as well as big costs, particularly if those small costs may be recurrent. Therefore, even for a large corporation, a relatively modest punitive damage award may be sufficient to induce an end to the offensive conduct. Moreover, above a certain level, the precise amount of a defendant’s wealth becomes less relevant, compared to other factors, in determining an appropriate punitive damage award. Accordingly, trial courts should not permit attorneys to argue, as plaintiffs’ attorney argued here, that punitive damages should be a fixed percentage of the defendant’s total net worth.
On the other hand, punitive damage awards should not be a routine cost of doing business that an industry can simply pass on to its customers through price increases, while continuing the conduct the law proscribes. In contract law, we recognize “ ‘efficient breach[]’ ” as being socially beneficial (Freeman & Mills, Inc. v. Belcher Oil Co. (1995) 11 Cal.4th 85, 98 [44 Cal.Rptr.2d 420, 900 P.2d 669]), but in the law of punitive damages there is no such thing as “efficient” oppression, fraud or malice. The purpose of punitive damages is to prevent oppression, fraud and malice, not merely to force defendants to internalize the social costs of that conduct. Or, put more accurately, the hidden or unquantifiable social costs of oppression, fraud and malice are so high that the Legislature has determined such conduct has no social benefit. In sum, the law of punitive damages does not punish a large corporation simply for being large; it takes wealth into consideration so as to ensure the award creates an adequate deterrent, even though the award may still be small in relation to the corporation’s net worth. (See, e.g., Neal, supra, 21 Cal.3d at p. 928 and cases cited therein; see also BAJI No. 14.72.2 (8th ed. 1994).) In light of these principles, the full extent of a defendant’s wealth provides too unbounded an upper limit to be a suitable means to calibrate punitive damage awards. (See 42 U.S.C. § 1981a(b)(3) [taking defendant’s size into consideration, but imposing $300,000 limit applicable regardless of size].)
We read often in the popular press of large punitive damage awards. In practice, however, such awards rarely exceed three times compensatory damages. (See, e.g., Punitive Awards, supra, Wis. L.Rev. at pp. 420, 422.) Commentators have cited this lack of a punitive damages crisis as a reason to go slow with respect to reforms. (See, e.g., Rustad, Unraveling Punitive Damages: Current Data and Further Inquiry (1998) Wis. L.Rev. 15, 69.) But arbitrariness in punitive damage awards, even if only occasional, does not *428affect jury verdicts alone; it distorts settlements by adding disproportionate leverage to a plaintiff’s demands, as well as uncertainty. Rather than weighing in favor of no reform, the absence of a punitive damages crisis establishes that a “soft ceiling” is unobjectionable—it will have a direct impact on relatively few cases. The standard advanced here would not affect small awards at all, and in the case of large awards, it would screen out only those that are especially suspect because of the extent they exceed the level of the plaintiff’s harm. And even in those few cases, no firm limit is proposed; rather, a suggestion for greater scrutiny of the award, whereby the trial court would favorably consider a new trial motion unless it could articulate specific evidence in the record justifying the high award.
We need not decide the precise measure of punitive damages in this case, but in my view the foregoing discussion is essential to the new trial issue that is before us. When setting punitive damages, a jury does not have the perspective, and the resulting sense of proportionality, that a court has after observing many trials. One of the primary checks against excessive punitive damage awards, essential to ensuring procedural fairness with respect to these awards, is the discretion—and, in some cases, the duty—of trial judges to order new trials. (See, e.g., Schelbauer v. Butler Manufacturing Co. (1984) 35 Cal.3d 442, 453 [198 Cal.Rptr. 155, 673 P.2d 743, 38 A.L.R.4th 566] [“trial courts have traditionally exercised the authority to reduce excessive punitive damage awards”].) As California’s highest court, it is our responsibility to give guidance to the trial courts.
Justice Mosk writes separately, not to comment on the majority opinion, but to accuse me of a “generalized attack on punitive damages.” (Conc. opn. of Mosk, J., ante, at p. 419.) He misapprehends the point of this exercise. The Legislature has authorized the award of punitive damages—an act clearly within its purview. By enacting section 657, the Legislature has also, as Justice Mosk concedes, conferred on courts the responsibility to define and restrain excessive damages, including punitive damages. (Conc. opn. of Mosk, J., ante, at p. 421.) Given the infinite range of misconduct and harm which can give rise to claims for punitive damages, it seems both reasonable and necessary to leave the fine-tuning to the courts. It seems equally reasonable for the courts, in developing case law pursuant to the statute, to look to what the Legislature has done in an analogous context. That context, of course, is the civil penalty statutes in which ceilings on recoverable damages have been imposed. These statutes furnish a reliable guide to what the Legislature regards as not excessive. And the fact the Legislature imposed no damages limitation in particular penalty statutes says nothing about the kind of limitations appropriate in the case of punitive damages. Why? Because the Legislature has already imposed a statutory ceiling on such awards by expressly providing for new trials in the case of “excessive damages.”
*429The target of “attack” is thus not punitive damages, but arbitrariness—a due process concern. Nor does the standard outlined here ignore the other Neal factors; it acknowledges that particularly egregious circumstances may justify very large punitive damage awards. By asking trial judges to spell out those circumstances, it gives appellate courts something concrete to review and provides the common law the opportunity to define the kinds of outrageous conduct that should be punished more harshly.
Because this opinion attempts to formulate a common law standard by drawing inferences from analogous legislative activity, Justice Mosk brands it “judicial lawmaking.” (Conc. opn. of Mosk, J., ante, at p. 416.) The language is unfortunate because it betrays a misconception of the responsibility of judges in the common law tradition. When exercising common law powers, all judges “make law,” for the techniques of the common law involve reaching beyond positive law for a standard to guide the creation of a new legal rule or to supplement an old one. The real question is not whether judges “make law,” it is whether they respect the boundaries imposed on their discretion by precedent and statute, and by constitutional text, design and tradition. It is arrogance, carelessness, and a lack of candor that constitute impermissible judicial practice, not creativity.
As noted, the Legislature has expressly authorized trial courts to grant new trials when damages are “excessive.” The standard proposed here would provide trial courts much needed guidance as to the meaning of that word, removing a specter of arbitrariness that hovers over punitive damage awards without, in most cases, substantively affecting the result. Here, defendant based its new trial motion in part on excessive punitive damages. With respect to this issue, I would hold that, in the case of large awards, punitive damages should rarely exceed compensatory damages by more than a factor of three, and then only in the most egregious circumstances clearly evident in the record. By this standard, the punitive damages awarded in this case were excessive, and therefore the new trial order was proper.
Chin, J., concurred.
The petition of all appellants for a rehearing was denied May 10, 2000, and the opinion was modified to read as printed above. Mosk, J., was of the opinion that the petition should be granted.