POSNER, Circuit Judge.

This is a tort case under the federal admiralty jurisdiction. We are called upon to decide questions of contributory negligence and damage assessment, in particular the question — one of first impression in this circuit — whether, and if so how, to account for inflation in computing lost future wages.

On the day of the accident, Margaret O’Shea was coming off duty as a cook on a towboat plying the Mississippi River. A harbor boat operated by the defendant, Riverway Towing Company, carried Mrs. O’Shea to shore and while getting off the boat she fell and sustained the injury complained of. The district judge found River-way negligent and Mrs. O’Shea free from contributory negligence, and assessed • damages in excess of $150,000. Riverway appeals only from the finding that there was no contributory negligence and from the part of the damage award that was intended to compensate Mrs. O’Shea for her lost future wages.

The accident happened in the following way. When the harbor boat reached shore it tied up to a seawall the top of which was several feet above the boat’s deck. There was no ladder. The other passengers, who were seamen, clambered up the seawall without difficulty, but Mrs. O’Shea, a 57-year-old woman who weighs 200 pounds (she is five foot seven), balked. According to Mrs. O’Shea’s testimony, which the district court believed, a deckhand instructed her to climb the stairs to a catwalk above the deck and disembark from there. But the catwalk was three feet above the top of the seawall, and again there was no ladder. The deckhand told her that she should jump and that the men who had already disembarked would help her land safely. She did as told, but fell in landing, carrying the assisting seamen down with her, and broke her leg.

Riverway concedes that the instruction to Mrs. O’Shea to jump was negligent, but argues that she was contributorily negligent to jump down the three feet to the seawall, given her age and weight, and therefore that her damages should have been reduced. But if her testimony is believed — and Riverway does not ask us to redetermine any issue of credibility — the finding that she was not contributorily negligent cannot be said to be clearly erroneous. There she was on the boat with no apparent way to get off. The deckhand told her to jump and that the men already on the shore would make sure she landed safely. It was reasonable for her to assume that Riverway knew how to disembark passengers, even overweight middle-aged female passengers, safely in the absence of a ladder (with which the boat curiously was not equipped). The deckhand’s apparent expertise made the risks as they reasonably appeared to Mrs. O’Shea small relative to the inconvenience of remaining aboard indefinitely, at some cost in embarrassment as well as time, while the crew rigged some alternative method of exit tailored to her lack of agility.

The more substantial issues in this appeal relate to the computation of lost wages. Mrs. O’Shea’s job as a cook paid her $40 a day, and since the custom was to work 30 days consecutively and then have the next 30 days off, this comes to $7200 a year although, as we shall see, she never had earned that much in a single year. She testified that when the accident occurred she had been about to get another cook’s job on a Mississippi towboat that would have paid her $60 a day ($10,800 a year). She also testified that she had been intending to work as a boat’s cook until she was 70 — longer if she was able. An economist who testified on Mrs. O’Shea’s behalf used the foregoing testimony as the basis for estimating the wages that she lost because *1197of the accident. He first subtracted federal income tax from yearly wage estimates based on alternative assumptions about her wage rate (that it would be either $40 or $60 a day); assumed that this wage would have grown by between six and eight percent a year; assumed that she would have worked either to age 65 or to age 70; and then discounted the resulting lost-wage estimates to present value, using a discount rate of 8.5 percent a year. These calculations, being based on alternative assumptions concerning starting wage rate, annual wage increases, and length of employment, yielded a range of values rather than a single value. The bottom of the range was $50,000. This is the present value, computed at an 8.5 percent discount rate, of Mrs. O’Shea’s lost future wages on the assumption that her starting wage was $40 a day and that it would have grown by six percent a year until she retired at the age of 65. The top of the range was $114,000, which is the present value (again discounted at 8.5 percent) of her lost future wages assuming she would have worked till she was 70 at a wage that would have started at $60 a day and increased by eight percent a year. The judge awarded a figure — $86,-033 — near the midpoint of this range. He did not explain in his written opinion how he had arrived at this figure, but in a preceding oral opinion he stated that he was “not certain that she would work until age 70 at this type of work,” although “she certainly was entitled to” do so and “could have earned something”; and that he had not “felt bound by [the economist’s] figure of eight per cent increase in wages” and had “not found the wages based on necessarily a 60 dollar a day job.” If this can be taken to mean that he thought Mrs. O’Shea would probably have worked till she was 70, starting at $40 a day but moving up from there at six rather than eight percent a year, the economist’s estimate of the present value of her lost future wages would be $75,000.

There is no doubt that the accident disabled Mrs. O’Shea from working as a cook on a boat. The break in her leg was very serious: it reduced the stability of the leg and caused her to fall frequently. It is impossible to see how she could have continued working as a cook, a job performed mostly while standing up, and especially on a boat, with its unsteady motion. But Riverway argues that Mrs. O’Shea (who has not worked at all since the accident, which occurred two years before the trial) could have gotten some sort of job and that the wages in that job should be deducted from the admittedly higher wages that she could have earned as a cook on a boat.

The question is not whether Mrs. O’Shea is totally disabled in the sense, relevant to social security disability cases but not tort cases, that there is no job in the American economy for which she is medically fit. Compare Cummins v. Schweiker, 670 F.2d 81 (7th Cir. 1982), with New Orleans (Gulfwide) Stevedores v. Turner, 661 F.2d 1031, 1037-38 (5th Cir. 1981). It is whether she can by reasonable diligence find gainful employment, given the physical condition in which the accident left her. See, e.g., Baker v. Baltimore & Ohio R. R., 502 F.2d 638, 644 (6th Cir. 1974). Here is a middle-aged woman, very overweight, badly scarred on one arm and one leg, unsteady on her feet, in constant and serious pain from the accident, with no education beyond high school and no work skills other than cooking, a job that happens to require standing for long periods which she is incapable of doing. It seems unlikely that someone in this condition could find gainful work at the minimum wage. True, the probability is not zero; and a better procedure, therefore, might have been to subtract from Mrs. O’Shea’s lost future wages as a boat’s cook the wages in some other job, discounted (i.e., multiplied) by the probability — very low — that she would in fact be able to get another job. But the district judge cannot be criticized for having failed to use a procedure not suggested by either party. The question put to him was the dichotomous one, would she or would she not get another job if she made reasonable efforts to do so? This required him to decide whether there was a more than 50 percent probability that she would. We cannot say that the negative answer he gave to that question was clearly erroneous.

*1198 Riverway argues next that it was wrong for the judge to award damages on the basis of a wage not validated, as it were, by at least a year’s employment at that wage. Mrs. O’Shea had never worked full time, had never in fact earned more than $3600 in a full year, and in the year preceding the accident had earned only $900. But previous wages do not put a cap on an award of lost future wages. If a man who had never worked in his life graduated from law school, began working at a law firm at an annual salary of $35,000, and was killed the second day on the job, his lack of a past wage history would be irrelevant to computing his lost future wages. The present case is similar if less dramatic. Mrs. O’Shea did not work at all until 1974, when her husband died. She then lived on her inheritance and worked at a variety of part-time jobs till January 1979, when she started working as a cook on the towboat. According to her testimony, which the trial judge believed, she was then working full time. It is immaterial that this was her first full-time job and that the accident occurred before she had held it for a full year. Her job history was typical of women who return to the labor force after their children are grown or, as in Mrs. O’Shea’s case, after their husband dies, and these women are, like any tort victims, entitled to damages based on what they would have earned in the future rather than on what they may or may not have earned in the past.

If we are correct so far, Mrs. O’Shea was entitled to have her lost wages determined on the assumption that she would have earned at least $7200 in the first year after the accident and that the accident caused her to lose that entire amount by disabling her from any gainful employment. And since Riverway neither challenges the district judge’s (apparent) finding that Mrs. O’Shea would have worked till she was 70 nor contends that the lost wages for each year until then should be discounted by the probability that she would in fact have been alive and working as a boat’s cook throughout the damage period, we may also assume that her wages would have been at least $7200 a year for the 12 years between the date of the accident and her seventieth birthday. But Riverway does argue that we cannot assume she might have earned $10,800 a year rather than $7200, despite her testimony that at the time of the accident she was about to take another job as a boat’s cook where she would have been paid at the rate of $60 rather than $40 a day. The point is not terribly important since the trial judge gave little weight to this testimony, but we shall discuss it briefly. Mrs. O’Shea was asked on direct examination what “pay you would have worked” for in the new job. Riverway’s counsel objected on the ground of hearsay, the judge overruled his objection, and she answered $60 a day. The objection was not well taken. Riverway argues that only her prospective employer knew what her wage was, and hence when she said it was $60 she was testifying to what he had told her. But an employee’s wage is as much in the personal knowledge of the employee as of the employer. If Mrs. O’Shea’s prospective employer had testified that he would have paid her $60, Riverway’s counsel could have made the converse hearsay objection that the employer was really testifying to what Mrs. O’Shea had told him she was willing to work for. Riverway’s counsel could on cross-examination have probed the basis for Mrs. O’Shea’s belief that she was going to get $60 a day in a new job, but he did not do so and cannot complain now that the judge may have given her testimony some (though little) weight.

We come at last to the most important issue in the case, which is the proper treatment of inflation in calculating lost future wages. Mrs. O’Shea’s economist based the six to eight percent range which he used to estimate future increases in the wages of a boat’s cook on the general pattern of wage increases in service occupations over the past 25 years. During the second half of this period the rate of inflation has been substantial and has accounted for much of the increase in nominal wages in this period; and to use that increase to project future wage increases is therefore to assume that inflation will continue, and continue to push up wages. Riverway argues *1199that it is improper as a matter of law to take inflation into account in projecting lost future wages. Yet Riverway itself wants to take inflation into account — one-sidedly, to reduce the amount of the damages computed. For Riverway does not object to the economist’s choice of an 8.5 percent discount rate for reducing Mrs. O’Shea’s lost future wages to present value, although the rate includes an allowance — a very large allowance — for inflation.

To explain, the object of discounting lost future wages to present value is to give the plaintiff an amount of money which, invested safely, will grow to a sum equal to those wages. So if we thought that but for the accident Mrs. O’Shea would have earned $7200 in 1990, and we were computing in 1980 (when this case was tried) her damages based on those lost earnings, we would need to determine the sum of money that, invested safely for a period of 10 years, would grow to $7200. Suppose that in 1980 the rate of interest on ultra-safe (i.e., federal government) bonds or notes maturing in 10 years was 12 percent. Then we would consult a table of present values to see what sum of money invested at 12 percent for 10 years would at the end of that time have grown to $7200. The answer is $2318. But a moment’s reflection will show that to give Mrs. O’Shea $2318 to compensate her for lost wages in 1990 would grossly undercompensate her. People demand 12 percent to lend money risklessly for 10 years because they expect their principal to have much less purchasing power when they get it back at the end of the time. In other words, when long-term interest rates are high, they are high in order to compensate lenders for the fact that they will be repaid in cheaper dollars. In periods when no inflation is anticipated, the risk-free interest rate is between one and three percent. See references in Doca v. Marina Mercante Nicaraguense, S.A., 634 F.2d 30, 39 n.2 (2d Cir. 1980). Additional percentage points above that level reflect inflation anticipated over the life of the loan. But if there is inflation it will affect wages as well as prices. Therefore to give Mrs. O’Shea $2318 today because that is the present value of $7200 10 years hence, computed at a discount rate — 12 percent — that consists mainly of an allowance for anticipated inflation, is in fact to give her less than she would have been earning then if she was earning $7200 on the date of the accident, even if the only wage increases she would have received would have been those necessary to keep pace with inflation.

There are (at least) two ways to deal with inflation in computing the present value of lost future wages. One is to take it out of both the wages and the discount rate — to say to Mrs. O’Shea, “we are going to calculate your probable wage in 1990 on the assumption, unrealistic as it is, that there will be zero inflation between now and then; and, to be consistent, we are going to discount the amount thus calculated by the interest rate that would be charged under the same assumption of zero inflation.” Thus, if we thought Mrs. O’Shea’s real (i.e., inflation-free) wage rate would not rise in the future, we would fix her lost earnings in 1990 as $7200 and, to be consistent, we would discount that to present (1980) value using an estimate of the real interest rate. At two percent, this procedure would yield a present value of $5906. Of course, she would not invest this money at a mere two percent. She would invest it at the much higher prevailing interest rate. But that would not give her a windfall; it would just enable her to replace her lost 1990 earnings with an amount equal to what she would in fact have earned in that year if inflation continues, as most people expect it to do. (If people did not expect continued inflation, long-term interest rates would be much lower; those rates impound investors’ inflationary expectations.)

An alternative approach, which yields the same result, is to use a (higher) discount rate based on the current risk-free 10-year interest rate, but apply that rate to an estimate of lost future wages that includes expected inflation. Contrary to Riverway’s argument, this projection would not require gazing into a crystal ball. The expected rate of inflation can, as just suggested, be read off from the current long-term interest rate. If that rate is 12 percent, and if as suggested earlier the real or inflation-*1200free interest rate is only one to three percent, this implies that the market is anticipating 9-11 percent inflation over the next 10 years, for a long-term interest rate is simply the sum of the real interest rate and the anticipated rate of inflation during the term.

Either approach to dealing with inflation is acceptable (they are, in fact, equivalent) and we by no means rule out others; but it is illogical and indefensible to build inflation into the discount rate yet ignore it in calculating the lost future wages that are to be discounted. That results in systematic undercompensation, just as building inflation into the estimate of future lost earnings and then discounting using the real rate of interest would systematically overcompensate. The former error is committed, we respectfully suggest, by those circuits, notably the Fifth, that refuse to allow inflation to be used in projecting lost future earnings but then use a discount rate that has built into it a large allowance for inflation. See, e.g., Culver v. Slater Boat Co., 644 F.2d 460, 464 (5th Cir. 1981) (using a 9.125 percent discount rate). We align ourselves instead with those circuits (a majority, see Doca v. Marina Mercante Nicaraguense, S.A., supra, 634 F.2d at 35- 36), notably the Second, that require that inflation be treated consistently in choosing a discount rate and in estimating the future lost wages to be discounted to present value using that rate. See id. at 36- 39. We note that in Byrd v. Reederei, 638 F.2d 1300, 1307-08 (5th Cir. 1981), a panel of the Fifth Circuit indicated misgivings over that circuit’s position and that rehearing en banc has been granted. 650 F.2d 1324 (1981).

Applying our analysis to the present case, we cannot pronounce the approach taken by the plaintiff’s economist unreasonable. He chose a discount rate — 8.5 percent — well above the real rate of interest, and therefore containing an allowance for inflation. Consistency required him to inflate Mrs. O’Shea’s starting wage as a boat’s cook in calculating her lost future wages, and he did so at a rate of six to eight percent a year. If this rate had been intended as a forecast of purely inflationary wage changes, his approach would be open to question, especially at the upper end of his range. For if the estimated rate of inflation were eight percent, the use of a discount rate of 8.5 percent would imply that the real rate of interest was only .5 percent, which is lower than most economists believe it to be for any substantial period of time. But wages do not rise just because of inflation. Mrs. O’Shea could expect her real wages as a boat’s cook to rise as she became more experienced and as average real wage rates throughout the economy rose, as they usually do over a decade or more. It would not be outlandish to assume that even if there were no inflation, Mrs. O’Shea’s wages would have risen by three percent a year. If we subtract that from the economist’s six to eight percent range, the inflation allowance built into his estimated future wage increases is only three to five percent; and when we subtract these figures from 8.5 percent we see that his implicit estimate of the real rate of interest was very high (3.5-5.5 percent). This means he was conservative, because the higher the discount rate used the lower the damages calculated.

