dissenting:
This appeal presents an apparently novel equal protection issue in the area of gender discrimination. The issue is whether the use of gender-based mortality tables to value reversionary interests for estate tax purposes denies female decedents the equal protection of the laws. The Government seeks to uphold this use of gender-distinct tables on two grounds: The tables do not discriminate on the basis of gender and, even if they do, the discrimination is not unconstitutional. I agree with the majority, though for different reasons, that the first argument is unavailing, but I respectfully dissent because I conclude that the second argument also fails. Both contentions merit extended discussion.
I.
Is There Gender Discrimination?
The Government contends that there has been no gender discrimination against the female decedent. There are two aspects to this contention. First, the Government points out that if the decedent had been a male, her estate would still have been liable for tax upon the trust corpus. Second, the Government argues that whether a trust is *470indudable in a decedent’s estate depends on a variety of factors — the age of the decedent at death, the age of the beneficiary at the decedent’s death, the gender of the decedent, and the gender of the beneficiary. Recognizing that there is “some reliance on male-female mortality data,” the Government maintains that there is “no disparate treatment of groups on the basis of that data” and “no classification or creation of groups solely identifiable by sex.” Brief for Appellant at 15.
Before considering these specific arguments, I acknowledge my agreement with a basic premise that underlies each of them — a government practice does not constitute gender discrimination unless it burdens a class identified solely on the basis of gender. That premise was at the core of the Supreme Court’s decision in Geduldig v. Aiello, 417 U.S. 484, 94 S.Ct. 2485, 41 L.Ed.2d 256 (1974), rejecting a gender-discrimination challenge to a state’s exclusion of pregnancy from coverage under a disability insurance program:
The lack of identity between the excluded disability and gender as such under this insurance program becomes clear upon the most cursory analysis. The program divides potential recipients into two groups — pregnant women and nonpregnant persons. While the first group is exclusively female, the second includes members of both sexes. The fiscal and actuarial benefits of the program thus accrue to members of both sexes. •
Id. at 496-97 n. 20, 94 S.Ct. at 2492 n. 20. See Personnel Administrator v. Feeney, 442 U.S. 256, 271-74, 99 S.Ct. 2282, 2291-93, 60 L.Ed.2d 870 (1979); Parham v. Hughes, 441 U.S. 347, 352-57, 97 S.Ct. 451, 50 L.Ed.2d 397 (1979). Successful equal protection challenges based on gender discrimination have involved practices that burdened a class identified solely by gender. E.g., Mississippi University for Women v. Hogan, 458 U.S. 718, 102 S.Ct. 3331, 73 L.Ed.2d 1090 (1982); Craig v. Boren, 429 U.S. 190, 99 S.Ct. 1742, 1746-48, 60 L.Ed.2d 269 (1976).
A. The effect on the decedent. The present value of the decedent’s reversion-ary interest in the trust just prior to her death, calculated on the basis of gender-distinct mortality tables, was 6.654% of the trust corpus. If the decedent had been a male, the present value of the reversionary interest, calculated in the same way, would have been 6.415% of the trust corpus. The Government argues that since both percentages are greater than the 5% threshold figure at which a trust corpus is includable in the decedent’s gross estate, there has been no gender discrimination against Charlotte Wallace; her trust was not included in her gross estate because of her gender.
That argument carries the premise of Geduldig far beyond its legitimate scope. It is one thing to say, as the Court did in Geduldig, that a practice constitutes a gender-based discrimination only when it disadvantages a class identified solely on the basis of gender. It is quite another thing to say, as the Government urges here, that a challenged practice does not disadvantage a class identified solely on the basis of gender because some members of the class, including the decedent, are not sufficiently disadvantaged to suffer adverse consequences at the precise level at which the Government has elected to attach such consequences. In some cases, the absence of adverse consequences may defeat standing for a particular plaintiff, but it does not show the absence of a gender-based discrimination. For example, if a government employer awarded five extra points to all males who took a civil service examination and hired all persons who passed, that practice would be a gender-based discrimination against females, regardless of what score the employer selected as a passing grade.