If conservative in one sense, the economist was most liberal in another. He made no allowance for the fact that Mrs. O’Shea, whose health history quite apart from the accident is not outstanding, might very well not have survived — let alone survived and been working as a boat’s cook or in an equivalent job — until the age of 70. The damage award is a sum certain, but the lost future wages to which that award is equated by means of the discount rate are mere probabilities. If the probability of her being employed as a boat’s cook full time in 1990 was only 75 percent, for example, then her estimated wages in that year should have been multiplied by .75 to determine the value of the expectation that she lost as a result of the accident; and so with each of the other future years. Cf. Conte v. Flota Mercante del Estado, 277 F.2d 664, 670 (2d Cir. 1960). The economist did not do this, and by failing to do this he overstated the loss due to the accident.

*1201But Riverway does not make an issue of this aspect of the economist’s analysis. Nor of another: the economist selected the 8.5 percent figure for the discount rate because that was the current interest rate on Triple A 10-year state and municipal bonds, but it would not make sense in Mrs. O’Shea’s federal income tax bracket to invest in tax-free bonds. If he wanted to use nominal rather than real interest rates and wage increases (as we said was proper), the economist should have used a higher discount rate and a higher expected rate of inflation. But as these adjustments would have been largely or entirely offsetting, the failure to make them was not a critical error.

Although we are not entirely satisfied with the economic analysis on which the judge, in the absence of any other evidence of the present value of Mrs. O’Shea’s lost future wages, must have relied heavily, we recognize that the exactness which economic analysis rigorously pursued appears to offer is, at least in the litigation setting, somewhat delusive. Therefore, we will not reverse an award of damages for lost wages because of questionable assumptions unless it yields an unreasonable result — especially when, as in the present case, the defendant does not offer any economic evidence himself and does not object to the questionable steps in the plaintiff’s economic analysis. We cannot say the result here was unreasonable. If the economist’s method of estimating damages was too generous to Mrs. O’Shea in one important respect it was, as we have seen, niggardly in another. Another error against Mrs. O’Shea should be noted: the economist should not have deducted her entire income tax liability in estimating her future lost wages. Cf. Norfolk & W. Ry. v. Liepelt, 444 U.S. 490, 495, 100 S.Ct. 755, 758, 62 L.Ed.2d 689 (1980). While it is true that the damage award is not taxable, the interest she earns on it will be (a point the economist may have ignored because of his erroneous assumption that she would invest the award in tax-exempt bonds), so that his method involved an element of double taxation.

If we assume that Mrs. O’Shea could have expected a three percent annual increase in her real wages from a base of $7200, that the real risk-free rate of interest (and therefore the appropriate discount rate if we are considering only real wage increases) is two percent, and that she would have worked till she was 70, the present value of her lost future wages would be $91,310. This figure ignores the fact that she did not have a 100 percent probability of actually working till age 70 as a boat’s cook, and fails to make the appropriate (though probably, in her bracket, very small) net income tax adjustment; but it also ignores the possibility, small but not totally negligible, that the proper base is really $10,800 rather than $7200.

So we cannot say that the figure arrived at by the judge, $86,033, was unreasonably high. But we are distressed that he made no attempt to explain how he had arrived at that figure, since it was not one contained in the economist’s testimony though it must in some way have been derived from that testimony. Unlike many other damage items in a personal injury case, notably pain and suffering, the calculation of damages for lost earnings can and should be an analytical rather than an intuitive undertaking. Therefore, compliance with Rule 52(a) of the Federal Rules of Civil Procedure requires that in a bench trial the district judge set out the steps by which he arrived at his award for lost future earnings, in order to assist the appellate court in reviewing the award. Cf. Rucker v. Higher Educ. Aids Bd., 669 F.2d 1179, 1183-84 (7th Cir. 1982). The district judge failed to do that here. We do not consider this reversible error, because our own analysis convinces us that the award of damages for lost future wages was reasonable. But for the future we ask the district judges in this circuit to indicate the steps by which they arrive at damage awards for lost future earnings.

JUDGMENT AFFIRMED.

LASKER, District Judge:

On June 7, 1971, an Allegheny Airlines flight crashed in fog while approaching New Haven Airport. Nancy Feldman, a passenger, died, in the crash. Allegheny conceded liability, and the parties submitted the issue of damages to Judge Blumenfeld of the United States District Court for the District of Connecticut.1 The airline appeals2 from Judge Blumenfeld’s judgment awarding $444,056. to Reid Laurence Feldman, as administrator of the estate of his late wife.

Determination of damages in this diversity, wrongful death action is governed by Connecticut law, specifically Conn.Gen.Stats. § 52-555, which measures recovery by the loss to the decedent of the value of her life rather than by the value of the estate she would have left had she lived a full life. Perry v. Allegheny Airlines, Inc., 489 F.2d 1349, 1351 (2d Cir. 1974); Floyd v. Fruit Industries, Inc., 144 Conn. 659, 669—671, 136 A.2d 918, 924 (1957). In accordance with Connecticut law, the judgment represented the sum of (1) the value of Mrs. Feldman’s lost earning capacity and (2) the destruction of her capacity to enjoy life’s non-remunerative activities, less (3) deductions for her necessary personal living expenses. No award was made for conscious pain and suffering before Mrs. Feldman’s death because the evidence on this point was too speculative, nor did the award include pre-judgment interest.

Damages in a wrongful death action must of necessity represent a crude monetary forecast of how the decedent’s life would have evolved. Prior to stating his specific findings, the district judge noted, and we agree, that “[t]he whole problem of assessing damages for wrongful death . . . defies any precise mathematical computation,” citing Floyd v. Fruit Industries, Inc., supra, 144 Conn. at 675, 136 A.2d at 927 (382 F.Supp. at 1282).

It is clear from Judge Blumenfeld’s remarkably detailed and precise analysis that he nevertheless made a prodigious effort to reduce the intangible elements of an award to measurable quantities. It is with reluctance, therefore, that we conclude that his determination of loss of earnings and personal living expenses must be remanded.

I.

Damages for Destruction of Earning Capacity.

Nancy Feldman was 25 years old at the time of her death. From 1968 until shortly before the plane crash, she lived and worked in New Haven while her husband studied at Yale Law School. On Mr. Feldman’s graduation from law school in the spring of 1971 the Feldmans moved to Washington, D. C., where they intended to settle. At the time of her death, Mrs. Feldman had neither accepted nor formally applied for employment in Washington, although she had been accepted by George Washington Law School for admission in the Fall of 1971 and had made inquiries about the availability of employment.

A key objection of appellant Allegheny runs to Judge Blumenfeld’s calculation of the discount rate at lV2% in determining the present value of Mrs. Feldman’s lost earning capacity on the grounds that the court has no right to take inflation into account in any way in its assessment of damages. The district court decided that the appropriate rate of discount would be the “price of capital,” such to be “obtained by adjusting inter*387est rates on ‘risk-free’ investments so as to exclude the additional interest demanded by the investment market as compensation for investors’ assumption of the risk of inflation.” 382 F.Supp. at 1293.

In calculating the discount rate, the appellee’s expert, relied on by the district court, used an average earnings of 4.14% (from mutual savings bank investments) as representative of a prudent, non-sophisticated investment and subtracted 2.87% as the average yearly inflation rate revealed in the Department of Labor’s Consumer Price Index over an 18-year period, yielding a 1.27% difference which was rounded up to 1.5%. Judge Blumenfeld corroborated this “inflation-adjusted discount rate” of 1.5% by calculating the real yields of investments since 1940 in federal government securities (with inflation factored out) from the 1974 Economic Report of the President, a source referred to by appellant Allegheny’s expert. The district court made this calculation according to its view of Connecticut’s law and policies on the subject of inflation accounting in wrongful death damages.

We agree with the district court’s interpretation of Connecticut law as leaving open the question how inflation may be accounted for in such damages.3 We believe that Judge Blumenfeld, a long-time Connecticut lawyer and district court judge for 14 years, appropriately hypothesized the Connecticut Supreme Court’s favorable reaction to a discount rate adjustment, since Connecticut, unlike most jurisdictions, reduces what would otherwise be inflated judgments for wrongful death injuries by requiring deduction of income taxes payable on future earnings. Floyd v. Fruit Industries, Inc., supra, 144 Conn. at 673, 136 A.2d at 926.

The district court was fully aware that in a way it was being speculative in what it was doing, as every trier of fact is required to some extent to be whenever it engages in calculating future earnings and a lump sum discount rate. 382 F.Supp. at 1291—92. As a matter of federal law we do not necessarily vouchsafe either the principle of making an “inflation adjustment” in setting a discount rate 4 or the means by which it was done in this instance. Yet we note that consideration of inflation has historically been approved in a number of state courts. See, e. g., Halloran v. New England Telephone & Telegraph Co., 95 Vt. 273, 276, 115 A. 143, 144 (1921) and cases cited in Judge Blumenfeld’s opinion, 382 F.Supp. at 1290. As a matter of federal law, at least one circuit has approved jury consideration of the impact of inflation and even reversed for charging that it should not consider “future increases or decreases in the purchasing power of money.” Bach v. Penn Central Transportation Co., 502 F.2d 1117, 1122 (6th Cir. 1974); see also Sleeman v. Chesapeake & Ohio Railway Co., 414 F.2d 305 (6th Cir. 1969). Our own Perry v. Allegheny Airlines, Inc., supra, 489 F.2d 1349, affirmed a $369,400. judgment on a jury verdict for the estate of another victim of the very same crash here involved; while the point was not discussed specifically in the opinion it is interesting that Judge Blumenfeld’s charge to the jury referred to the plaintiff’s expert’s testimony on a 1.5% discount rate and the underlying rationale therefor, a reference duly attacked on *388appeal by Allegheny. Commentators have supported an accounting for inflation in damage awards, see Econometrics and Damages, 44 Wash.L.Rev. 351, 360— 61 (1969); Comment, 6 U.S.F.L.Rev. 311 (1972). It has even been suggested that a trial court may be in error in failing to account fully for inflation in wrongful death damages in a non-diversity case. See Mills v. Tucker, 499 F.2d 866, 868 (9th Cir. 1974).5 As Judge Friendly himself said in McWeeney v. New York, New Haven and Hartford Railroad Co., 282 F.2d 34, 38 (2d Cir. 1960):

“There are few who do not regard some degree of continuing inflation as here to stay and would be willing to translate their own earning power into a fixed annuity, and it is scarcely to be expected that the average personal injury plaintiff will have the acumen to find investments that are proof against both inflation and depression — a task formidable for the most expert investor.” (Footnote omitted.)

Within the latitude afforded by the Connecticut decisions, note 3 supra, and with the support in the historical and other economic evidence before him that Judge Blumenfeld had, we cannot fault him for computing the discount rate by offsetting the anticipated rate of earnings from investment of the lump sum to be awarded, by an inflation factor.

In computing the value of Mrs. Feldman’s lost earning capacity, the trial judge found that Mrs. Feldman’s professional earnings in her first year of employment would have been $15,040. and that with the exception of eight years during which she intended to raise a family and to work only part time, she would have continued in full employment for forty years until she retired at age 65. The judge further found that during the period in which she would be principally occupied in raising her family, Mrs. Feldman would have remained sufficiently in contact with her profession to maintain, but not increase, her earning ability. Pointing out that under Connecticut law damages are to be based on “the loss of earning capacity, not future earnings per se . . . ” (382 F.Supp. at 1282) (emphasis in original), the judge concluded that when a person such as Mrs. Feldman, who possesses significant earning capacity, chooses to forego remunerative employment in order to raise a family, she manifestly values child rearing as highly as work in her chosen profession and her loss of the opportunity to engage in child rearing “may thus fairly be measured by reference to the earning capacity possessed by the decedent” (382 F.Supp. at 1283). Applying this rationale, the trial judge made an award for the eight year period of $17,044. per year, the salary which he computed Mrs. Feldman would have reached in the year preceding the first child-bearing year, but did not increase the amount during the period.

We believe the trial judge erred in automatically valuing Mrs. Feldman’s loss for the child-bearing period at the level of her salary. As Judge Blumenfeld’s opinion points out, the Connecticut cases distinguish clearly between loss of earning capacity and loss of capacity to carry on life’s non-remunerative activities. As we read Connecticut law, where a decedent suffers both kinds of loss for the same period each must be valued independently in relation to the elements particular to it.

The court in Floyd v. Fruit Industries, Inc., supra, equated “earning capacity” with “the capacity to carry on the particular activity of earning money.” 144 Conn, at 671, 136 A.2d at 925. Here the evidence established, and the trial court found, that Mrs. Feldman would have worked only part-time while raising a family. In the circumstances, we believe that under the Connecticut rule the plaintiff is entitled to recover “loss of earnings” for the child raising years only to the extent that the court finds that Mrs. Feldman would actually have *389worked during those years. For example, if the court finds that she would have worked 25% of the time during that period, the plaintiff would properly be credited only with 25% of her salary for each of the eight years.

This conclusion is consistent with the other leading authority in Connecticut. In Chase v. Fitzgerald, 132 Conn. 461, 45 A.2d 789 (1946), an award for “loss of future earnings” was denied in respect of a decedent who had been employed as a housekeeper, but who at the time of her death was a housewife with no intention of seeking outside employment. The court held that any award for wrongful death in such a case should be based not on the decedent’s loss of earning capacity, but rather on her “loss of the enjoyment of life’s activities.” 132 Conn. at 470, 45 A.2d at 793. Consistently with the holding in Chase, we conclude that any award in relation to the portion of the child-raising period during which Mrs. Feldman would not have been working must be predicated on her “loss of the enjoyment of life’s activities” rather than on loss of earnings, and on remand the district judge should reevaluate the elements accordingly.

We recognize that thus computed the total award for Mrs. Feldman’s child-raising years may be similar to that already made, but conclude that the conceptual framework we have described is required by Connecticut’s distinctive law of damages.

II.

Deductions for Decedent’s Necessary-Personal Living Expenses.

Where the decedent had been subject to the expense of self-maintenance, Connecticut case law provides for the deduction of “personal living expenses” from damages otherwise recoverable for the loss of earning capacity. Floyd v. Fruit Industries, Inc., supra, 144 Conn. at 674, 136 A.2d at 926. Judge Blumenfeld properly held that although a husband under Connecticut law has a duty to support his spouse, (see, e. g., Conn.Gen.Stats. §§ 46-10; 53-304), that duty does not exempt an income-earning wife from an obligation to apportion a part of her income for her own support. The Floyd court defined the term “personal living expenses” as:

“ . . . those personal expenses which, under the standard of living followed by a given decedent, it would have been reasonably necessary for him to incur in order to keep himself in such a condition of health and well-being that he could maintain his capacity to enjoy life’s activities, including the capacity to earn money.” 144 Conn. at 675, 136 A.2d at 926-927.

The trial judge concluded that, under Connecticut law, deductions for Mrs. Feldman’s personal living expenses should include the cost, at a level commensurate with her standard of living, of food, shelter, clothing and health care. The judge fixed such costs in Washington, D. C. for the year following her death at $2,750., increasing that figure by 3% per year to the age of retirement. After retirement, living expenses were deducted at the rate of $5,000. annually. These figures were discounted annually by 1.5% to reduce the deduction to present value. Although the process by which the trial judge determined the level of Mrs. Feldman’s living expenses was proper, we believe that he substantially underestimated the actual costs of food, shelter, clothing and health care.

On direct examination, Mr. Feldman testified that his wife’s personal living expenses in New Haven had been approximately $2,120. per year. On cross-examination, this figure was shown to have been unduly conservative with regard to clothing and food, and the trial judge rounded the amount to $2,200. He found that the Feldmans’ cost of living would have increased after they moved to Washington, where living expenses were higher and their social and economic status would have changed from that of students to that of young professionals. Accordingly, the judge adjusted the $2,200. figure upward by 25% for the *390first year Mrs. Feldman would have resided in Washington, and by 3% annually until she would have reached the age of sixty-five and retired. Personal living expenses for that year were calculated to be $6,675, but during the years of retirement deductions were lowered to $5,000., a level which the trial judge felt was consistent with a high standard of living but also reflected the fact that the cessation of work often produces a reduction in personal expenditures.

We recognize the perils involved in an appellate court dealing de novo with factual matters. We would not venture to do so in this case if we did not feel we have the right to take judicial notice of the facts of life, including the cost of living for those in the position of the Feldmans in such metropolitan areas as Washington, D. C. We reluctantly conclude that the trial judge was in error in computing living expenses at $2,750. for the year after Mrs. Feldman’s death, and building 6n that base for later years.

Without attempting to specify what the results of such a computation should be, we believe that it would fall more nearly in the area of $4,000., including approximately $25. per week for food, $125. per month for rent, $1,000. annually for clothing and $400. annually for health care. For one year the difference between the trial judge’s figure of $2,750. and the suggested figure of $4,000. may be considered de minimis in relation to the total award. However, projected over the 52 years of Mrs. Feldman’s life expectancy, and at an annual increase of 3%, the difference is sufficiently large to require us to remand the matter for further determination by the trial judge.

We have considered the other points raised by Allegheny and find them to be without merit.

The judgment is affirmed in part, reversed in part and remanded.

FRIENDLY, Circuit Judge

(concurring dubitante):

This case is another example of a federal court’s being compelled by the Congressional grant of diversity jurisdiction to determine a novel and important question of state law on which state decisions do not shed even a glimmer of light.1 The question here, how far awards of damages for disabling personal injury or for death shall attempt to make allowance for future inflation, is of great concern to the states since awards like that made here will further escalate the heavily mounting burden of liability insurance costs. The state decisions and the federal cases endeavoring to ascertain state law are in a stage of uncertainty and flux. So too are the decisions with respect to federal law. Compare Sleeman v. Chesapeake & Ohio Railway Co., 414 F.2d 305 (6 Cir. 1969), with Bach v. Penn Central Transportation Co., 502 F.2d 1117, 1122 (6 Cir. 1974). In a case of federal law, the Fifth Circuit recently granted en banc consideration and by a vote of twelve to three expressly disapproved a district court’s effort, in computing lost future earnings, to take account of possible inflationary trends over a period of several decades on the ground that “the influence on future damages of possible inflation or deflation is too speculative a matter for judicial determination.” Johnson v. Penrod Drilling Co., 510 F.2d 234, 236, 241 (5 Cir. 1975) (en banc), petition for certiorari filed, 43 U.S.L.W. 3684.

Both plaintiff’s expert and the court allowed for inflation not by building cost-of-living increases into future earnings but by applying a rate of only 1.5% in discounting to present value estimated *391lost future earnings and other recoverable values calculated in 1971 dollars. The district court derived this 1.5% figure by comparing rates of return on a number of “risk-free” securities issued by the federal government (plaintiff’s expert examined other types of “risk-free” fixed income investments and reached similar conclusions) since 1940 with rates of inflation during the same period as reflected by annual changes in the Consumer Price Index. The court subtracted the latter from the former on the theory that the latter amounted to that portion of the return representing what investors have historically demanded as protection against inflation. The difference varied from year to year, but the court determined that 1.5% was a representative figure for the period. This was deemed to be “that part of the annual yield which constitutes payment for the use of capital” or “real yield”— presumably the rate of return which investors would be willing to accept in an inflation-free economy; while the rate of inflation might rise and fall, investors could be expected to demand about 1.5% return on safe investments in addition to protection against expected inflation. The court recognized that other courts have been reluctant to take explicit account of the effects of future inflation (see 510 F.2d at 236 n.l) but stated that its approach in fact was “a means to avoid undue speculation” with respect to future inflation and even suggested that when, as in this case, the effects of future inflation have been expressly excluded in the calculation of the amount to be discounted, the “appropriate rate of discount” must necessarily be adjusted so that “the additional interest demanded by the investment market as compensation for investors’ assumption of the risk of inflation” is excluded. The distinction drawn between this method and the one more commonly used — adjusting the amount to be discounted so as to include a sum reflecting assumptions about future inflation — apparently given some weight by the majority — is more apparent than real. Plaintiff’s expert acknowledged that “another approach” or “alternative calculation” for this problem would be to increase estimated lost future compensation and living expenses to take account of the effects of future inflation and not reduce the rate of return used for discounting by any amount reflecting inflation. The outcome of the calculation under either approach would be very nearly identical. Indeed, at one point counsel for plaintiff asked his expert to calculate the present value in 1971 dollars of the deceased’s lost earnings based upon the “speculative” assumption of an inflation rate of 4.5% and a rate of return of 6%. This approach would have reduced the recovery by less than $2,000 out of approximately a quarter of a million dollars.

In any event, plaintiff’s expert came up with a $253,424 present value of Mrs. Feldman’s projected earnings. Using higher starting and ending salary figures and a different percentage deduction for income taxes, the judge arrived at an initial sum of $499,953 to which he added $100,000, admittedly drawn from the atmosphere, “for the destruction of the decedent’s capacity to enjoy life’s activities”, an element recognized as appropriate for consideration by Connecticut law, and from which he subtracted $155,897 as the discounted sum of personal living expenses, yielding a total recovery of $444,056. On the judge’s computations, Mrs. Feldman, who had been earning $10,000 a year at the time of her death in 1971, would be earning $33,757 in 1971 dollars in 2011 as a “legislative analyst” for the National League of Cities and United States Conference of Mayors (NLC/USCM), when she would have attained the age of 65. However, as counsel for Allegheny points out, without dispute from counsel for Mr. Feldman and apparently based upon the testimony of plaintiff’s expert,, about an alternative method of calculation discussed above, a calculation deducting 4.5% for inflation from a 6% interest rate assumed to be attainable on an investment free from risks other than inflation implicitly car*392ríes the prediction that Mrs. Feldman, and also all federal employees in the GS 16 — 7 category (which Mrs. Feldman hypothetically would reach after 40 years under the scheme for predicting merit pay increases adopted by the court), would in fact be earning $122,823 in the year 2011. Similar calculations based upon maintaining the 1.5% differential could yield even more striking results, which are largely veiled by the court’s approach. One point that immediately occurs is why, if Mrs. Feldman’s salary would rise to such a figure, income tax, deductible from damages under Connecticut law, should be computed at only 25%, a rate which the court found would achieve “substantial justice.” It is common knowledge that one effect of inflation is that the same progressive rates of income tax take an ever larger bite out of real income, and it is unlikely in the last degree that, in an era of increasing budgets, due in considerable part to inflation, Congress would make the accommodation needed to prevent this.

Save for this important point not urged by Allegheny and the two corrections made by the majority, I have no reason to question the meticulous calculations of the able district judge. Indeed one could argue that, at a time when the national goal is simply to bring back the golden age of single rather than double digit inflation, without too much question what the single digit should be, the entire interest return on otherwise risk-free investments, today probably in excess of 6%, represents compensation against the risk of inflation; in other words, investors in fixed income securities are willing, for the time being, to forego any return if they can keep the real amount of their investment intact. Indeed, insofar as the return is subject to income tax, they are not even achieving that. Yet common sense suggests that investors will not tolerate such a situation indefinitely.

I doubt whether judges, or anyone else, can peer so far into the future; the district court’s computations suffer from what Mr. Justice Holmes, in another context, called “[t]he dangers of a delusive exactness,” Truax v. Corrigan, 257 U.S. 312, 342, 42 S.Ct. 124, 133, 66 L.Ed. 254 (1921) (dissenting opinion). Instead of recognizing the plethora of uncertainties as the Fifth Circuit has done, see Johnson v. Penrod Drilling Co., supra, 510 F.2d at 236, compare Frankel v. United States, 321 F.Supp. 1331, 1346 (E.D.Pa.1970), aff’d, 466 F.2d 1226 (3 Cir. 1972), the court below endeavored to construct an iron-clad guaranty against the unknown and unknowable future effects of inflation. The estate of a young woman without dependents is hardly an outstanding candidate for a forty-year protection against inflation not enjoyed at all by millions of Americans who depend on pensions or investment income and not fully enjoyed by millions more whose salaries have in no wise kept pace with inflation.

The court necessarily assumed not only continued inflation, which unhappily seems likely in some degree, but continued responsiveness to it by equivalent wage increases. Yet we have seen in recent months that employers, particularly municipalities, simply cannot maintain these. Thousands of New York City’s employees have been dismissed and the rest are being subjected to a wage freeze. Other important cities may not be far behind in having to resort to similar measures. Under such conditions can we be sure that NLC/USCM would continue to grant automatic cost-of-living pay increases for 40 years, as the court assumed? Perhaps so, since the main business of NLC/USCM is seeking to obtain federal funds for cities, which surely is a boom industry if any there be; but perhaps not.

I would also question the likelihood— indeed, the certainty as found by the court — that, despite her ability, determination and apparent good health, Mrs. Feldman would have worked full time for forty years until attaining age 65, except for the eight years she was expected to devote to the bearing and early rearing of two children. Apart from *393the danger of disabling illness, temporary or permanent, there would be many attractions to which the wife of a successful lawyer might yield: devoting herself to various types of community service, badly needed but unpaid, or to political activity; accompanying her husband on business trips — often these days to far-off foreign countries; making pleasure trips for periods and at times of the year inconsistent with the demands of her job; perhaps, as the years went on, simply taking time off for reflection and enjoyment. Granted that in an increasing number of professional households both spouses work full time until retirement age, in more they do not. Surely some discount can and should be applied to the recovery for these reasons.

My guess is also that, even if inflation should be taken into account, neither a Connecticut nor a federal jury would have made an award as large as was made here. I say this despite the $369,-400 jury verdict for another death arising out of the same crash which we sustained in Perry v. Allegheny Airlines, Inc., supra, 489 F.2d 1349, where we did not expressly discuss the inflation question. Even though the existence of dependents is legally irrelevant under the Connecticut survival statute, a jury would hardly have ignored that, whereas Perry was survived by a dependent wife and five children ranging from 6 to 14 years in age, Mrs. Feldman had no dependents. More significant to me is that in Perry’s case the jury awarded only $369,400 as against the $535,000 estimate of Mrs. Perry’s expert for economic loss alone; here the judge was more generous in important respects than plaintiff’s expert.

However, I am loathe to require a busy federal judge to spend still more time on this diversity case, especially when I do not know what instructions to give him about Connecticut law. Some of the questions I have raised are not open for exploitation by the defendant since its own expert made his calculations on the basis that Mrs. Feldman would work until age 65. Although intuition tells me that the Supreme Court of Connecticut would not sustain the award made here, I cannot prove it. I therefore go along with the majority, although with the gravest doubts. I do this on the basis that, as far as I am concerned, the decision will not constitute a precedent on the inflation problem in a case arising under federal law. Judgments like Mr. Feldman’s and Mrs. Perry’s also inevitably raise serious policy questions with respect to damages in airline accident cases beyond those here considered, but these are for Congress and not for the courts.

POSNER, Chief Judge.

Jeffrey Kemezy sued a Muncie, Indiana policeman named James Peters under 42 U.S.C. § 1983, claiming that Peters had wantonly beaten him with the officer’s nightstick in an altercation in a bowling alley where Peters was moonlighting as a security guard. The jury awarded Kemezy $10,000 in compensatory damages and $20,000 in punitive damages. Peters’ appeal challenges only the award of punitive damages, and that on the narrowest of grounds: that it was the plaintiffs burden to introduce evidence concerning the defendant’s net worth for purposes of equipping the jury with information essential to a just measurement of punitive damages.

Two courts have adopted the position that Peters advocates. Adams v. Murakami, 54 Cal.3d 105, 284 Cal.Rptr. 318, 327-330, 813 P.2d 1348, 1357-60 (1991); Adel v. Parkhurst, 681 P.2d 886, 892 (Wyo.1984); and see the dissent in Keenan v. City of Philadelphia, 983 F.2d 459, 483-84 (3d Cir.1992). But the majority view is opposed, as noted in Hutchinson v. Stuckey, 952 F.2d 1418, 1422 *34n. 4 (D.C.Cir.1992); see, e.g., Smith v. Lightning Bolt Productions, Inc., 861 F.2d 363, 373 (2d Cir.1988); Fishman v. Clancy, 763 F.2d 485, 490 (1st Cir.1985); Woods-Drake v. Lundy, 667 F.2d 1198, 1203 n. 9 (5th Cir.1982). Our decision in Littlefield v. McGuffey, 954 F.2d 1337, 1349 (7th Cir.1992), can be read as aligning us with the majority, although as Peters points out the plaintiff there had presented some evidence of the defendant’s net worth and it is possible (though not necessary) to read our opinion as placing some minimal burden of production on the plaintiff. See id. at 1349-50. But we think the majority rule, which places no burden of production on the plaintiff, is sound, and we take this opportunity to make clear that it is indeed the law of this circuit.

The standard judicial formulation of the purpose of punitive damages is that it is to punish the defendant for reprehensible conduct and to deter him and others from engaging in similar conduct. E.g., Memphis Community School District v. Stachura, 477 U.S. 299, 307 n. 9, 106 S.Ct. 2537, 2542 n. 9, 91 L.Ed.2d 249 (1986); Smith v. Wade, 461 U.S. 30, 54, 103 S.Ct. 1625, 1639, 75 L.Ed.2d 632 (1983); City of Newport v. Fact Concerts, Inc., 453 U.S. 247, 266-67, 101 S.Ct. 2748, 2759-60, 69 L.Ed.2d 616 (1981); Gertz v. Robert Welch, Inc., 418 U.S. 323, 350, 94 S.Ct. 2997, 3012, 41 L.Ed.2d 789 (1974). This formulation is cryptic, since deterrence is a purpose of punishment, rather than, as the formulation implies, a parallel purpose, along with punishment itself, for imposing the specific form of punishment that is punitive damages. An extensive academic literature, however, elaborates on the cryptic judicial formula, offering a number of reasons for awards of punitive damages. See, e.g., Symposium: Punitive Damages, 40 Ala.L.Rev. 687 (1989); Symposium: Punitive Damages, 56 S.Cal.L.Rev. 1 (1982); 1 Dan B. Dobbs, Law of Remedies: Damages-Equity-Restitution § 3.11(3) (2d ed. 1993); William M. Landes & Richard A. Posner, The Economic Structure of Tort Law, ch. 6 (1987); W. Page Keeton et al., Prosser and Keeton on the Law of Torts § 2, pp. 9, 11-12 (5th ed. 1984). Some of these reasons are mentioned in our cases. See, e.g., Zazú Designs v. L’Oréal, S.A., 979 F.2d 499, 508 (7th Cir.1992); Fortino v. Quasar Co., 950 F.2d 389, 398 (7th Cir.1991); FDIC v. W.R. Grace & Co., 877 F.2d 614, 623 (7th Cir.1989). A review of the reasons will point us toward a sound choice between the majority and minority views.