The challenged practice in this case — use of gender-distinct mortality tables — produces a disadvantage to a class identified solely on the basis of gender: The rever-sionary interest of every female decedent is valued higher than the reversionary inter*471est of every similarly situated male.1 Since the higher the present value of the rever-sionary interest, the more likely that the trust will be included in the decedent’s gross estate and therefore taxed, the existence of the 5% dividing line does not defeat the claim of gender-based discrimination. Nor is standing defeated in this case by the fact that, with gender-distinct tables in use, the decedent’s trust would have been included in her gross estate even if she had been a male. Her executor is presenting the claim that the Government has violated equal protection requirements by using gender-distinct mortality tables instead of a unisex table. Since the decedent’s trust would not have been included in her gross estate if a unisex table had been used, the executor plainly has standing to make this claim.
B. The effect on the class of female decedents. Determining whether the use of gender-distinct mortality tables is a gender-based discrimination against the class of female decedents is complicated by the formula for calculating the present value of a decedent’s reversionary interest. The formula calls for calculation of the probability that a person at the age of the decedent, just prior to death, will outlive the beneficiary of the decedent’s trust. The present value of the reversionary interest is thus a function of the life expectancies of both the decedent and the beneficiary of the trust, estimated on the basis of their ages just prior to the decedent’s death. Though it is undisputed that women as a class live on the average longer than men as a class, the use of gender-distinct mortality tables reflecting this statistical fact has conflicting effects upon the calculation of the present value of a reversionary interest because of the following circumstances. The ages and consequent life expectancies of both the decedent and the beneficiary affect that value. Whatever increases the predicted life expectancy of the trust settlor increases the likelihood that the settlor will outlive the beneficiary; on the other hand, whatever increases the predicted life expectancy of the beneficiary decreases the likelihood that the settlor will outlive the beneficiary. Gender-based tables yield higher life expectancies for females than for males at every age. Therefore, when such tables are used, the female gender of the decedent increases the present value of the reversionary interest, but the female gender of the beneficiary decreases the present value of that interest.
The situation is further complicated when one compares the outcome under gender-based tables with the outcome under a unisex table. Female decedents with male beneficiaries of their trusts will have their reversionary interests valued higher under gender-based tables than under a unisex table and thus will be more likely to have the trusts included in their gross estates. However, female decedents with female beneficiaries of their trusts will normally have their reversionary interests valued lower under gender-based tables than under a unisex table and thus will be less likely to have the trusts included in their gross estates.2 The Government contends that this assortment of results3 undercuts *472the claim that the IRS regulations discriminate on the basis of gender.
It is clear that female decedents are not uniformly worse off when gender-based tables are used than when a unisex table is used for calculating the present value of a reversionary interest. However, it is also true that the present value of the rever-sionary interest of every decedent, calculated under gender-distinct tables, is always higher when the decedent is a female than when the decedent is a male. These two facts require consideration of the issue, rarely presented in discrimination cases, whether the pertinent comparison for equal protection purposes is between the treatment of a class under the challenged practice and the treatment of that class under a neutral practice, or between the treatment of a class under the challenged practice and the treatment of all others similarly situated under the challenged practice. This issue normally does not arise in discrimination cases because a gender-based practice that disadvantages women compared to men will usually also disadvantage women compared to how they would fare under a neutral rule.4
Once the analytical issue is clearly in focus, its resolution is clear. For purposes of equal protection analysis, the pertinent inquiry in determining whether a discrimination exists is whether members of a protected class are disadvantaged by the challenged practice compared to similarly situated non-members of the class. Decisions of the Supreme Court have implicitly adopted this view, even though the precise choice we face was not explicitly posed. In Wengler v. Druggists Mutual Insurance Co., 446 U.S. 142, 100 S.Ct. 1540, 64 L.Ed.2d 107 (1980), for instance, the Court invalidated on equal protection grounds a Missouri law that denied a widower workers’ compensation death benefits on his wife’s work-related death, unless he either was mentally or physically incapacitated or proved dependence on his wife’s earnings, but granted a widow death benefits without her having to prove dependence on her husband’s earnings. The Court recognized that “the statute discriminates against both men and women.” Id. at 147, 100 S.Ct. at 1543. It discriminated against widowers vis-a-vis similarly situated widows and discriminated against female wage earners vis-a-vis similarly situated male wage earners. The Court did not reject the equal protection claim simply because under the discriminatory law, some women (widows) fared better while others (wage earners) fared worse than they would have fared under a gender-neutral law. The relevant inquiry was how women fared compared to similarly situated men. See also Califano v. Goldfarb, 430 U.S. 199, 208, 97 S.Ct. 1021, 1027, 51 L.Ed.2d 270 (1977); Wein-berger v. Wiesenfeld, 420 U.S. 636, 651-53, 95 S.Ct. 1225, 1234-35, 43 L.Ed.2d 514 (1975).