1. Compensatory damages do not always compensate fully. Because courts insist that an award of compensatory damages have an objective basis in evidence, such awards are likely to fall short in some cases, especially when the injury is of an elusive or intangible character. If you spit upon another person in anger, you inflict a real injury but one exceedingly difficult to quantify. If the court is confident that the injurious conduct had no redeeming social value, so that “overdeter-ring” such conduct by an “excessive” award of damages is not a concern, a generous award of punitive damages will assure full compensation without impeding socially valuable conduct.

2. By the same token, punitive damages are necessary in such cases in order to make sure that tortious conduct is not underdet-erred, as it might be if compensatory damages fell short of the actual injury inflicted by the tort.

These two points bring out the close relation between the compensatory and deterrent objectives of tort law, or, more precisely perhaps, its rectificatory and regulatory purposes. Knowing that he will have to pay compensation for harm inflicted, the potential injurer will be deterred from inflicting that harm unless the benefits to him are greater. If we do not want him to balance costs and benefits in this fashion, we can add a dollop of punitive damages to make the costs greater.

3. Punitive damages are necessary in some cases to make sure that people channel transactions through the market when the costs of voluntary transactions are low. We do not want a person to be able to take his neighbor’s car and when the neighbor complains tell him to go sue for its value. Guido Calabresi & A. Douglas Melamed, “Property Rules, Liability Rules, and Inalienability: One View of the Cathedral,” 85 Harv.L.Rev. 1089, 1124-27 (1972). We want to make such expropriations valueless to the expropriator *35and we can do this by adding a punitive exaction to the judgment for the market value of what is taken. This function of punitive damages is particularly important in areas such as defamation and sexual assault, where the tortfeasor may, if the only price of the tort is having to compensate his victim, commit the tort because he derives greater pleasure from the act than the victim incurs pain.

4. When a tortious act is concealable, a judgment equal to the harm done by the act will underdeter. Suppose a person who goes around assaulting other people is caught only half the time. Then in comparing the costs, in the form of anticipated damages, of the assaults with the benefits to him, he will discount the costs (but not the benefits, because they are realized in every assault) by 50 percent, and so in deciding whether to commit the next assault he will not be confronted by the fall social cost of his activity.

5. An award of punitive damages expresses the community’s abhorrence at the defendant’s act. We understand that otherwise upright, decent, law-abiding people are sometimes careless and that their carelessness can result in unintentional injury for which compensation should be required. We react far more strongly to the deliberate or reckless wrongdoer, and an award of punitive damages commutes our indignation into a kind of civil fine, civil punishment.

Some of these functions are also performed by the criminal justice system. Many legal systems do not permit awards of punitive damages at all, believing that such awards anomalously intrude the principles of criminal justice into civil cases. Even our cousins the English allow punitive damages only in an excruciatingly narrow category of cases. See, e.g., AB v. South West Water Services Ltd., [1993] 1 All E.R. 609 (Ct.App.1992). But whether because the American legal and political cultures are unique, or because the criminal justice system in this country is overloaded and some of its functions have devolved upon the tort system, punitive damages are a regular feature of American tort cases, though reserved generally for intentional torts, including the deliberate use of excess force as here. This suggests additional functions of punitive damages:

6. Punitive damages relieve the pressures on the criminal justice system. They do this not so much by creating an additional sanction, which could be done by increasing the fines imposed in criminal cases, as by giving private individuals — the tort victims themselves — a monetary incentive to shoulder the costs of enforcement.

7. If we assume realistically that the criminal justice system could not or would not take up the slack if punitive damages were abolished, then they have the additional function of heading off breaches of the peace by giving individuals injured by relatively minor outrages a judicial remedy in lieu of the violent self-help to which they might resort if their complaints to the criminal justice authorities were certain to be ignored and they had no other legal remedy.

What is striking about the purposes that are served by the awarding of punitive damages is that none of them depends critically on proof that the defendant’s income or wealth exceeds some specified level. The more wealth the defendant has, the smaller is the relative bite that an award of punitive damages not actually geared to that wealth will take out of his pocketbook, while if he has very little wealth the award of punitive damages may exceed his ability to pay and perhaps drive him into bankruptcy. To a very rich person, the pain of having to pay a heavy award of damages may be a mere pinprick and so not deter him (or people like him) from continuing to engage in the same type of wrongdoing. Zazú Designs v. L’Oréal, S.A., supra, 979 F.2d at 508. What in economics is called the principle of diminishing marginal utility teaches, what is anyway obvious, that losing $1 is likely to cause less unhappiness (disutility) to a rich person than to a poor one. (This point, as the opinion in Zazú Designs emphasizes, does not apply to institutions as distinct from natural persons. Id. at 508-09.) But rich people are not famous for being indifferent to money, and if they are forced to pay not merely the cost of the harm to the victims of their torts but also some multiple of that cost they are likely to think twice before engaging in such expen*36sive behavior again. Juries, rightly or wrongly, think differently, so plaintiffs who are seeking punitive damages often present evidence of the defendant’s wealth. The question is whether they must present such evidence — whether it is somehow unjust to allow a jury to award punitive damages without knowing that the defendant really is a wealthy person. The answer, obviously, is no. A plaintiff is not required to seek punitive damages in the first place, so he should not be denied an award of punitive damages merely because he does not present evidence that if believed would persuade the jury to award him even more than he is asking.

Take the question from the other side: if the defendant is not as wealthy as the jury might in the absence of any evidence suppose, should the plaintiff be required .to show this? That seems an odd suggestion too. The reprehensibility of a person’s conduct is not mitigated by his not being a rich person, and plaintiffs are never required to apologize for seeking damages that if awarded will precipitate the defendant into bankruptcy. A plea of poverty is a classic appeal to the mercy of the judge or jury, and why the plaintiff should be required to make the plea on behalf of his opponent eludes us.

The usual practice with respect to fines is not to proportion the fine to the defendant’s wealth, but to allow him to argue that the fine should be waived or lowered because he cannot possibly pay it. U.S.S.G. § 5E1.2(a); United States v. Young, 66 F.3d 830, 839 (7th Cir.1995). (For a rare exception, based on a special statute, see Merritt v. United States, 960 F.2d 15, 18 (2d Cir.1992).) The practice with regard to sanctions under Fed.R.Civ.P. 11 is similar. E.g., Johnson v. A.W. Chesterton Co., 18 F.3d 1362, 1366 (7th Cir.1994). It is unnecessary to multiply examples. Given the close relation between fines and punitive damages, this is the proper approach to punitive damages as well. The defendant who cannot pay a large award of punitive damages can point this out to the jury so that they will not waste their time and that of the bankruptcy courts by awarding an amount that exceeds his ability to pay.

It ill becomes defendants to argue that plaintiffs must introduce evidence of the defendant’s wealth. Since most tort defendants against whom punitive damages are sought are enterprises rather than individuals, the effect of such a rule would be to encourage plaintiffs to seek punitive damages whether or not justified, in order to be able to put before the jury evidence that the defendant has a deep pocket and therefore should be made to pay a large judgment regardless of any nice calculation of actual culpability. (The judge might not allow this, if persuaded by the suggestion in Zazú Designs that the defendant’s net worth is irrelevant to the size of the award of punitive damages when the defendant is a corporation or other institution rather than an individual.) Individual defendants, as in the present case, are reluctant to disclose their net worth in any circumstances, so that compelling plaintiffs to seek discovery of that information would invite a particularly intrusive and resented form of pretrial discovery and disable the defendant from objecting. Sinee, moreover, information about net worth is in the possession of the person whose net wealth is in issue, the normal principles of pleading would put the burden of production on the defendant — which, as we have been at pains to stress, is just where defendants as a whole would want it.

Peters argues that a defendant who presents evidence of his net worth to the jury in an effort to minimize any award of punitive damages will be understood by the jury to be conceding the appropriateness of awarding punitive damages in at least the amount suggested by the defendant. This is just the kind of thinking that has so often led defendants into disaster when they decided not to put into evidence their own estimate of the damages to which the plaintiff was entitled, but instead played the equivalent of double or nothing. See, e.g., Avitia v. Metropolitan Club of Chicago, Inc., 49 F.3d 1219, 1230 (7th Cir.1995). Most jurors should be able to understand the structure of an argument to the effect that the defendant does not concede liability, let alone liability for punitive damages, but that if the jury disagrees it should award only a nominal amount of punitive damages because the defendant is a person of limited means.

*37The defendant should not be allowed to plead poverty if his employer or an insurance company is going to pick up the tab. DeLoach v. Bevers, 922 F.2d 618, 624 (10th Cir.1990); Garnes v. Fleming Landfill, Inc., 186 W.Va. 656, 669, 413 S.E.2d 897, 910 (1991); DeMatteo v. Simon, 112 N.M. 112, 115, 812 P.2d 361, 364 (Ct.App.1991). The contrary argument, accepted in Michael v. Cole, 122 Ariz. 450, 452, 595 P.2d 995, 997 (1979), that the insurance contract is a purely private matter between the defendant and his insured, ignores the consequence of such a view for the deterrent efficacy of punitive damages. It is bad enough that insurance or other indemnification reduces the financial incentive to avoid wrongdoing — which is why insuring against criminal liability is prohibited. It would be worse if the cost of the insurance fell, reducing the financial disincentive to engage in wrongful behavior, because the insurance company knew that its insured could plead poverty to the jury.

We were told by Peters’ lawyer without contradiction that Peters will not be indemnified for the punitive damages that he has been ordered to pay. We have noted the inappropriateness of allowing the defendant to plead poverty if he will be indemnified not because such a plea was attempted here, but to underscore the anomaly of requiring plaintiffs seeking punitive damages always to put in evidence of the defendant’s net worth. When the defendant is to be fully indemnified, such evidence, far from being required, is inadmissible. Thus, in some cases it is inadmissible, but in no cases is it required.

Affirmed.

*412Justice Kennedy

delivered the opinion of the Court.

We address once again the measure of punishment, by means of punitive damages, a State may impose upon a defendant in a civil case. The question is whether, in the circumstances we shall recount, an award of $145 million in punitive damages, where full compensatory damages are $.1 million, is excessive and in violation of the Due Process Clause of the Fourteenth Amendment to the Constitution of the United States.

I

In 1981, Curtis Campbell (Campbell) was driving with his wife, Inez Preece Campbell, in Cache County, Utah. He decided to pass six vans traveling ahead of them on a two-lane highway. Todd Ospital was driving a small car approaching from the opposite direction. To avoid a head-on collision with Campbell, who by then was driving on the wrong side of the highway and toward oncoming traffic, Ospital swerved onto the shoulder, lost control of his automobile, and col*413lided with a vehicle driven by Robert G. Slusher. Ospital was killed, and Slusher was rendered permanently disabled. The Campbells escaped unscathed.

In the ensuing wrongful death and tort action, Campbell insisted he was not at fault. Early investigations did support differing conclusions as to who caused the accident, but “a consensus was reached early on by the investigators and witnesses that Mr. Campbell’s unsafe pass had indeed caused the crash.” 65 P. Bd 1134, 1141 (Utah 2001). Campbell’s insurance company, petitioner State Farm Mutual Automobile Insurance Company (State Farm), nonetheless decided to contest liability and declined offers by Slusher and Ospital’s estate (Ospital) to settle the claims for the policy limit of $50,000 ($25,000 per claimant). State Farm also ignored the advice of one of its own investigators and took the case to trial, assuring the Campbells that “their assets were safe, that they had no liability for the accident, that [State Farm] would represent their interests, and that they did not need to procure separate counsel.” Id., at 1142. To the contrary, a jury determined that Campbell was 100 percent at fault, and a judgment was returned for $185,849, far more than the amount offered in settlement.

At first State Farm refused to cover the $135,849 in excess liability. Its counsel made this clear to the Campbells: “ ‘You may want to put for sale signs on your property to get things moving.’ ” Ibid. Nor was State Farm willing to post a su-persedeas bond to allow Campbell to appeal the judgment against him. Campbell obtained his own counsel to appeal the verdict. During the pendency of the appeal, in late 1984, Slusher, Ospital, and the Campbells reached an agreement whereby Slusher and Ospital agreed not to seek satisfaction of their claims against the Campbells. In exchange the Campbells agreed to pursue a bad-faith action against State Farm and to be represented by Slusher’s and Ospital’s attorneys. The Campbells also agreed that Slusher and Ospital would have a right to play a part in all major decisions con*414cerning the bad-faith action. No settlement could be concluded without Slusher’s and Ospital’s approval, and Slusher and Ospital would receive 90 percent of any verdict against State Farm.

In 1989, the Utah Supreme Court denied Campbell’s appeal in the wrongful-death and tort actions. Slusher v. Ospital, 777 P. 2d 437. State Farm then paid the entire judgment, including the amounts in excess of the policy limits. The Campbells nonetheless filed a complaint against State Farm alleging bad faith, fraud, and intentional infliction of emotional distress. The trial court initially granted State Farm’s motion for summary judgment because State Farm had paid the excess verdict, but that ruling was reversed on appeal. 840 P. 2d 130 (Utah App. 1992). On remand State Farm moved in limine to exclude evidence of alleged conduct that occurred in unrelated cases outside of Utah, but the trial court denied the motion. At State Farm’s request the trial court bifurcated the trial into two phases conducted before different juries. In the first phase the jury determined that State Farm’s decision not to settle was unreasonable because there was a substantial likelihood of an excess verdict.

Before the second phase of the action against State Farm we decided BMW of North America, Inc. v. Gore, 517 U. S. 559 (1996), and refused to sustain a $2 million punitive damages award which accompanied a verdict of only $4,000 in compensatory damages. Based on that decision, State Farm again moved for the exclusion of evidence of dissimilar out-of-state conduct. App. to Pet. for Cert. 168a-172a. The trial court denied State Farm’s motion. Id., at 189a.

The second phase addressed State Farm’s liability for fraud and intentional infliction of emotional distress, as well as compensatory and punitive damages. The Utah Supreme Court aptly characterized this phase of the trial:

“State Farm argued during phase II that its decision to take the case to trial was an ‘honest mistake’ that did *415not warrant punitive damages. In contrast, the Camp-bells introduced evidence that State Farm’s decision to take the case to trial was a result of a national scheme to meet corporate fiscal goals by capping payouts on claims company wide. This scheme was referred to as State Farm’s ‘Performance, Planning and Review,’ or PP & R, policy. To prove the existence of this scheme, the trial court allowed the Campbells to introduce extensive expert testimony regarding fraudulent practices by State Farm in its nation-wide operations. Although State Farm moved prior to phase II of the trial for the exclusion of such evidence and continued to object to it at trial, the trial court ruled that such evidence was admissible to determine whether State Farm’s conduct in the Campbell case was indeed intentional and sufficiently egregious to warrant punitive damages.” 65 P. 3d, at 1143.

Evidence pertaining to the PP&R policy concerned State Farm’s business practices for over 20 years in numerous States. Most of these practices bore no relation to third-party automobile insurance claims, the type of claim underlying the Campbells’ complaint against the company. The jury awarded the Campbells $2.6 million in compensatory damages and $145 million in punitive damages, which the trial court reduced to $1 million and $25 million respectively. Both parties appealed.

The Utah Supreme Court sought to apply the three guideposts we identified in Gore, supra, at 574-575, and it reinstated the $145 million punitive damages award. Relying in large part on the extensive evidence concerning the PP&R policy, the court concluded State Farm’s conduct was reprehensible. The court also relied upon State Farm’s “massive wealth” and on testimony indicating that “State Farm’s actions, because of their clandestine nature, will be punished at most in one out of every 50,000 cases as a matter of statistical probability,” 65 P. 3d, at 1153, and concluded that the ratio *416between punitive and compensatory damages was not unwarranted. Finally, the court noted that the punitive damages award was not excessive when compared to various civil and criminal penalties State Farm could have faced, including $10,000 for each act of fraud, the suspension of its license to conduct business in Utah, the disgorgement of profits, and imprisonment. Id., at 1154-1155. We granted certiorari. 535 U. S. 1111 (2002).

II

We recognized in Cooper Industries, Inc. v. Leatherman Tool Group, Inc., 532 U. S. 424 (2001), that in our judicial system compensatory and punitive damages, although usually awarded at the same time by the same decisionmaker, serve different purposes. Id., at 432. Compensatory damages “are intended to redress the concrete loss that the plaintiff has suffered by reason of the defendant’s wrongful conduct.” Ibid, (citing Restatement (Second) of Torts §903, pp. 453-454 (1979)). By contrast, punitive damages serve a broader function; they are aimed at deterrence and retribution. Cooper Industries, supra, at 432; see also Gore, supra, at 568 (“Punitive damages may properly be imposed to further a State’s legitimate interests in punishing unlawful conduct and deterring its repetition”); Pacific Mut. Life Ins. Co. v. Haslip, 499 U. S. 1, 19 (1991) (“[Pjunitive damages are imposed for purposes of retribution and deterrence”).

While States possess discretion over the imposition of punitive damages, it is well established that there are procedural and substantive constitutional limitations on these awards. Cooper Industries, supra; Gore, supra, at 559; Honda Motor Co. v. Oberg, 512 U. S. 415 (1994); TXO Production Corp. v. Alliance Resources Corp., 509 U. S. 443 (1993); Haslip, supra. The Due Process Clause of the Fourteenth Amendment prohibits the imposition of grossly excessive or arbitrary punishments on a tortfeasor. Cooper Industries, supra, at 433; Gore, 517 U. S., at 562; see also id., at 587 (Breyer, J., concurring) (“This constitutional concern, itself *417harkening back to the Magna Carta, arises out of the basic unfairness of depriving citizens of life, liberty, or property, through the application, not of law and legal processes, but of arbitrary coercion”). The reason is that “[elementary notions of fairness enshrined in our constitutional jurisprudence dictate that a person receive fair notice not only of the conduct that will subject him to punishment, but also of the severity of the penalty that a State may impose.” Id., at 574; Cooper Industries, supra, at 433 (“Despite the broad discretion that States possess with respect to the imposition of criminal penalties and punitive damages, the Due Process Clause of the Fourteenth Amendment to the Federal Constitution imposes substantive limits on that discretion”). To the extent an award is grossly excessive, it furthers no legitimate purpose and constitutes an arbitrary deprivation of property. Haslip, supra, at 42 (O’Connor, J., dissenting) (“Punitive damages are a powerful weapon. Imposed wisely and with restraint, they have the potential to advance legitimate state interests. Imposed indiscriminately, however, they have a devastating potential for harm. Regrettably, common-law procedures for awarding punitive damages fall into the latter category”).

Although these awards serve the same purposes as criminal penalties, defendants subjected to punitive damages in civil cases have not been accorded the protections applicable in a criminal proceeding. This increases our concerns over the imprecise manner in which punitive damages systems are administered. We have admonished that “[pjunitive damages pose an acute danger of arbitrary deprivation of property. Jury instructions typically leave the jury with wide discretion in choosing amounts, and the presentation of evidence of a defendant’s net worth creates the potential that juries will use their verdicts to express biases against big businesses, particularly those without strong local presences.” Honda Motor, supra, at 432; see also Haslip, supra, at 59 (O’Connor, J., dissenting) (“[T]he Due Process Clause *418does not permit a State to classify arbitrariness as a virtue. Indeed, the point of due process — of the law in general — is to allow citizens to order their behavior. A State can have no legitimate interest in deliberately making the law so arbitrary that citizens will be unable to avoid punishment based solely upon bias or whim”). Our concerns are heightened when the decisionmaker is presented, as we shall discuss, with evidence that has little bearing as to the amount of punitive damages that should be awarded. Vague instructions, or those that merely inform the jury to avoid “passion or prejudice,” App. to Pet. for Cert. 108a-109a, do little to aid the decisionmaker in its task of assigning appropriate weight to evidence that is relevant and evidence that is tangential or only inflammatory.

In light of these concerns, in Gore, supra, we instructed courts reviewing punitive damages to consider three guideposts: (1) the degree of reprehensibility of the defendant’s misconduct; (2) the disparity between the actual or potential harm suffered by the plaintiff and the punitive damages award; and (3) the difference between the punitive damages awarded by the jury and the civil penalties authorized or imposed in comparable cases. Id., at 575. We reiterated the importance of these three guideposts in Cooper Industries and mandated appellate courts to conduct de novo review of a trial court’s application of them to the jury’s award. 532 U. S. 424. Exacting appellate review ensures that an award of punitive damages is based upon an “‘application of law, rather than a decisionmaker’s caprice.’” Id., at 436 (quoting Gore, supra, at 587 (Breyer, J., concurring)).

III

Under the principles outlined m BMW of North America, Inc. v. Gore, this case is neither close nor difficult. It was error to reinstate the jury’s $145 million punitive damages award. We address each guidepost of Gore in some detail.

*419A

“[T]he most important indicium of the reasonableness of a punitive damages award is the degree of reprehensibility of the defendant’s conduct.” Gore, 517 U. S., at 575. We have instructed courts to determine the reprehensibility of a defendant by considering whether: the harm caused was physical as opposed to economic; the tortious conduct evinced an indifference to or a reckless disregard of the health or safety of others; the target of the conduct had financial vulnerability; the conduct involved repeated actions or was an isolated incident; and the harm was the result of intentional malice, trickery, or deceit, or mere accident. Id., at 576-577. The existence of any one of these factors weighing in favor of a plaintiff may not be sufficient to sustain a punitive damages award; and the absence of all of them renders any award suspect. It should be presumed a plaintiff has been made whole for his injuries by compensatory damages, so punitive damages should only be awarded if the defendant’s culpability, after having paid compensatory damages, is so reprehensible as to warrant the imposition of further sanctions to achieve punishment or deterrence. Id., at 575.

Applying these factors in the instant case, we must acknowledge that State Farm’s handling of the claims against the Campbells merits no praise. The trial court found that State Farm’s employees altered the company’s records to make Campbell appear less culpable. State Farm disregarded the overwhelming likelihood of liability and the near-certain probability that, by taking the case to trial, a judgment in excess of the policy limits would be awarded. State Farm amplified, the harm by at first assuring the Campbells their assets would be safe from any verdict and by later telling them, postjudgment, to put a for-sale sign on their house. While we do not suggest there was error in awarding punitive damages based upon State Farm’s conduct toward the Campbells, a more modest punishment for this *420reprehensible conduct could have satisfied the State’s legitimate objectives, and the Utah courts should have gone no further.

This case, instead, was used as a platform to expose, and punish, the perceived deficiencies of State Farm’s operations throughout the country. The Utah Supreme Court’s opinion makes explicit that State Farm was being condemned for its nationwide policies rather than for the conduct directed toward the Campbells. 65 P. 3d, at 1143 (“[T]he Campbells introduced evidence that State Farm’s decision to take the case to trial was a result of a national scheme to meet corporate fiscal goals by capping payouts on claims company wide”). This was, as well, an explicit rationale of the trial court’s decision in approving the award, though reduced from $145 million to $25 million. App. to Pet. for Cert. 120a (“[T]he Campbells demonstrated, through the testimony of State Farm employees who had worked outside of Utah, and through expert testimony, that this pattern of claims adjustment under the PP&R program was not a local anomaly, but was a consistent, nationwide feature of State Farm’s business operations, orchestrated from the highest levels of corporate management”).

The Campbells contend that State Farm has only itself to blame for the reliance upon dissimilar and out-of-state conduct evidence. The record does not support this contention. From their opening statements onward the Campbells framed this case as a chance to rebuke State Farm for its nationwide activities. App. 208 (“You’re going to hear evidence that even the insurance commission in Utah and around the country are unwilling or inept at protecting people against abuses”); id., at 242 (“[T]his is a very important case. . . . [I]t transcends the Campbell file. It involves a nationwide practice. And you, here, are going to be evaluating and assessing, and hopefully requiring State Farm to stand accountable for what it’s doing across the country, which is the purpose of punitive damages”). This was a po*421sition maintained throughout the litigation. In opposing State Farm’s motion to exclude such evidence under Gore, the Campbells’ counsel convinced the trial court that there was no limitation on the scope of evidence that could be considered under our precedents. App. to Pet. for Cert. 172a (“As I read the case [Gore], I was struck with the fact that a clear message in the case . . . seems to be that courts in punitive damages cases should receive more evidence, not less. And that the court seems to be inviting an even broader area of evidence than the current rulings of the court would indicate”); id., at 189a (trial court ruling).

A State cannot punish a defendant for conduct that may have been lawful where it occurred. Gore, supra, at 572; Bigelow v. Virginia, 421 U. S. 809, 824 (1975) (“A State does not acquire power or supervision over the internal affairs of another State merely because the welfare and health of its own citizens may be affected when they travel to that State”); New York Life Ins. Co. v. Head, 234 U. S. 149, 161 (1914) (“[I]t would be impossible to permit the statutes of Missouri to operate beyond the jurisdiction of that State ... without throwing down the constitutional barriers by which all the States are restricted within the orbits of their lawful authority and upon the preservation of which the Government under the Constitution depends. This is so obviously the necessary result of the Constitution that it has rarely been called in question and hence authorities directly dealing with it do not abound”); Huntington v. Attrill, 146 U. S. 657, 669 (1892) (“Laws have no force of themselves beyond the jurisdiction of the State which enacts them, and can have extra-territorial effect only by the comity of other States”). Nor, as a general rule, does a State have a legitimate concern in imposing punitive damages to punish a defendant for unlawful acts committed outside of the State’s jurisdiction. Any proper adjudication of conduct that occurred outside Utah to other persons would require their inclusion, and, to those parties, the Utah courts, in the usual case, would need *422to apply the laws of their relevant jurisdiction. Phillips Petroleum Co. v. Shutts, 472 U. S. 797, 821-822 (1985).

Here, the Campbells do not dispute that much of the out-of-state conduct was lawful where it occurred. They argue, however, that such evidence was not the primary basis for the punitive damages award and was relevant to the extent it demonstrated, in a general sense, State Farm’s motive against its insured. Brief for Respondents 46-47 (“[E]ven if the practices described by State Farm were not malum in se or malum prohibitum, they became relevant to punitive damages to the extent they were used as tools to implement State Farm’s wrongful PP&R policy”). This argument misses the mark. Lawful out-of-state conduct may be probative when it demonstrates the deliberateness and culpability of the defendant’s action in the State where it is tortious, but that conduct must have a nexus to the specific harm suffered by the plaintiff. A jury must be instructed, furthermore, that it may not use evidence of out-of-state conduct to punish a defendant for action that was lawful in the jurisdiction where it occurred. Gore, 517 U. S., at 572-573 (noting that a State “does not have the power... to punish [a defendant] for conduct that was lawful where it occurred and that had no impact on [the State] or its residents”). A basic principle of federalism is that each State may make its own reasoned judgment about what conduct is permitted or proscribed within its borders, and each State alone can determine what measure of punishment, if any, to impose on a defendant who acts within its jurisdiction. Id., at 569 (“[T]he States need not, and in fact do not, provide such protection in a uniform manner”).

For a more fundamental reason, however, the Utah courts erred in relying upon this and other evidence: The courts awarded punitive damages to punish and deter conduct that bore no relation to the Campbells’ harm. A defendant’s dissimilar acts, independent from the acts upon which liability was premised, may not serve as the basis for punitive dam*423ages. A defendant should be punished for the conduct that harmed the plaintiff, not for being an unsavory individual or business. Due process does not permit courts, in the calculation of punitive damages, to adjudicate the merits of other parties’ hypothetical claims against a defendant under the guise of the reprehensibility analysis, but we have no doubt the Utah Supreme Court did that here. 65 P. 3d, at 1149 (“Even if the harm to the Campbells can be appropriately characterized as minimal, the trial court’s assessment of the situation is on target: ‘The harm is minor to the individual but massive in the aggregate’”). Punishment on these bases creates the possibility of multiple punitive damages awards for the same conduct; for in the usual case nonparties are not bound by the judgment some other plaintiff obtains. Gore, supra, at 593 (Breyer, J., concurring) (“Larger damages might also ‘double count’ by including in the punitive damages award some of the compensatory, or punitive, damages that subsequent plaintiffs would also recover”).

The same reasons lead us to conclude the Utah Supreme Court’s decision cannot be justified on the grounds that State Farm was a recidivist. Although “[o]ur holdings that a recidivist may be punished more severely than a first offender recognize that repeated misconduct is more reprehensible than an individual instance of malfeasance,” Gore, supra, at 577, in the context of civil actions courts must ensure the conduct in question replicates the prior transgressions. TXO, 509 U. S., at 462, n. 28 (noting that courts should look tó “ ‘the existence and frequency of similar past conduct’ ” (quoting Haslip, 499 U. S., at 21-22)).

The Campbells have identified scant evidence of repeated misconduct of the sort that injured them. Nor does our review of the Utah courts’ decisions convince us that State Farm was only punished for its actions toward the Camp-bells. Although evidence of other acts need not be identical to have relevance in the calculation of punitive damages, the Utah court erred here because evidence pertaining to claims *424that had nothing to do with a third-party lawsuit was introduced at length. Other evidence concerning reprehensibility was even more tangential. For example, the Utah Supreme Court criticized State Farm’s investigation into the personal life of one of its employees and, in a broader approach, the manner in which State Farm’s policies corrupted its employees. 65 P. 3d, at 1148, 1150. The Campbells attempt to justify the courts’ reliance upon this unrelated testimony on the theory that each dollar of profit made by underpaying a third-party claimant is the same as a dollar made by underpaying a first-party one. Brief for Respondents 45; see also 65 P. 3d, at 1150 (“State Farm’s continuing illicit practice created market disadvantages for other honest insurance companies because these practices increased profits. As plaintiffs’ expert witnesses established, such wrongfully obtained competitive advantages have the potential to pressure other companies to adopt similar fraudulent tactics, or to force them out of business. Thus, such actions cause distortions throughout the insurance market and ultimately hurt all consumers”). For the reasons already stated, this argument is unconvincing. The reprehensibility guidepost does not permit courts to expand the scope of the case so that a defendant may be punished for any malfeasance, which in this case extended for a 20-year period. In this case, because the Campbells have shown no conduct by State Farm similar to that which harmed them, the conduct that harmed them is the only conduct relevant to the reprehensibility analysis.