*473Since the use of gender-distinct tables causes the present value of the reversion-ary interests of all female decedents to be higher than the present value of the rever-sionary interests of all similarly situated male decedents and therefore increases the likelihood that the trusts created by female decedents will be included in their gross estates and consequently taxed, the use of these tables is a discrimination based on gender. Whether that discrimination is unconstitutional under the Equal Protection Clause is a different issue, to which I now turn.
II.
Is the Discrimination Unconstitutional?
To withstand constitutional challenge under the equal protection component of either the Fifth or Fourteenth Amendments, regulations that discriminate on the basis of gender must be “substantially related” to the promotion of “important governmental objectives.” Craig v. Boren, supra, 429 U.S. at 197, 97 S.Ct. at 456.5 In this case the Government has the burden of demonstrating that the regulations requiring use of gender-based tables in implementing section 2037 meet this standard, for, as the Supreme Court has stated, “[T]he party seeking to uphold a statute that classifies individuals on the basis of their gender must carry the burden of showing an ‘exceedingly persuasive justification’ for the classification.” Mississippi University for Women v. Hogan, supra, 458 U.S. at 724, 102 S.Ct. at 3336 (quoting Personnel Administrator v. Feeney, supra, 442 U.S. at 273, 99 S.Ct. at 2293).
The Government contends that the “important governmental objective” it seeks to promote by the use of gender-distinct mortality tables is actuarial accuracy in valuing reversionary interests of decedents’ trusts. It is the achievement of that objective to which the use of gender-distinct tables must be “substantially related.” Since “[a]s a class, women live longer than men,” Los Angeles Department of Water & Power v. Manhart, 435 U.S. 702, 704, 98 S.Ct. 1370, 1373, 55 L.Ed.2d 657 (1978), there is a surface plausibility to the Government’s contention that use of gender-distinct tables is sufficiently related to the objective of promoting actuarial accuracy to justify the gender-based discrimination inherent in such use. Analysis of the Government’s contention, however, demonstrates that the requisite justification for using gender-distinct tables is not present in the limited context illustrated by this litigation — the valuation of the reversionary interest of a decedent’s trust.
In the first place, it should be noted that the valuation for which gender-distinct tables are used in implementing the 5% cutoff of section 2037 differs significantly from most other purposes for which such tables are used. Typically gender-distinct tables are used to determine the amount of dollars that women as a class and men as a class should pay (or have paid for them by their employer) in order to fund some benefit such as a pension. In that context, the average life expectancies of those in each class determine the amount of dollars that need to be collected to fund the benefit. Whether or not separate calculations of life expectancies for a class of male beneficiaries and a class of female beneficiaries *474promote sufficient accuracy, beyond what would be achieved by a unisex table, to justify a gender-based discrimination, the calculations in that context at least determine the valuation of something real — the amount of money needed to fund the benefit. The cost of providing future benefits for a class of females is higher than the cost of promoting the same level of periodic benefits for a class of similarly situated males because the class of females will receive the benefits for a longer period of time. That is the argument that economically justifies a higher dollar payment for females than for males, see Kimball, Reverse Sex Discrimination: Manhart, 1979 Am.B.Found.Research J. 83, 106, whether or not the threshold decision to price the cost of coverage for males and females separately is sound social policy or is consistent with legal requirements, see Los Angeles Department of Water & Power v. Manhart, supra (pricing employment benefits higher for females than for males violates Title VII of the Civil Rights Act of 1964).