B

Turning to the second Gore guidepost, we have been reluctant to identify concrete constitutional limits on the ratio between harm, or potential harm, to the plaintiff and the punitive damages award. 517 U. S., at 582 (“[W]e have consistently rejected the notion that the constitutional line is marked by a simple mathematical formula, even one that compares actual and potential damages to the punitive *425award”); TXO, supra, at 458. We decline again to impose a bright-line ratio which a punitive damages award cannot exceed. Our jurisprudence and the principles it has now established demonstrate, however, that, in practice, few awards exceeding a single-digit ratio between punitive and compensatory damages, to a significant degree, will satisfy due process. In Haslip, in upholding a punitive damages award, we concluded that an award of more than four times the amount of compensatory damages might be close to the line of constitutional impropriety. 499 U. S., at 23-24. We cited that 4-to-l ratio again in Gore. 517 U. S., at 581. The Court further referenced a long legislative history, dating back over 700 years and going forward to today, providing for sanctions of double, treble, or quadruple damages to deter and punish. Id., at 581, and n. 33. While these ratios are not binding, they are instructive. They demonstrate what should be obvious: Single-digit multipliers are more likely to comport with due process, while still achieving the State’s goals of deterrence and retribution, than awards with ratios in range of 500 to 1, id., at 582, or, in this case, of 145 to 1.

Nonetheless, because there are no rigid benchmarks that a punitive damages award may not surpass, ratios greater than those we have previously upheld may comport with due process where “a particularly egregious act has resulted in only a small amount of economic damages.” Ibid.; see also ibid, (positing that a higher ratio might be necessary where “the injury is hard to detect or the monetary value of non-economic harm might have been difficult to determine”). The converse is also true, however. When compensatory damages are substantial, then a lesser ratio, perhaps only equal to compensatory damages, can reach the outermost limit of the due process guarantee. The precise award in any case, of course, must be based upon the facts and circumstances of the defendant’s conduct and the harm to the plaintiff.

*426In sum, courts must 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. In the context of this case, we have no doubt that there is a presumption against an award that has a 145-to-l ratio. The compensatory award in this case was substantial; the Campbells were awarded $1 million for a year and a half of emotional distress. This was complete compensation. The harm arose from a transaction in the economic realm, not from some physical assault or trauma; there were no physical injuries; and State Farm paid the excess verdict before the complaint was filed, so the Campbells suffered only minor economic injuries for the 18-month period in which State Farm refused to resolve the claim against them. The compensatory damages for the injury suffered here, moreover, likely were based on a component which was duplicated in the punitive award. Much of the distress was caused by the outrage and humiliation the Campbells suffered at the actions of their insurer; and it is a major role of punitive damages to condemn such conduct. Compensatory damages, however, already contain this punitive element. See Restatement (Second) of Torts § 908, Comment c, p. 466 (1977) (“In many cases in which compensatory damages include an amount for emotional distress, such as humiliation or indignation aroused by the defendant’s act, there is no clear line of demarcation between punishment and compensation and a verdict for a specified amount frequently includes elements of both”).

The Utah Supreme Court sought to justify the massive award by pointing to State Farm’s purported failure to report a prior $100 million punitive damages award in Texas to its corporate headquarters; the fact that State Farm’s policies have affected numerous Utah consumers; the fact that State Farm will only be punished in one out of every 50,000 cases as a matter of statistical probability; and State Farm’s enormous wealth. 65 P. 3d, at 1153. Since the Supreme *427Court of Utah discussed the Texas award when applying the ratio guidepost, we discuss it here. The Texas award, however, should have been analyzed in the context of the reprehensibility guidepost only. The failure of the company to report the Texas award is out-of-state conduct that, if the conduct were similar, might have had some bearing on the degree of reprehensibility, subject to the limitations we have described. Here, it was dissimilar, and of such marginal relevance that it should have been accorded little or no weight. The award was rendered in a first-party lawsuit; no judgment was entered in the case; and it was later settled for a fraction of the verdict. With respect to the Utah Supreme Court’s second justification, the Campbells’ inability to direct us to testimony demonstrating harm to the people of Utah (other than those directly involved in this case) indicates that the adverse effect on the State’s general population was in fact minor.

The remaining premises for the Utah Supreme Court’s decision bear no relation to the award’s reasonableness or proportionality to the harm. They are, rather, arguments that seek to defend a departure from well-established constraints on punitive damages. While States enjoy considerable discretion in deducing when punitive damages are warranted, each award must comport with the principles set forth in Gore. Here the argument that State Farm will be punished in only the rare case, coupled with reference to its assets (which, of course, are what other insured parties in Utah and other States must rely upon for payment of claims) had little to do with the actual harm sustained by the Campbells. The wealth of a defendant cannot justify an otherwise unconstitutional punitive damages award. Gore, 517 U. S., at 585 (“The fact that BMW is a large corporation rather than an impecunious individual does not diminish its entitlement to fair notice of the demands that the several States impose on the conduct of its business”); see also id., at 591 (BREYER, J., concurring) (“[Wealth] provides an open-ended basis for *428inflating awards when the defendant is wealthy .... That does not make its use unlawful or inappropriate; it simply means that this factor cannot make up for the failure of other factors, such as ‘reprehensibility,’ to constrain significantly an award that purports to punish a defendant’s conduct”). The principles set forth in Gore must be implemented with care, to ensure both reasonableness and proportionality.

C

The third guidepost in Gore is the disparity between the punitive damages award and the “civil penalties authorized or imposed in comparable cases.” Id., at 575. We note that, in the past, we have also looked to criminal penalties that could be imposed. Id., at 583; Haslip, 499 U. S., at 23. The existence of a criminal penalty does have bearing on the seriousness with which a State views the wrongful action. When used to determine the dollar amount of the award, however, the criminal penalty has less utility. Great care must be taken to avoid use of the civil process to assess criminal penalties that can be imposed only after the heightened protections of a criminal trial have been observed, including, of course, its higher standards of proof. Punitive damages are not a substitute for the criminal process, and the remote possibility of a criminal sanction does not automatically sustain a punitive damages award.

Here, we need not dwell long on this guidepost. The most relevant civil sanction under Utah state law for the wrong done to the Campbells appears to be a $10,000 fine for an act of fraud, 65 P. 3d, at 1154, an amount dwarfed by the $145 million punitive damages award. The Supreme Court of Utah speculated about the loss of State Farm’s business license, the disgorgement of profits, and possible imprisonment, but here again its references were to the broad fraudulent scheme drawn from evidence of out-of-state and dissimilar conduct. This analysis was insufficient to justify the award.

*429>

An application of the Gore guideposts to the facts of this case, especially in light of the substantial compensatory damages awarded (a portion of which contained a punitive element), likely would justify a punitive damages award at or near the amount of compensatory damages. The punitive award of $145 million, therefore, was neither reasonable nor proportionate to the wrong committed, and it was an irrational and arbitrary deprivation of the property of the defendant. The proper calculation of punitive damages under the principles we have discussed should be resolved, in the first instance, by the Utah courts.

The judgment of the Utah Supreme Court is reversed, and the ease is remanded for further proceedings not inconsistent with this opinion.

It is so ordered.

Justice Scalia,

dissenting.

I adhere to the view expressed in my dissenting opinion in BMW of North America, Inc. v. Gore, 517 U. S. 559, 598-599 (1996), that the Due Process Clause provides no substantive protections against “excessive” or “'unreasonable’” awards of punitive damages. I am also of the view that the punitive damages jurisprudence which has sprung forth from BMW v. Gore is insusceptible of principled application; accordingly, I do not feel justified in giving the case stare deci-sis effect. See id., at 599. I would affirm the judgment of the Utah Supreme Court.

Justice Thomas,

dissenting.

I would affirm the judgment below because “I continue to believe that the Constitution does not constrain the size of punitive damages awards.” Cooper Industries, Inc. v. Leatherman Tool Group, Inc., 532 U. S. 424, 443 (2001) (Thomas, J., concurring) (citing BMW of North America, *430Inc. v. Gore, 517 U. S. 559, 599 (1996) (Scalia, J., joined by Thomas, J., dissenting)). Accordingly, I respectfully dissent.

Justice Ginsburg,

dissenting.

Not long ago, this Court was hesitant to impose a federal check on state-court judgments awarding punitive damages. In Browning-Ferris Industries of Vt., Inc. v. Kelco Disposal, Inc., 492 U. S. 257 (1989), the Court held that neither the Excessive Fines Clause of the Eighth Amendment nor federal common law circumscribed awards of punitive damages in civil cases between private parties. Id., at 262-276, 277-280. Two years later, in Pacific Mut. Life Ins. Co. v. Haslip, 499 U. S. 1 (1991), the Court observed that “unlimited jury [or judicial] discretion ... in the fixing of punitive damages may invite extreme results that jar one’s constitutional sensibilities,” id., at 18; the Due Process Clause, the Court suggested, would attend to those sensibilities and guard against unreasonable awards, id., at 17-24. Nevertheless, the Court upheld a punitive damages award in Haslip “more than 4 times the amount of compensatory damages, . . . more than 200 times [the plaintiff’s] out-of-pocket expenses,” and “much in excess of the fine that could be imposed.” Id., at 23. And in TXO Production Corp. v. Alliance Resources Corp., 509 U. S. 443 (1993), the Court affirmed a state-court award “526 times greater than the actual damages awarded by the jury.” Id., at 453;1 cf. Browning-Ferris, 492 U. S., at 262 (ratio of punitive to compensatory damages over 100 to 1).

It was not until 1996, in BMW of North America, Inc. v. Gore, 517 U. S. 559, that the Court, for the first time, invalidated a state-court punitive damages assessment as un*431reasonably large. See id., at 599 (Scalia, J., dissenting). If our activity in this domain is now “well established,” see ante, at 416, 427, it takes place on ground not long held.

In Gore, I stated why I resisted the Court’s foray into punitive damages “territory traditionally within the States’ domain.” 517 U. S., at 612 (dissenting opinion). I adhere to those views, and note again that, unlike federal habeas corpus review of state-court convictions under 28 U. S. C. §2254, the Court “work[s] at this business [of checking state courts] alone,” unaided by the participation of federal district courts and courts of appeals. 517 U. S., at 613. It was once recognized that “the laws of the particular State must suffice [to superintend punitive damages awards] until judges or legislators authorized to do so initiate system-wide change.” Haslip, 499 U. S., at 42 (Kennedy, J., concurring in judgment). I would adhere to that traditional view.

► — I

The large size of the award upheld by the Utah Supreme Court in this case indicates why damages-capping legislation may be altogether fitting and proper. Neither the amount of the award nor the trial record, however, justifies this Court’s substitution of its judgment for that of Utah’s competent de-cisionmakers. In this regard, I count it significant that, on the key criterion “reprehensibility,” there is a good deal more to the story than the Court’s abbreviated account tells.

Ample evidence allowed the jury to find that State Farm’s treatment of the Campbells typified its “Performance, Planning and Review” (PP&R) program; implemented by top management in 1979, the program had “the explicit objective of using the claims-adjustment process as a profit center.” App. to Pet. for Cert. 116a. “[Tjhe Campbells presented considerable evidence,” the trial court noted, documenting “that the PP&R program . . . has functioned, and continues to function, as an unlawful scheme ... to deny benefits owed consumers by paying out less than fair value in order to meet *432preset, arbitrary payout targets designed to enhance corporate profits.” Id., at 118a-119a. That policy, the trial court observed, was encompassing in scope; it “applied equally to the handling of both third-party and first-party claims.” Id., at 119a. But cf. ante, at 423-424, 427 (suggesting that State Farm’s handling of first-party claims has “nothing to do with a third-party lawsuit”).

Evidence the jury could credit demonstrated that the PP&R program regularly and adversely affected Utah residents. Ray Summers, “the adjuster who handled the Campbell case and who was a State Farm employee in Utah for almost twenty years,” described several methods used by State Farm to deny claimants fair benefits, for example, “falsifying or withholding of evidence in claim files.” App. to Pet. for Cert. 121a. A common tactic, Summers recounted, was to “unjustly attac[k] the character, reputation and credibility of a claimant and mak[e] notations to that effect in the claim file to create prejudice in the event the claim ever came before a jury.” Id., at 130a (internal quotation marks omitted). State Farm manager Bob Noxon, Summers testified, resorted to a tactic of this order in the Campbell case when he “instructed] Summers to write in the file that Todd Os-pital (who was killed in the accident) was speeding because he was on his way to see a pregnant girlfriend.” Ibid. In truth, “[t]here was no pregnant girlfriend.” Ibid. Expert testimony noted by the trial court described these tactics as “completely improper.” Ibid.

The trial court also noted the testimony of two Utah State Farm employees, Felix Jensen and Samantha Bird, both of whom recalled “intolerable” and “recurrent” pressure to reduce payouts below fair value. Id., at 119a (internal quotation marks omitted). When Jensen complained to top managers, he was told to “get out of the kitchen” if he could not take the heat; Bird was told she should be “more of a team player.” Ibid, (internal quotation marks omitted). At times, Bird said, she “was forced to commit dishonest acts *433and to knowingly underpay claims.” Id., at 120a. Eventually, Bird quit. Ibid. Utah managers superior to Bird, the evidence indicated, were improperly influenced by the PP&R program to encourage insurance underpayments. For example, several documents evaluating the performance of managers Noxon and Brown “contained explicit preset average payout goals.” Ibid.

Regarding liability for verdicts in excess of policy limits, the trial court referred to a State Farm document titled the “Excess Liability Handbook”; written before the Campbell accident, the handbook instructed adjusters to pad files with “self-serving” documents, and to leave critical items out of files, for example, evaluations of the insured’s exposure. Id., at 127a-128a (internal quotation marks omitted). Divisional superintendent Bill Brown used the handbook to train Utah employees. Id., at 134a. While overseeing the Campbell case, Brown ordered adjuster Summers to change the portions of his report indicating that Mr. Campbell was likely at fault and that the settlement cost was correspondingly high. Id., at 3a. The Campbells’ case, according to expert testimony the trial court recited, “was a classic example of State Farm’s application of the improper practices taught in the Excess Liability Handbook.” Id., at 128a.

The trial court further determined that the jury could find State Farm’s policy “deliberately crafted” to prey on consumers who would be unlikely to defend themselves. Id., at 122a. In this regard, the trial court noted the testimony of several former State Farm employees affirming that they were trained to target “the weakest of the herd” — “the elderly, the poor, and other consumers who are least knowledgeable about their rights and thus most vulnerable to trickery or deceit, or who have little money and hence have no real alternative but to accept an inadequate offer to settle a claim at much less than fair value.” Ibid, (internal quotation marks omitted).

*434The Campbells themselves could be placed within the “weakest of the herd” category. The couple appeared economically vulnerable and emotionally fragile. App. 3360a-3361a (Order Denying State Farm’s Motion for Judgment NOV and New Trial Regarding Intentional Infliction of Emotional Distress). At the time of State Farm’s wrongful conduct, “Mr. Campbell had residuary effects from a stroke and Parkinson’s disease.” Id., at 3360a.

To further insulate itself from liability, trial evidence indicated, State Farm made “systematic” efforts to destroy internal company documents that might reveal its scheme, App. to Pet. for Cert. 123a, efforts that directly affected the Campbells, id., at 124a. For example, State Farm had “a special historical department that contained a copy of all past manuals on claim-handling practices and the dates on which each section of each manual was changed.” Ibid. Yet in discovery proceedings, State Farm failed to produce any claim-handling practice manuals for the years relevant to the Campbells’ bad-faith case. Id., at 124a-125a.

State Farm’s inability to produce the manuals, it appeared from the evidence, was not accidental. Documents retained by former State Farm employee Samantha Bird, as well as Bird’s testimony, showed that while the Campbells’ case was pending, Janet Cammack, “an in-house attorney sent by top State Farm management, conducted a meeting ... in Utah during which she instructed Utah claims management to search their offices and destroy a wide range of material of the sort that had proved damaging in bad-faith litigation in the past — in particular, old claim-handling manuals, memos, claim school notes, procedure guides and other similar documents.” Id., at 125a. “These orders were followed even though at least one meeting participant, Paul Short, was personally aware that these kinds of materials had been requested by the Campbells in this very case.” Ibid.