By contrast, the valuation for which gender-distinct tables are used under section 2037 does not concern anything of reality, but instead concerns only a theoretical construct. When section 2037 focuses attention on the present value of a decedent’s reversionary interest in a trust the moment before death, the statute is not concerned with the actual fair market value of that interest. In most cases (all in which the decedent succumbs to a last illness) the actual market value of such an interest will be virtually zero. See Estate of Allen, 558 F.2d 14, 21, 214 Ct.Cl. 630 (1977) (Nichols, J., concurring). Unless a decedent dies from sudden and totally unanticipated causes, his realistic life expectancy just prior to death, as distinguished from the average life expectancy of all persons of his age, is too brief to give any real value to his reversionary interest. Moreover, the present value just prior to death is calculated only for those persons who die, a circumstance that eliminates even a theoretical present value for their rever-sionary interest. The Government recognizes this circumstance:
In all cases where IRC § 2037 comes into play, the grantor has in fact predeceased the beneficiary. In reality, therefore, the decedent’s reversionary interest in a Section 2037 situation has no real value, since reversion can no longer occur. What IRC § 2037 does is provide for the valuation of the reversionary interest hypothetically — without regard to the fact of the decedent’s death.
Appellant’s Brief at 34. What is being measured in this case is the theoretical value of an interest that the Government concedes has no value in reality. It is difficult to see what important governmental objective is being substantially advanced by using a gender-based discrimination to place a theoretical value on a rever-sionary interest whose actual value is almost always zero.
I recognize, of course, that Congress had a legitimate reason for wanting the I.R.S. to determine the theoretical value of a re-versionary interest just prior to a decedent’s death. The reason is tax equity. Prior to the enactment of section 2037, a trust was included in the gross estate no matter how remote the likelihood that the corpus would ever have reverted to the decedent during his lifetime. See Estate of Spiegel v. Commissioner, 335 U.S. 701, 707, 69 S.Ct. 301, 303-04, 93 L.Ed. 330 (1949). Congress adopted the 5% cutoff rule of section 2037 to prevent taxation of trusts in which the interest of the decedent, just prior to death, valued on the basis of the average life expectancy for persons of the decedent’s age, was too slight to warrant taxation of the entire trust corpus. See 1949 U.S.Code Cong.Serv., 81st Cong., 1st Sess., 2172, 2185; Estate of Allen, supra, 558 F.2d 17-18 & n. 8.
Thus the inquiry becomes whether tax equity will be substantially advanced by valuing reversionary interests, for purposes of applying the 5% cutoff rule, by means of gender-distinct tables. One important indication that tax equity will not be significantly promoted by gender-distinct tables is the recent decision of the Treasury to abandon the use of such ta*475bles. Effective December 1, 1983, the Treasury resumed the use of a unisex mortality table, returning to the practice that had been followed prior to 1970. See Treas.Reg. § 20.2031-7(a). I do not suggest that taxation is a field requiring the Government to use the least restrictive means in accomplishing its legitimate objectives. And what the Government elects to do is of course not the determinant of what the Constitution minimally requires it to do. Nevertheless, in the inquiry as to whether gender-distinct tables “substantially” promote an “important” governmental objective, the fact that the Government has come to realize that it can adequately meet its objectives without the use of the challenged discrimination is at least relevant.
Finally, I examine the degree to which the use of gender-distinct tables achieves the asserted interest in promoting the accuracy of even the theoretical valuation being made. This requires consideration of what we mean when we say that women as a class live longer than men. We surely do not mean that all women live longer than all men. We do not even mean that most women live longer than most men. One thing we do mean has been expressed by those knowledgeable in the field in the following statistical terms: A woman at the mean of the distribution of mortality for all women will live longer than a man at the mean of the distribution of mortality for all men. See Brilmayer, Hekeler, Lay-cock & Sullivan, Sex Discrimination in Employer-Sponsored Insurance Plans: A Legal and Demographic Analysis, 47 U.Chi.L.Rev. 505, 513 (1980) (“Demographic Analysis”). That observation, however, tells us little about the degree of accuracy achieved by gender-distinct tables compared to a unisex table. More helpful is the fact that, for a group of people chosen at random, more than 80% of all female deaths can be matched with the deaths of males of the same age.6 Demographic Analysis, supra, at 531. It was this fact that led Judge Stewart to conclude in this case that “only 20% of all women actually do outlive men.”7 576 F.Supp. at 842.