Consistent with Bird’s testimony, State Farm admitted that it destroyed every single copy of claim-handling manu*435als on file in its historical department as of 1988, even though these documents could have been preserved at minimal expense. Ibid. Fortuitously, the Campbells obtained a copy of the 1979 PP&R manual by subpoena from a former employee. Id., at 132a. Although that manual has been requested in other cases, State Farm has never itself produced the document. Ibid.

‘As a final, related tactic,” the trial court stated, the jury could reasonably find that “in recent years State Farm has gone to extraordinary lengths to stop damaging documents from being created in the first place.” Id., at 126a. State Farm kept no records at all on excess verdicts in third-party cases, or on bad-faith claims or attendant verdicts. Ibid. State Farm alleged “that it has no record of its punitive damage payments, even though such payments must be reported to the [Internal Revenue Service] and in some states may not be used to justify rate increases.” Ibid. Regional Vice President Buck Moskalski testified that “he would not report a punitive damage verdict in [the Campbells’] case to higher management, as such reporting was not set out as part of State Farm’s management practices.” Ibid.

State Farm’s “wrongful profit and evasion schemes,” the trial court underscored, were directly relevant to the Camp-bells’ case, id., at 132a:

“The record fully supports the conclusion that the bad-faith claim handling that exposed the Campbells to an excess verdict in 1983, and resulted in severe damages to them, was a product of the unlawful profit scheme that had been put in place by top management at State Farm years earlier. The Campbells presented substantial evidence showing how State Farm’s improper insistence on claims-handling employees’ reducing their claim payouts ... regardless of the merits of each claim, manifested itself ... in the Utah claims operations during the period when the decisions were made not to offer to settle the Campbell case for the $50,000 policy limits— *436indeed, not to make any offer to settle at a lower amount. This evidence established that high-level manager Bill Brown was under heavy pressure from the PP&R scheme to control indemnity payouts during the time period in question. In particular, when Brown declined to pay the excess verdict against Curtis Campbell, or even post a bond, he had a special need to keep his year-end numbers down, since the State Farm incentive scheme meant that keeping those numbers down was important to helping Brown get a much-desired transfer to Colorado.. .. There was ample evidence that the concepts taught in the Excess Liability Handbook, including the dishonest alteration and manipulation of claim files and the policy against posting any superse-deas bond for the full amount of an excess verdict, were dutifully carried out in this case. . . . There was ample basis for the jury to find that everything that had happened to the Campbells — when State Farm repeatedly refused in bad-faith to settle for the $50,000 policy limits and went to trial, and then failed to pay the ‘excess’ verdict, or at least post a bond, after trial — was a direct application of State Farm’s overall profit scheme, operating through Brown and others.” Id., at 133a-134a.

State Farm’s “policies and practices,” the trial evidence thus bore out, were “responsible for the injuries suffered by the Campbells,” and the means used to implement those policies could be found “callous, clandestine, fraudulent, and dishonest.” Id., at 136a; see id., at 113a (finding “ample evidence” that State Farm’s reprehensible corporate policies were responsible for injuring “many other Utah consumers during the past two decades”). The Utah Supreme Court, relying on the trial, court’s record-based recitations, understandably characterized State Farm’s behavior as “egregious and malicious.” Id., at 18a.

*437I > — I

The Court dismisses the evidence describing and documenting State Farm’s PP&R policy and practices as essentially irrelevant, bearing “no relation to the Campbells’ harm.” Ante, at 422; see ante, at 424 (“conduct that harmed [the Campbells] is the only conduct relevant to the reprehensibility analysis”). It is hardly apparent why that should be so. What is infirm about the Campbells’ theory that their experience with State Farm exemplifies and reflects an overarching underpayment scheme, one that caused “repeated misconduct of the sort that injured them,” ante, at 423? The Court’s silence on that score is revealing: Once one recognizes that the Campbells did show “conduct by State Farm similar to that which harmed them,” ante, at 424, it becomes impossible to shrink the reprehensibility analysis to this sole case, or to maintain, at odds with the determination of the trial court, see App. to Pet. for Cert. 113a, that “the adverse effect on the State’s general population was in fact minor,” ante, at 427.

Evidence of out-of-state conduct, the Court acknowledges, may be “probative [even if the conduct is lawful in the State where it occurred] when it demonstrates the deliberateness and culpability of the defendant’s action in the State where it is tortious....” Ante, at 422; cf. ante, at 419 (reiterating this Court’s instruction that trial courts assess whether “the harm was the result of intentional malice, trickery, or deceit, or mere accident”). “Other acts” evidence concerning practices both in and out of State was introduced in this case to show just such “deliberateness” and “culpability.” The evidence was admissible, the trial court ruled: (1) to document State Farm’s “reprehensible” PP&R program; and (2) to “rebut [State Farm’s] assertion that [its] actions toward the Campbells were inadvertent errors or mistakes in judgment.” App. 3329a (Order Denying Various Motions of State Farm to Exclude Plaintiffs’ Evidence). Viewed in this light, there surely was “a nexus” between much of the “other *438acts” evidence and “the specific harm suffered by [the Camp-bells].” Ante, at 422.

Ill

When the Court first ventured to override state-court punitive damages awards, it did so moderately. The Court recalled that “[i]n our federal system, States necessarily have considerable flexibility in determining the level of punitive damages that they will allow in different classes of cases and in any particular case.” Gore, 517 U. S., at 568. Today’s decision exhibits no such respect and restraint. No longer content to accord state-court judgments “a strong presumption of validity,” TXO, 509 U. S., at 457, the Court announces that “few awards exceeding a single-digit ratio between punitive and compensatory damages, to a significant degree, will satisfy due process.” Ante, at 425.2 Moreover, the Court adds, when compensatory damages are substantial, doubling those damages “can reach the outermost limit of the due process guarantee.” Ibid,.; see ante, at 429 (“facts of this case ... likely would justify a punitive damages award at or near the amount of compensatory damages”). In a legislative scheme or a state high court’s design to cap punitive damages, the handiwork in setting single-digit and 1-to-l benchmarks could hardly be questioned; in a judicial decree imposed on the States by this Court under the banner of substantive due process, the numerical controls today’s decision installs seem to me boldly out of order.

* * *

1 remain of the view that this Court has no warrant to reform state law governing awards of punitive damages. *439Gore, 517 U. S., at 607 (Ginsburg, J., dissenting). Even if I were prepared to accept the flexible guides prescribed in Gore, I would not join the Court’s swift conversion of those guides into instructions that begin to resemble marching orders. For the reasons stated, I would leave the judgment of the Utah Supreme Court undisturbed.

*349Justice Breyer

delivered the opinion of the Court.

The question we address today concerns a large state-court punitive damages award. We are asked whether the Constitution’s Due Process Clause permits a jury to base that award in part upon its desire to punish the defendant for harming persons who are not before the court (e. g., victims whom the parties do not represent). We hold that such an award would amount to a taking of “property” from the defendant without due process.

I

This lawsuit arises out of the death of Jesse Williams, a heavy cigarette smoker. Respondent, Williams’ widow, represents his estate in this state lawsuit for negligence and deceit against Philip Morris, the manufacturer of Marlboro, the brand that Williams favored. A jury found that Williams’ death was caused by smoking; that Williams smoked in significant part because he thought it was safe to do so; *350and that Philip Morris knowingly and falsely led him to believe that this was so. The jury ultimately found that Philip Morris was negligent (as was Williams) and that Philip Morris had engaged in deceit. In respect to deceit, the claim at issue here, it awarded compensatory damages of about $821,000 (about $21,000 economic and $800,000 noneconomic) along with $79.5 million in punitive damages.

The trial judge subsequently found the $79.5 million punitive damages award “excessive,” see, e. g., BMW of North America, Inc. v. Gore, 517 U. S. 559 (1996), and reduced it to $32 million. Both sides appealed. The Oregon Court of Appeals rejected Philip Morris’ arguments and restored the $79.5 million jury award. Subsequently, Philip Morris sought review in the Oregon Supreme Court (which denied review) and then here. We remanded the case in light of State Farm Mut. Automobile Ins. Co. v. Campbell, 538 U. S. 408 (2003). 540 U. S. 801 (2003). The Oregon Court of Appeals adhered to its original views. And Philip Morris sought, and this time obtained, review in the Oregon Supreme Court.

Philip Morris then made two arguments relevant here. First, it said that the trial court should have accepted, but did not accept, a proposed “punitive damages” instruction that specified the jury could not seek to punish Philip Morris for injury to other persons not before the court. In particular, Philip Morris pointed out that the plaintiff’s attorney had told the jury to “think about how many other Jesse Williams in the last 40 years in the State of Oregon there have been. ... In Oregon, how many people do we see outside, driving home . . . smoking cigarettes? . . . [Cigarettes . . . are going to kill ten [of every hundred]. [And] the market share of Marlboros [i e., Philip Morris] is one-third [i e., one of every three killed].” App. 197a, 199a. In light of this argument, Philip Morris asked the trial court to tell the jury that “you may consider the extent of harm suffered by others in determining what [the] reasonable relationship is” be*351tween any punitive award and “the harm caused to Jesse Williams” by Philip Morris’ misconduct, “[but] you are not to punish the defendant for the impact of its alleged misconduct on other persons, who may bring lawsuits of their own in which other juries can resolve their claims . . . Id., at 280a. The judge rejected this proposal and instead told the jury that “[p]unitive damages are awarded against a defendant to punish misconduct and to deter misconduct,” and “are not intended to compensate the plaintiff or anyone else for damages caused by the defendant’s conduct.” Id., at 283a. In Philip Morris’ view, the result was a significant likelihood that a portion of the $79.5 million award represented punishment for its having harmed others, a punishment that the Due Process Clause would here forbid.

Second, Philip Morris pointed to the roughly 100-to-l ratio the $79.5 million punitive damages award bears to $821,000 in compensatory damages. Philip Morris noted that this Court in BMW emphasized the constitutional need for punitive damages awards to reflect (1) the “reprehensibility” of the defendant’s conduct, (2) a “reasonable relationship” to the harm the plaintiff (or related victim) suffered, and (3) the presence (or absence) of “sanctions,” e. g., criminal penalties, that state law provided for comparable conduct, 517 U. S., at 575-585. And in State Farm, this Court said that the longstanding historical practice of setting punitive damages at two, three, or four times the size of compensatory damages, while “not binding,” is “instructive,” and that “[s]ingle-digit multipliers are more likely to comport with due process.” 538 U. S., at 425. Philip Morris claimed that, in light of this case law, the punitive award was “grossly excessive.” See TXO Production Corp. v. Alliance Resources Corp., 509 U. S. 443, 458 (1993) (plurality opinion); BMW, supra, at 574-575; State Farm, supra, at 416-417.

The Oregon Supreme Court rejected these and other Philip Morris arguments. In particular, it rejected Philip Morris’ claim that the Constitution prohibits a state jury *352“from using punitive damages to punish a defendant for harm to nonparties.” 340 Ore. 35, 51-52, 127 P. 3d 1165, 1175 (2006). And in light of Philip Morris’ reprehensible conduct, it found that the $79.5 million award was not “grossly excessive.” Id., at 63-64, 127 P. 3d, at 1181-1182.

Philip Morris then sought certiorari. It asked us to consider, among other things, (1) its claim that Oregon had unconstitutionally permitted it to be punished for harming nonparty victims; and (2) whether Oregon had in effect disregarded “the constitutional requirement that punitive damages be reasonably related to the plaintiff’s harm.” Pet. for Cert. (I). We granted certiorari limited to these two questions.

For reasons we shall set forth, we consider only the first of these questions. We vacate the Oregon Supreme Court’s judgment, and we remand the case for further proceedings.

II

This Court has long made clear that “[pjunitive damages may properly be imposed to further a State’s legitimate interests in punishing unlawful conduct and deterring its repetition.” BMW, supra, at 568. See also Gertz v. Robert Welch, Inc., 418 U. S. 323, 350 (1974); Newport v. Fact Concerts, Inc., 453 U. S. 247, 266-267 (1981); Pacific Mut. Life Ins. Co. v. Haslip, 499 U. S. 1, 22 (1991). At the same time, we have emphasized the need to avoid an arbitrary determination of an award’s amount. Unless a State insists upon proper standards that will cabin the jury’s discretionary authority, its punitive damages system may deprive a defendant of “fair notice ... of the severity of the penalty that a State may impose,” BMW, supra, at 574; it may threaten “arbitrary punishments,” i. e., punishments that reflect not an “application of law” but “a decisionmaker’s caprice,” State Farm, supra, at 416, 418 (internal quotation marks omitted); and, where the amounts are sufficiently large, it may impose one State’s (or one jury’s) “policy choice,” say, as to the condi*353tions under which (or even whether) certain products can be sold, upon “neighboring States” with different public policies, BMW, supra, at 571-572.

For these and similar reasons, this Court has found that the Constitution imposes certain limits, in respect both to procedures for awarding punitive damages and to amounts forbidden as “grossly excessive.” See Honda Motor Co. v. Oberg, 512 U. S. 415, 432 (1994) (requiring judicial review of the size of punitive awards); Cooper Industries, Inc. v. Leatherman Tool Group, Inc., 532 U. S. 424, 443 (2001) (review must be de novo); BMW, supra, at 574-585 (excessiveness decision depends upon the reprehensibility of the defendant’s conduct, whether the award bears a reasonable relationship to the actual and potential harm caused by the defendant to the plaintiff, and the difference between the award and sanctions “authorized or imposed in comparable cases”); State Farm, supra, at 425 (excessiveness more likely where ratio exceeds single digits). Because we shall not decide whether the award here at issue is “grossly excessive,” we need now only consider the Constitution’s procedural limitations.

III

In our view, the Constitution’s 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. For one thing, the Due Process Clause prohibits a State from punishing an individual without first providing that individual with “an opportunity to present every available defense.” Lindsey v. Normet, 405 U. S. 56, 66 (1972) (internal quotation marks omitted). Yet a defendant threatened with punishment for injuring a nonparty victim has no opportunity to defend against the charge, by showing, for example in a case such as this, that the other victim was not entitled to dam*354ages because he or she knew that smoking was dangerous or did not rely upon the defendant’s statements to the contrary.

For another, to permit punishment for injuring a nonparty victim would add a near standardless dimension to the punitive damages equation. How many such victims are there? How seriously were they injured? Under what circumstances did injury occur? The trial will not likely answer such questions as to nonparty victims. The jury will be left to speculate. And the fimdamental due process concerns to which our punitive damages cases refer — risks of arbitrariness, uncertainty, and lack of notice — will be magnified. State Farm, 538 U. S., at 416, 418; BMW, 517 U. S., at 574.

Finally, we can find no authority supporting the use of punitive damages awards for the purpose of punishing a defendant for harming others. We have said that it may be appropriate to consider the reasonableness of a punitive damages award in light of the potential harm the defendant’s conduct could have caused. But we have made clear that the potential harm at issue was harm potentially caused the plaintiff. See State Farm, supra, at 424 (“[W]e have been reluctant to identify concrete constitutional limits on the ratio between harm, or potential harm, to the plaintiff and the punitive damages award” (emphasis added)). See also TXO, 509 U. S., at 460-462 (plurality opinion) (using same kind of comparison as basis for finding a punitive award not unconstitutionally excessive). We did use the term “error-free” (in BMW) to describe a lower court punitive damages calculation that likely included harm to others in the equation. 517 U. S., at 568, n. 11. But context makes clear that the term “error-free” in the BMW footnote referred to errors relevant to the case at hand. Although elsewhere in BMW we noted that there was no suggestion that the plaintiff “or any other BMW purchaser was threatened with any additional potential harm” by the defendant’s conduct, we did not purport to decide the question of harm to others. Id., at 582. Rather, the opinion appears to have left the question open.

*355Respondent argues that she is free to show harm to other vietims because it is relevant to a different part of the punitive damages constitutional equation, namely, reprehensibility. That is to say, harm to others shows more reprehensible conduct. Philip Morris, in turn, does not deny that a plaintiff may show harm to others in order to demonstrate reprehensibility. Nor do we. Evidence of actual harm to nonparties can help to show that the conduct that harmed the plaintiff also posed a substantial risk of harm to the general public, and so was particularly reprehensible — although counsel may argue in a particular case that conduct resulting in no harm to others nonetheless posed a grave risk to the public, or the converse. Yet for the reasons given above, a jury may not go further than this and use a punitive damages verdict to punish a defendant directly on account of harms it is alleged to have visited on nonparties.

Given the risks of unfairness that we have mentioned, it is constitutionally important for a court to provide assurance that the jury will ask the right question, not the wrong one. And given the risks of arbitrariness, the concern for adequate notice, and the risk that punitive damages awards can, in practice, impose one State’s (or one jury’s) policies (e. g., banning cigarettes) upon other States — all of which accompany awards that, today, may be many times the size of such awards in the 18th and 19th centuries, see id., at 594-595 (Breyer, J., concurring) — it is particularly important that States avoid procedure that unnecessarily deprives juries of proper legal guidance. We therefore conclude that the Due Process Clause requires States to provide assurance that juries are not asking the wrong question, i. e., seeking, not simply to determine reprehensibility, but also to punish for harm caused strangers.

IV

Respondent suggests as well that the Oregon Supreme Court, in essence, agreed with us, that it did not authorize punitive damages awards based upon punishment for harm caused to nonparties. We concede that one might read some *356portions of the Oregon Supreme Court’s opinion as focusing only upon reprehensibility. See, e.g., 340 Ore., at 51, 127 P. 3d, at 1175 (“[T]he jury could consider whether Williams and his misfortune were merely exemplars of the harm that Philip Morris was prepared to inflict on the smoking public at large”). But the Oregon court’s opinion elsewhere makes clear that that court held more than these few phrases might suggest.

The instruction that Philip Morris said the trial court should have given distinguishes between using harm to others as part of the “reasonable relationship” equation (which it would allow) and using it directly as a basis for punishment. The instruction asked the trial court to tell the jury that “you may consider the extent of harm suffered by others in determining what [the] reasonable relationship is” between Philip Morris’ punishable misconduct and harm caused to Jesse Williams, “[but] you are not to punish the defendant for the impact of its alleged misconduct on other persons, who may bring lawsuits of their own in which other juries can resolve their claims . . . .” App. 280a (emphasis added). And as the Oregon Supreme Court explicitly recognized, Philip Morris argued that the Constitution “prohibits the state, acting through a civil jury, from using punitive damages to punish a defendant for harm to nonparties.” 340 Ore., at 51-52, 127 P. 3d, at 1175.

The court rejected that claim. In doing so, it pointed out (1) that this Court in State Farm had held only that a jury could not base its award upon “dissimilar” acts of a defendant. 340 Ore., at 52-53, 127 P. 3d, at 1175-1176. It added (2) that “[i]f a jury cannot punish for the conduct, then it is difficult to see why it may consider it at all.” Id., at 52, n. 3, 127 P. 3d, at 1175, n. 3. And it stated (3) that “[i]t is unclear to us how a jury could ‘consider’ harm to others, yet withhold that consideration from the punishment calculus.” Ibid.

The Oregon court’s first statement is correct. We did not previously hold explicitly that a jury may not punish for the *357harm caused others. But we do so hold now. We do not agree with the Oregon court’s second statement. We have explained why we believe the Due Process Clause prohibits a State’s inflicting punishment for harm caused strangers to the litigation. At the same time we recognize that conduct that risks harm to many is likely more reprehensible than conduct that risks harm to only a few. And a jury consequently may take this fact into account in determining reprehensibility. Cf., e. g., Witte v. United States, 515 U. S. 389, 400 (1995) (recidivism statutes taking into account a criminal defendant’s other misconduct do not impose an “ ‘additional penalty for the earlier crimes,’ but instead ... ‘a stiffened penalty for the latest crime, which is considered to be an aggravated offense because a repetitive one’” (quoting Gryger v. Burke, 334 U. S. 728, 732 (1948))).

The Oregon court’s third statement raises a practical problem. How can we know whether a jury, in taking account of harm caused others under the rubric of reprehensibility, also seeks to punish the defendant for having caused injury to others? Our answer is that state courts cannot authorize procedures that create an unreasonable and unnecessary risk of any such confusion occurring. In particular, we believe that where the risk of that misunderstanding is a significant one — because, for instance, of the sort of evidence that was introduced at trial or the kinds of argument the plaintiff made to the jury — a court, upon request, must protect against that risk. Although the States have some flexibility to determine what kind of procedures they will implement, federal constitutional law obligates them to provide some form of protection in appropriate cases.

V

As the preceding discussion makes clear, we believe that the Oregon Supreme Court applied the wrong constitutional standard when considering Philip Morris’ appeal. We remand this ease so that the Oregon Supreme Court can apply *358the standard we have set forth. Because the application of this standard may lead to the need for a new trial, or a change in the level of the punitive damages award, we shall not consider whether the award is constitutionally “grossly excessive.” We vacate the Oregon Supreme Court’s judgment and remand the case for further proceedings not inconsistent with this opinion.

It is so ordered.

Justice Stevens,

dissenting.

The Due Process Clause of the Fourteenth Amendment imposes both substantive and procedural constraints on the power of the States to impose punitive damages on tortfeasors. See State Farm Mut. Automobile Ins. Co. v. Campbell, 538 U. S. 408 (2003); Cooper Industries, Inc. v. Leather-man Tool Group, Inc., 532 U. S. 424 (2001); BMW of North America, Inc. v. Gore, 517 U. S. 559 (1996); Honda Motor Co. v. Oberg, 512 U. S. 415 (1994); TXO Production Corp. v. Alliance Resources Corp., 509 U. S. 443 (1993). I remain firmly convinced that the cases announcing those constraints were correctly decided. In my view the Oregon Supreme Court faithfully applied the reasoning in those opinions to the egregious facts disclosed by this record. I agree with Justice Ginsburg’s explanation of why no procedural error even arguably justifying reversal occurred at the trial in this case. See post, p. 362 (dissenting opinion).

Of greater importance to me, however, is the Court’s imposition of a novel limit on the State’s power to impose punishment in civil litigation. Unlike the Court, I see no reason why an interest in punishing a wrongdoer “for harming persons who are not before the court,” ante, at 349, should not be taken into consideration when assessing the appropriate sanction for reprehensible conduct.

Whereas compensatory damages are measured by the harm the defendant has caused the plaintiff, punitive damages are a sanction for the public harm the defendant’s con*359duct has caused or threatened. There is little difference between the justification for a criminal sanction, such as a fine or a term of imprisonment, and an award of punitive damages. See Cooper Industries, 532 U. S., at 432. In our early history either type of sanction might have been imposed in litigation prosecuted by a private citizen. See Steel Co. v. Citizens for Better Environment, 523 U. S. 83, 127-128 (1998) (Stevens, J., concurring in judgment). And while in neither context would the sanction typically include a pecuniary award measured by the harm that the conduct had caused to any third parties, in both contexts the harm to third parties would surely be a relevant factor to consider in evaluating the reprehensibility of the defendant’s wrongdoing. We have never held otherwise.

In the case before us, evidence attesting to the possible harm the defendant’s extensive deceitful conduct caused other Oregonians was properly presented to the jury. No evidence was offered to establish an appropriate measure of damages to compensate such third parties for their injuries, and no one argued that the punitive damages award would serve any such purpose. To award compensatory damages to remedy such third-party harm might well constitute a taking of property from the defendant without due process, cf. ante, at 349. But a punitive damages award, instead of serving a compensatory purpose, serves the entirely different purposes of retribution and deterrence that underlie every criminal sanction. State Farm, 538 U. S., at 416. This justification for punitive damages has even greater salience when, as in this case, see Ore. Rev. Stat. §31.735(1) (2003), the award is payable in whole or in part to the State rather than to the private litigant.1

*360While apparently recognizing the novelty of its holding, ante, at 356-357, the majority relies on a distinction between taking third-party harm into account in order to assess the reprehensibility of the defendant’s conduct — which is permitted — and doing so in order to punish the defendant “directly” — which is forbidden. Ante, at 355. This nuance eludes me. When a jury increases a punitive damages award because injuries to third parties enhanced the reprehensibility of the defendant’s conduct, the jury is by definition punishing the defendant — directly—for third-party harm.2 A murderer who kills his victim by throwing a bomb that injures dozens of bystanders should be punished more severely than one who harms no one other than his intended victim. Similarly, there is no reason why the measure of the appropriate punishment for engaging in a campaign of deceit in distributing a poisonous and addictive substance to thousands of cigarette smokers statewide should not include consideration of the harm to those “bystanders” as well as the harm to the individual plaintiff. The Court endorses a contrary conclusion without providing us with any reasoned justification.

It is far too late in the day to argue that the Due Process Clause merely guarantees fair procedure and imposes no *361substantive limits on a State’s lawmaking power. See, e. g., Moore v. East Cleveland, 431 U. S. 494, 544 (1977) (White, J., dissenting); Poe v. Ullman, 367 U. S. 497, 540-541 (1961) (Harlan, J., dissenting); Whitney v. California, 274 U. S. 357, 373 (1927) (Brandéis, J., concurring). It remains true, however, that the Court should be “reluctant to expand the concept of substantive due process because guideposts for responsible decisionmaking in this unchartered area are scarce and open-ended.” Collins v. Harker Heights, 503 U. S. 115, 125 (1992). Judicial restraint counsels us to “exercise the utmost care whenever we are asked to break new ground in this field.” Ibid. Today the majority ignores that sound advice when it announces its new rule of substantive law.

Essentially for the reasons stated in the opinion of the Supreme Court of Oregon, I would affirm its judgment.

Justice Thomas,

dissenting.

I join Justice Ginsbueg’s dissent in full. I write separately to reiterate my view that “‘the Constitution does not constrain the size of punitive damages awards.’” State Farm Mut. Automobile Ins. Co. v. Campbell, 538 U. S. 408, 429-430 (2003) (Thomas, J., dissenting) (quoting Cooper Industries, Inc. v. Leatherman Tool Group, Inc., 532 U. S. 424, 443 (2001) (Thomas, J., concurring)). It matters not that the Court styles today’s holding as “procedural” because the “procedural” rule is simply a confusing implementation of the substantive due process regime this Court has created for punitive damages. See Pacific Mut. Life Ins. Co. v. Has-lip, 499 U. S. 1, 26-27 (1991) (Scalia, J., concurring in judgment) (“In 1868 . . . punitive damages were undoubtedly an established part of the American common law of torts. It is ... clear that no particular procedures were deemed necessary to circumscribe a jury’s discretion regarding the award of such damages, or their amount”). Today’s opinion proves once again that this Court’s punitive damages jurisprudence is “insusceptible of principled application.” BMW of North *362 America, Inc. v. Gore, 517 U. S. 559, 599 (1996) (Scalia, J., joined by Thomas, J., dissenting).

Justice Ginsburg,

with whom Justice Scalia and Justice Thomas join, dissenting.

The purpose of punitive damages, it can hardly be denied, is not to compensate, but to punish. Punish for what? Not for harm actually caused “strangers to the litigation,” ante, at 353, the Court states, but for the reprehensibility of defendant’s conduct, ante, at 355. “[C]onduct that risks harm to many,” the Court observes, “is likely more reprehensible than conduct that risks harm to only a few.” Ante, at 357. The Court thus conveys that, when punitive damages are at issue, a jury is properly instructed to consider the extent of harm suffered by others as a measure of reprehensibility, but not to mete out punishment for injuries in fact sustained by nonparties. Ante, at 355-357. The Oregon courts did not rule otherwise. They have endeavored to follow our decisions, most recently in BMW of North America, Inc. v. Gore, 517 U. S. 559 (1996), and State Farm Mut. Automobile Ins. Co. v. Campbell, 538 U. S. 408 (2003), and have “deprive[d] [no jury] of proper legal guidance,” ante, at 355. Vacation of the Oregon Supreme Court’s judgment, I am convinced, is unwarranted.

The right question regarding reprehensibility, the Court acknowledges, ante, at 356, would train on “the harm that Philip Morris was prepared to inflict on the smoking public at large.” Ibid, (quoting 340 Ore. 35, 51,127 P. 3d 1165,1175 (2006)). See also id., at 55, 127 P. 3d, at 1177 (“[T]he jury, in assessing the reprehensibility of Philip Morris's actions, could consider evidence of similar harm to other Oregonians caused (or threatened) by the same conduct.” (emphasis added)). The Court identifies no evidence introduced and no charge delivered inconsistent with that inquiry.

The Court’s order vacating the Oregon Supreme Court’s judgment is all the more inexplicable considering that Philip *363Morris did not preserve any objection to the charges in fact delivered to the jury, to the evidence introduced at trial, or to opposing counsel’s argument. The sole objection Philip Morris preserved was to the trial court’s refusal to give defendant’s requested charge number 34. See id., at 54,127 P. 3d, at 1176. The proposed instruction read in pertinent part:

“If you determine that some amount of punitive damages should be imposed on the defendant, it will then be your task to set an amount that is appropriate. This should be such amount as you believe is necessary to achieve the objectives of deterrence and punishment. While there is no set formula to be applied in reaching an appropriate amount, I will now advise you of some of the factors that you may wish to consider in this connection.

“(1) The size of any punishment should bear a reasonable relationship to the harm caused to Jesse Williams by the defendant’s punishable misconduct. Although you may consider the extent of harm suffered by others in determining what that reasonable relationship is, you are not to punish the defendant for the impact of its alleged misconduct on other persons, who may bring lawsuits of their own in which other juries can resolve their claims and award punitive damages for those harms, as such other juries see fit.

“(2) The size of the punishment may appropriately reflect the degree of reprehensibility of the defendant’s conduct — that is, how far the defendant has departed from accepted societal norms of conduct.” App. 280a.

Under that charge, just what use could the jury properly make of “the extent of harm suffered by others”? The answer slips from my grasp. A judge seeking to enlighten rather than confuse surely would resist delivering the requested charge.

*364The Court ventures no opinion on the propriety of the charge proposed by Philip Morris, though Philip Morris preserved no other objection to the trial proceedings. Rather than addressing the one objection Philip Morris properly preserved, the Court reaches outside the bounds of the case as postured when the trial court entered its judgment. I would accord more respectful treatment to the proceedings and dispositions of state courts that sought diligently to adhere to our changing, less than crystalline precedent.

* * *

For the reasons stated, and in light of the abundant evidence of “the potential harm [Philip Morris’] conduct could have caused,” ante, at 354 (emphasis deleted), I would affirm the decision of the Oregon Supreme Court.