Gender-distinct tables uniformly estimate a longer life expectancy for all women compared to all men when in fact that estimate turns out to be true for no more than 20% of women.8 Whether or not the enhanced degree of accuracy in predicting the life expectancy of women as a class would satisfy equal protection standards in a case involving the purchase of retirement benefits, which must be priced prospectively, gender-based tables do not provide a sufficient increment of accuracy to justify the gender-based discrimination inherent in their use when applied to the theoretical valuation of a reversionary interest in a trust just prior to a decedent’s death, a valuation that is always made retrospectively. Cf. Craig v. Boren, supra, 429 U.S. at 199-204, 97 S.Ct. at 457-60 (invalidat*476ing discriminatory drinking age statute on grounds of insufficient statistical correlation between alcohol-related traffic offenses and gender).
The majority opinion, in contending that the discrimination involved in this case is not unlawful, rests heavily on the assertion that “the class of all directly affected women” has not been shown to be disadvantaged “compared to the class of all directly affected men.” 775 F.2d at 466. The majority makes this assertion after noting the indisputable fact that all other things being equal, under gender-distinct tables, the value of the reversionary interest when the beneficiary is a female is always lower than the value when the beneficiary is a male. From this fact the majority draws what I believe is the totally unsupportable conclusion that “the IRS practice places female beneficiaries in better positions than male beneficiaries.” Id. at 466. To appreciate the flaw in the majority’s argument, one must recognize that the “beneficiary” whose gender affects the value of the reversionary interest is not necessarily the same person who benefits from the tax saving in the event the trust is not included in the taxable estate. The “beneficiary” whose gender affects the value of the re-versionary interest is the second life tenant of the trust. The identity of the “beneficiaries” of a reduction in estate taxes depends on how the estate tax attributable to the trust corpus is allocated among all persons who would benefit from property of the taxable estate, who may be male, female, or neuter in the case of charities. The second life tenant benefits from the non-inclusion of the corpus in the decedent’s taxable estate only in those circumstances where the trust corpus would have borne all or a portion of the estate taxes thereby saved. The most that can be said is that changing the gender of the successor life tenant beneficiary from male to female tends to reduce the possibility of an estate tax on the value of the trust corpus, a consequence that will benefit all the beneficiaries of the estate whose interests would have borne the taxes that are saved and, in some circumstances, will benefit the successor life tenant of the trust. But there is no subclass of female beneficiaries that always comes out better.
Moreover, even if female beneficiaries sometimes fare better under gender-distinct tables (as in those instances when they happen to be included among the legatees who benefit from reduced estate taxes), that consequence does not eliminate the valid complaint of all female settlors that their gender always results in a higher valuation of the reversionary interest than would occur if these settlors had been males. The focus of an inquiry into whether a gender discrimination adversely affects all class members must be confined to the similarly situated members of that gender. A civil service scoring system that reduces the test scores of all female candidates by five points would not be saved by a proviso that the daughter of any applicant would be entitled to a five-point bonus.
The significance of the dispute in this case has been substantially reduced by the Government’s voluntary abandonment of gender-distinct tables to value reversionary interests. The consideration of gender-distinct tables, however, will be with us for many years to come. In this case, for all of the reasons set forth, I respectfully dissent.
. For purposes of this case, a male and a female decedent are similarly situated if they are the same age and the beneficiaries of their trusts are of the same age and gender.
. In those infrequent instances where the beneficiary is older than the decedent (and both are female), the reversionary interest of the decedent will be valued lower under a unisex table than under gender-based tables. For an older female beneficiary, the switch to a unisex table yields a lesser reduction in the life expectancy of a younger female decedent, thereby decreasing the likelihood that a person of the decedent's age would have survived the beneficiary.
. The variety of possible results is illustrated in the following table, which shows the percent to be used for calculating the present value of a reversionary interest of a trust created by a person who died at Charlotte’s age with a beneficiary at Howard’s age:
Percent Resulting From Decedent_Beneficiary Gender-Distinct Tables
Female aged 88 Male aged 57 6.654%
Male aged 88 Male aged 57 6.415%
Female aged 88 Female aged 57 3.520%
Male aged 88 Female aged 57 3.390%
*472 Percent Resulting From Decedent Beneficiara Unisex Table
Female aged 88 Male aged 57 4.9867%
Male aged 88 Male aged 57 4.9867%
Female aged 88 Female aged 57 4.9867%
Male aged 88 Female aged 57 4.9867%
The table illustrates the comparisons noted in the text. Under gender-based tables, the present value percentage is always higher for a female decedent than for a similarly situated male decedent (6.654% versus 6.415%; 3.520% versus 3.390%). Under such tables, the present value percentage is always lower when the beneficiary is female than when the beneficiary is male (3.520% versus 6.654%; 3.390% versus 6.415%). The present value percentage is always higher for a female decedent with a male beneficiary under gender-based tables than under a unisex table (6.654% versus 4.9867%). The present value percentage is always lower for a female decedent with a younger female beneficiary under gender-based tables than under a unisex table (3.520% versus 4.9867%).
. This case is distinctive, if not unique, because the effects of the gender-based mortality tables enter the calculation of the present value of the reversionary interest twice — once in estimating the life expectancy of the decedent, just before death, and again in estimating the life expectancy of the beneficiary, and because those gender-based effects are augmented or diminished depending on the combination of genders of the two persons involved. When decedents and beneficiaries are of the same gender, the effects tend to cancel each other out; when the genders are different, the effects are amplified.
. The Government argues that, despite the fact that the regulations requiring gender-distinct tables discriminate against women, they should be subjected only to "minimal scrutiny” because they are tax regulations. In support of this contention, the Government cites primarily equal protection challenges to tax regulations that did not involve gender discrimination. E.g., Regan v. Taxation With Representation, 461 U.S. 540, 103 S.Ct. 1997, 76 L.Ed.2d 129 (1983); Madden v. Kentucky, 309 U.S. 83, 60 S.Ct. 406, 84 L.Ed. 590 (1940); Burke Mountain Academy, Inc. v. United States, 715 F.2d 779 (2d Cir.1983). This authority is unpersuasive. Even Kahn v. Shevin, 416 U.S. 351, 94 S.Ct. 1734, 40 L.Ed.2d 189 (1974), which involved gender discrimination, is not dispositive. Unlike Kahn, federalism does not require deference to state tax policy in the instant case, and, unlike the statute in Kahn, the Treasury regulations at issue here do not employ gender discrimination in order to correct the effects of past gender discrimination. There is no reason to exempt tax regulations from the standard of scrutiny applicable in all other gender discrimination cases.
. The statistic is based on an analysis of a group of males and females aged 65. Bergmann & Gray, Equality in Retirement Benefits, Civ. Rights Dig., Fall 1975, at 25. The accuracy, though not the significance, of the statistic has been widely accepted. Compare Demographic Analysis, supra, at 531 (arguing in reliance on the statistic), with Kimball, Reverse Sex Discrimination: Manhart, supra, at 121-22 (accepting the statistic but characterizing the argument from it as a “fallacy” when applied to benefits that must be priced prospectively). For a group randomly selected from the entire population, without regard to age, the extent to which male and female deaths can be matched is probably somewhat lower than the figure reported from the group aged 65.
. The 20% figure appears to be the maximum possible consequence of the fact that 80% of male deaths can be matched with female deaths at the same age. Though most of the unmatched women die later than most of the unmatched men (a fact reflected in the higher average life expectancies for women as a class), there are undoubtedly some of the unmatched men who live longer than some of the unmatched women. The true percentage of women who outlive men is probably slightly less than 20%.
. Even this slight increase in predictive accuracy is brought into question by the tendency for the differential between male and female life expectancies to change rapidly. Demographic Analysis, supra, at 558; Brilmayer, Laycock & Sullivan, The Efficient Use of Group Averages as Nondiscrimination: A Rejoinder to Professor Benston, 50 U.Chi.L.Rev. 222, 236 (1983) (“Such an unstable [differential] means that sex is not even a reasonable predictor of the longevity of groups.”).